research article nonzero-sum stochastic differential game...
TRANSCRIPT
![Page 1: Research Article Nonzero-Sum Stochastic Differential Game ...downloads.hindawi.com/journals/aaa/2013/761306.pdf · Controller and Stopper for Jump Diffusions YanWang, 1,2 AiminSong,](https://reader033.vdocuments.mx/reader033/viewer/2022042410/5f26fe8f7005a2296769470b/html5/thumbnails/1.jpg)
Hindawi Publishing CorporationAbstract and Applied AnalysisVolume 2013 Article ID 761306 7 pageshttpdxdoiorg1011552013761306
Research ArticleNonzero-Sum Stochastic Differential Game betweenController and Stopper for Jump Diffusions
Yan Wang12 Aimin Song2 Cheng-De Zheng2 and Enmin Feng1
1 School of Mathematical Sciences Dalian University of Technology Dalian 116023 China2 School of Science Dalian Jiaotong University Dalian 116028 China
Correspondence should be addressed to Yan Wang wymath163com
Received 5 February 2013 Accepted 7 May 2013
Academic Editor Ryan Loxton
Copyright copy 2013 Yan Wang et al This is an open access article distributed under the Creative Commons Attribution Licensewhich permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited
We consider a nonzero-sum stochastic differential game which involves two players a controller and a stopper The controllerchooses a control process and the stopper selects the stopping rule which halts the gameThis game is studied in a jump diffusionssetting within Markov control limit By a dynamic programming approach we give a verification theorem in terms of variationalinequality-Hamilton-Jacobi-Bellman (VIHJB) equations for the solutions of the game Furthermore we apply the verificationtheorem to characterize Nash equilibrium of the game in a specific example
1 Introduction
In this paper we study a nonzero-sum stochastic differentialgame with two players a controller and a stopper The state119883(sdot) in this game evolves according to a stochastic differentialequation driven by jump diffusionsThe controller affects thecontrol process 119906(sdot) in the drift and volatility of119883(sdot) at time 119905and the stopper decides the duration of the game in the formof a stopping rule 120591 for the process119883(sdot) The objectives of thetwo players are to maximize their own expected payoff
In order to illustrate the motivation and the backgroundof application for this game we show a model in finance
Example 1 Let (ΩF F119905119905ge0 119875) be a filtered probability
space let 119861(119905) be a 119896-dimensional Brownian Motion and let(119889119905 119889119911) = (
1(119889119905 119889119911)
119896(119889119905 119889119911)) be 119896-independent
compensated Poisson randommeasures independent of 119861(119905)119905 isin [0infin) For 119894 = 1 119896
119894(119889119905 119889119911) = 119873
119894(119889119905 119889119911)minus]
119894(119889119911)119889119905
where ]119894is the Levy measure of a Levy process 120578
119894(119905) with
jump measure 119873119894such that 119864[1205782
119894(119905)] lt infin for all 119905 F
119905119905ge0
is the filtration generated by 119861(119905) and (119889119905 119889119911) (as usualaugmentedwith all the119875-null sets)We refer to [1 2] formoreinformation about Levy processes
We firstly define a financial market model as followsSuppose that there are two investment possibilities
(1) a risk-free asset (eg a bond) with unit price 1198780(119905) at
time 119905 given by1198891198780(119905) = 119903 (119905) 119878
0(119905) 119889119905 119878
0(0) = 1 (1)
(2) a risky asset (eg a stock) with unit price 1198781(119905) at time
119905 given by
1198891198781(119905)
= 1198781(119905minus) [119887 (119905) 119889119905 + 120590 (119905) 119889119861 (119905) + int
R
120574 (119905 119911) (119889119905 119889119911)]
1198781(0) gt 0
(2)
where 119903(119905) is F119905-adapted with int
119879
0|119903(119905)|119889119905 lt infin as 119879 gt 0
is a fixed given constant 119887 120590 120574 is F119905-predictable processes
satisfying 120574(119905 119911) gt minus1 for aa 119905 119911 as and
int
119879
0
|119887 (119905)| + 1205902(119905) + int
R
1205742(119905 119911) ] (119889119911) 119889119905 lt infin forall119879 lt infin
(3)
2 Abstract and Applied Analysis
Assume that an investor hires a portfolio manager tomanage his wealth 119883(119905) from investments The manager(controller) can choose a portfolio 119906(119905) which represents theproportion of the total wealth 119883(119905) invested in the stock attime 119905 And the investor (stopper) can halt the wealth process119883(119905) by selecting a stop-rule 120591 119862[0 119879] rarr [0 119879] Then thedynamics of the corresponding wealth process 119883(119905) = 119883
119906(119905)
is
119889119883119906(119905) = 119883
119906(119905minus) [(1 minus 119906 (119905)) 119903 (119905) + 119906 (119905) 119887 (119905)] 119889119905
+ 119906 (119905) 120590 (119905) 119889119861 (119905)
+119906 (119905minus) intR
120574 (119905 119911) (119889119905 119889119911)
119883119906(0) = 119909 gt 0
(4)
(see eg [3ndash5])We require that 119906(119905minus)120574(119905 119911) gt minus1 for aa 119905 119911as and that
int
119879
0
|(1 minus 119906 (119905)) 119903 (119905)| + |119906 (119905) 119887 (119905)| + 1199062(119905) 1205902(119905)
+1199062(119905) int
R
1205742(119905 119911) ] (119889119911) 119889119905 lt infin as
(5)
At terminal time 120591 the stopper gives the controller apayoff 119862(119883(120591)) where 119862 119862[0 119879] rarr R is a deterministicmapping Therefore the controller aims to maximize hisutility of the following form
J119909
1(119906 120591)=119864
119909[119890minus120575120591
1198801(119862 (119883 (120591)))minusint
120591
0
119890minus120575119905ℎ (119905 119883 (119905) 119906
119905) 119889119905]
(6)
where 120575 gt 0 is the discounting rate ℎ is a cost function and1198801is the controllerrsquos utility We denote 119864119909 the expectation
with respect to119875119909 and119875119909 the probability laws of119883(119905) startingat 119909
Meanwhile it is the stopperrsquos objective to choose thestopping time 120591 such that his own utility
J119909
2(119906 120591) = 119864
119909[119890minus120575120591
1198802119883 (120591) minus 119862 (119883 (120591))] (7)
is maximized where 1198802is the stopperrsquos utility
As this game is typically a nonzero sum we seek a Nashequilibrium namely a pair (119906lowast 120591lowast) such that
J119909
1(119906lowast 120591lowast) ge J
119909
1(119906 120591lowast) forall119906
J119909
2(119906lowast 120591lowast) ge J
119909
2(119906lowast 120591) forall120591
(8)
This means that the choice 119906lowast is optimal for the controllerwhen the stopper uses 120591lowast and vice verse
The game (4) and (8) are a nonzero-sum stochasticdifferential game between a controller and a stopper Theexistence of Nash equilibrium shows that by an appropriatestopping rule design 120591lowast the stopper can induce the controllerto choose the best portfolio he can Similarly by applying asuitable portfolio 119906lowast the controller can force the stopper tostop the employment relationship at a time of the controllerrsquoschoosing
There have been significant advances in the researchof stochastic differential games of control and stoppingFor example in [6ndash9] the authors considered the zero-sumstochastic differential games ofmixed type with both controlsand stopping between two players In these games each of theplayers chooses an optimal strategy which is composed by acontrol 119906(sdot) and a stopping 120591 Under appropriate conditionsthey constructed a saddle pair of optimal strategies For thenonsum case the games of mixed type were discussed in[6 10] The authors presented Nash equilibria rather thansaddle pairs of strategies [10] Moreover the papers [11ndash17]considered a zero-sum stochastic differential game betweencontroller and stopper where one player (controller) choosesa control process 119906(sdot) and the other (stopper) chooses astopping 120591 One player tries to maximize the reward and theother to minimize it They presented a saddle pair for thegame
In this paper we study a nonzero-sum stochastic differen-tial game between a controller and a stopper The controllerand the stopper have different payoffsThe objectives of themare tomaximize their own payoffsThis game is considered ina jumpdiffusion context under theMarkov control conditionWe prove a verification theorem in terms of VIHJB equationsfor the game to characterize Nash equilibrium
Our setup and approach are related to [3 17] Howevertheir games are different fromours In [3] the authors studiedthe stochastic differential games between two controllersThe work in [17] was carried out for a zero-sum stochasticdifferential game between a controller and a stopper
The paper is organized as follows in the next sectionwe formulate the nonzero-sum stochastic differential gamebetween a controller and a stopper and prove a generalverification theorem In Section 3 we apply the generalresults obtained in Section 2 to characterize the solutions of aspecial game Finally we conclude this paper in Section 4
2 A Verification Theorem forNonzero-Sum Stochastic Differential Gamebetween Controller and Stopper
Suppose the state 119884(119905) = 119884119906(119905) at time 119905 is given by the
following stochastic differential equation
119889119884 (119905) = 120572 (119884 (119905) 1199060(119905)) 119889119905 + 120573 (119884 (119905) 119906
0(119905)) 119889119861 (119905)
+ intR1198960
120579 (119884 (119905minus) 1199061(119905minus 119911) 119911) (119889119905 119889119911)
119884 (0) = 119910 isin R119896
(9)
where 120572 R119896 times U rarr R119896 120573 R119896 times U rarr R119896times119896 120579 R119896 times
U times R119896 rarr R119896times119896 are given functions and U denotes a givensubset of R119901
We regard 1199060(119905) = 119906
0(119905 120596) and 119906
1(119905 119911) = 119906
1(119905 119911 120596)
as the control processes assumed to be cadlag F119905-adapted
and with values in 1199060(119905) isin U 119906
1(119905 119911) isin U for aa 119905 119911 120596
And we put 119906(119905) = (1199060(119905) 1199061(119905 119911)) Then 119884(119905) = 119884
(119906)(119905) is
a controlled jump diffusion (see [18] for more informationabout stochastic control of jump diffusion)
Abstract and Applied Analysis 3
Fix an open solvency set S sub R119896 Let
120591S = inf 119905 gt 0 119884 (119905) notin S (10)
be the bankruptcy time 120591S is the first time at which thestochastic process 119884(119905) exits the solvency set S Similaroptimal control problems in which the terminal time isgoverned by a stopping criterion are considered in [19ndash21] inthe deterministic case
Let 119891119894 R119896 times 119870 rarr R and 119892
119894 R119896 rarr R be given
functions for 119894 = 1 2 Let A be a family of admissiblecontrols contained in the set of 119906(sdot) such that (9) has a uniquestrong solution and
119864119910[int
120591S
0
1003816100381610038161003816119891119894 (119884119905 119906119905)1003816100381610038161003816 119889119905] lt infin 119894 = 1 2 (11)
for all 119910 isin S where 119864119910 denotes expectation given that119884(0) =119910 isin R119896 Denote by Γ the set of all stopping times 120591 le 120591SMoreover we assume that
the family 119892minus
119894(119884120591) 120591 isin Γ is uniformly integrable
for 119894 = 1 2
(12)
Then for 119906 isin A and 120591 isin Γ we define the performancefunctionals as follows
J119910
119894(119906 120591) = 119864
119910[int
120591
0
119891119894(119884 (119905) 119906 (119905)) 119889119905 + 119892
119894(119884 (120591))]
119894 = 1 2
(13)
We interpret 119892119894(119884(120591)) as 0 if 120591 = infin We may regardJ
119910
1(119906 120591)
as the payoff to the controller who controls 119906 andJ1199102(119906 120591) as
the payoff to the stopper who decides 120591
Definition 2 (nash equilibrium) A pair (119906lowast 120591lowast) isin A times Γ iscalled a Nash equilibrium for the stochastic differential game(9) and (13) if the following holds
J119910
1(119906lowast 120591lowast) ge J
119910
1(119906 120591lowast) forall119906 isin U 119910 isin S (14)
J119910
2(119906lowast 120591lowast) ge J
119910
2(119906lowast 120591) forall120591 isin Γ 119910 isin S (15)
Condition (14) states that if the stopper chooses 120591lowast it isoptimal for the controller to use the control 119906lowast Similarlycondition (15) states that if the controller uses 119906lowast it is optimalfor the stopper to decide 120591lowast Thus (119906lowast 120591lowast) is an equilibriumpoint in the sense that there is no reason for each individualplayer to deviate from it as long as the other player does not
We restrict ourselves to Markov controls that is weassume that 119906
0(119905) =
0(119884(119905)) 119906
1(119905 119911) =
1(119884(119905) 119911) As
customary we do not distinguish between 1199060and
0 1199061and
1 Then the controls 119906
0and 119906
1can simply be identified with
functions 0(119910) and
1(119910 119911) where 119910 isin S and 119911 isin R119896
When the control 119906 is Markovian the correspondingprocess 119884(119906)(119905) becomes a Markov process with the generator119860119906 of 120601 isin 1198622(R119896) given by
119860120601 (119910) =
119896
sum
119894=1
120572119894(119910 1199060(119910))
120597120601
120597119910119894
(119910)
+1
2
119896
sum
119894119895=1
(120573120573119879)119894119895(119910 1199060(119910))
1205972120601
120597119910119894120597119910119895
(119910)
+
119896
sum
119895=1
intR
120601 (119910 + 120579119895(119910 1199061(119910 119911) 119911)) minus 120601 (119910)
minusnabla120601 (119910) 120579119895(119910 1199061(119910 119911) 119911) ] (119889119911)
(16)
where nabla120601 = (1205971206011205971199101 120597120601120597119910
119896) is the gradient of 120601 and 120579119895
is the 119895th column of the 119896 times 119896matrix 120579Now we can state the main result of this section
Theorem 3 (verification theorem for game (9) and (13))Suppose there exist two functions 120601
119894 S rarr R 119894 = 1 2 such
that
(i) 120601119894isin 1198621(S0)⋂119862(S) 119894 = 1 2 whereS is the closure of
S and S0 is the interior of S(ii) 120601119894ge 119892119894on S 119894 = 1 2
Define the following continuation regions
119863119894= 119910 isin S 120601
119894(119910) gt 119892
119894(119910) 119894 = 1 2 (17)
Suppose 119884(119905) = 119884119906(119905) spends 0 time on 120597119863
119894as 119894 =
1 2 that is(iii) 119864119910[int120591119878
0120594120597119863119894(119884(119905))119889119905] = 0 for all 119910 isin S 119906 isin U 119894 = 1 2
(iv) 120597119863119894is a Lipschitz surface 119894 = 1 2
(v) 120601119894isin 1198622(S120597119863
119894)The second-order derivatives of 120601
119894are
locally bounded near 120597119863119894 respectively 119894 = 1 2
(vi) 1198631= 1198632= 119863
(vii) there exists isin A such that for 119894 = 1 2
119860120601119894(119910) + 119891
119894(119910 (119910))
= sup119906isinA
119860119906120601119894(119910) + 119891
119894(119910 119906 (119910))
= 0 119910 isin 119863
le 0 119910 isin S 119863
(18)
(viii) 119864119910[|120601119894(119884120591)| + int120591
0|119860119906120601119894(119884119906(119905))|119889119905] lt infin 119894 = 1 2 for all
119906 isin U 120591 isin ΓFor 119906 isin A define
120591119863= 120591119906
119863= inf 119905 gt 0 119884
119906(119905) notin 119863 lt infin (19)
and in particular
120591119863= 120591
119863= inf 119905 gt 0 119884
(119905) notin 119863 lt infin (20)
4 Abstract and Applied Analysis
Suppose that(ix) the family 120601
119894(119884(120591)) 120591 isin Γ 120591 le 120591
119863 is uniformly
integrable for all 119906 isin A and 119910 isin S 119894 = 1 2
Then (120591119863 ) isin Γ timesA is a Nash equilibrium for game (9) (13)
and
1206011(119910) = sup
119906isinA
J119906120591119863
1(119910) = J
120591119863
1(119910) (21)
1206012(119910) = sup
120591isinΓ
J120591
2(119910) = J
120591119863
2(119910) (22)
Proof From (i) (iv) and (v) we may assume by an approx-imation theorem (see Theorem 21 in [18]) that 120601
119894isin 1198622(S)
119894 = 1 2For a given 119906 isin A we define with 119884(119905) = 119884
119906(119905)
120591119863= 120591119906
119863= inf 119905 gt 0 119884
119906(119905) notin 119863 (23)
In particular let be as in (vii) Then
120591119863= 120591
119863= inf 119905 gt 0 119884
(119905) notin 119863 (24)
We first prove that (21) holds Let 120591119863isin Γ be as in (24) For
arbitrary 119906 isin A by (vii) and the Dynkinrsquos formula for jumpdiffusion (see Theorem 124 in [18]) we have
1206011(119910) = 119864
119910[int
120591119863and119898
0
minus1198601199061206011(119884 (119905)) 119889119905 + 120601
1(119884 (120591119863and 119898))]
ge 119864119910[int
120591119863and119898
0
1198911(119884 (119905) 119906 (119905)) 119889119905 + 120601
1(119884 (120591119863and 119898))]
(25)
where119898 = 1 2 Therefore by (11) (12) (i) (ii) (viii) andthe Fatou lemma
1206011(119910)
ge lim inf119898rarrinfin
119864119910[int
120591119863and119898
0
1198911(119884 (119905) 119906 (119905)) 119889119905+120601
1(119884 (120591119863and 119898))]
ge 119864119910[int
120591119863
0
1198911(119884 (119905) 119906 (119905)) 119889119905 + 119892
1(119884 (120591119863))]
= J119910
1(119906 120591119863)
(26)
Since this holds for all 119906 isin A we have
1206011(119910) ge sup
119906isinA
J119910
1(119906 120591119863) (27)
In particular applying the Dynkinrsquos formula to 119906 = we getan equality that is
1206011(119910) = 119864
119910[int
120591119863and119898
0
minus1198601206011( (119905)) 119889119905 + 120601
1( (120591119863and 119898))]
= 119864119910[int
120591119863and119898
0
1198911( (119905) (119905)) 119889119905 + 120601
1( (120591119863and 119898))]
(28)
where (119905) = 119884(119905) and119898 = 1 2 Hence we deduce that
1206011(119910) = J
119910
1( 120591119863) (29)
Since we always have
J119910
1( 120591119863) le sup119906isinA
J119910
1(119906 120591119863) (30)
we conclude by combining (27) (29) and (30) that
1206011(119910) = J
119910
1( 120591119863) = sup119906isinA
J119910
1(119906 120591119863) (31)
which is (21)Next we prove that (22) holds Let isin A be as in (vii)
For 120591 isin Γ by the Dynkinrsquos formula and (vii) we have
119864119910[1206012( (120591119898))] = 120601
2(119910) + 119864
119910[int
120591119898
0
1198601206012( (119905)) 119889119905]
le 1206012(119910) minus 119864
119910[int
120591119898
0
1198912( (119905) (119905)) 119889119905]
(32)
where 120591119898= 120591 and 119898119898 = 1 2
Letting119898 rarr infin gives by (11) (12) (i) (ii) (viii) and theFatou Lemma
1206012(119910) ge lim inf
119898rarrinfin119864119910[int
120591and119898
0
1198912( (119905) (119905)) 119889119905 + 120601
2( (120591119898))]
ge 119864119910[int
120591
0
1198912( (119905) (119905)) 119889119905 + 119892
2( (120591)) 120594
120591ltinfin]
= J119910
2( 120591)
(33)
The inequality (33) holds for all 120591 isin Γ Then we have
1206012(119910) ge sup
120591isinΓ
J119910
2( 120591) (34)
Similarly applying the above argument to the pair ( 120591119863) we
get an equality in (34) that is
1206012(119910) = J
119910
2( 120591119863) (35)
We always have
J119910
2( 120591119863) le sup120591isinΓ
J119910
2( 120591) (36)
Therefore combining (34) (35) and (36) we get
1206012(119910) = J
119910
2( 120591119863) = sup120591isinΓ
J119910
2( 120591) (37)
which is (22) The proof is completed
Abstract and Applied Analysis 5
3 An Example
In this section we come back to Example 1 and useTheorem 3to study the solutions of game (4) and (8) Here and in thefollowing all the processes are assumed to be one-dimensionfor simplicity
To put game (4) and (8) into the framework of Section 2we define the process 119884(119905) = (119884
0(119905) 1198841(119905)) 119884(0) = 119910 = (119904 119910
1)
by
1198891198840(119905) = 119889119905 119884
0(0) = 119904 isin R
1198891198841(119905) = 119889119883 (119905)
= 1198841(119905) [(1 minus 119906 (119905)) 119903 (119905) + 119906 (119905) 119887 (119905)] 119889119905
+ 119906 (119905) 120590 (119905) 119889119861 (119905)
+119906 (119905minus) intR
120574 (119905 119911) (119889119905 119889119911)
1198841(0) = 119909 = 119910
1
(38)
Then the performance functionals to the controller (6) andthe stopper (7) can be formulated as follows
J119910
1(119906 120591) = 119864
1199041199101 [119890minus120575(119904+120591)
1198801(119862 (1198841(120591)))
minusint
120591
0
119890minus120575(119904+119905)
ℎ (1198840(119905) 1198841(119905) 119906 (119905)) 119889119905]
J119910
2(119906 120591) = 119864
1199041199101 [119890minus120575(119904+120591)
1198802(1198841(120591) minus 119862 (119884
1(120591)))]
(39)
In this case the generator 119860119906 in (16) has the form
119860119906120601 (119904 119909)
=120597120601
120597119904+ [(1 minus 119906) 119903119909 + 119906119887119909]
120597120601
120597119909+1
2119909211990621205902 1205972120601
1205971199092
+ intR0
120601 (119904 119909 + 119909119906120574 (119911)) minus 120601 (119904 119909) minus 119909119906120574 (119911)120597120601
120597119909
times ] (119889119911)
(40)
To obtain a possible Nash equilibrium ( 120591) isin A times Γ forgame (4) and (8) according to Theorem 3 it is necessary tofind a subset 119863 of S = R2
+= [0infin)
2 and 120601119894(119904 119909) 119894 = 1 2
such that
(i) 1206011(119904 119909) = 119890
minus1205751199041198801(119862(119909)) and 120601
2(119904 119909) = 119890
minus1205751199041198802(119909 minus
119862(119909)) for all (119904 119909) isin 119863(ii) 1206011(119904 119909) ge 119890
minus1205751199041198801(119862(119909)) and 120601
2(119904 119909) ge 119890
minus1205751199041198802(119909 minus
119862(119909)) for all (119904 119909) isin S(iii) 119860119906120601
1(119904 119909) minus ℎ(119904 119909 119906) le 0 and 119860119906120601
2(119904 119909) le 0 for all
(119904 119909) isin S 119863 and 119906(iv) there exists such that 119860120601
1(119904 119909) minus ℎ(119904 119909 ) =
1198601206012(119904 119909) = 0 for all (119904 119909) isin 119863
Imposing the first-order condition on 1198601199061206011(119904 119909) minus
ℎ(119904 119909 119906) and 1198601199061206012(119904 119909) we get the following equations for
the optimal control processes
(119887119909 minus 119903119909)1205971206011
120597119909(119904 119909) + 119909
2120590212059721206011
1205971199092(119904 119909)
+ intR
119903120574 (119911) [1205971206011
120597119909(119904 119909 + 119903120574 (119911)) minus
1205971206011
120597119909(119904 119909)]
minus120597ℎ
120597119906(119904 119909 ) = 0
(119887119909 minus 119903119909)1205971206012
120597119909(119904 119909) + 119909
2120590212059721206012
1205971199092(119904 119909)
+ intR
119903120574 (119911) [1205971206012
120597119909(119904 119909 + 119903120574 (119911)) minus
1205971206012
120597119909(119904 119909)] = 0
(41)
With as in (41) we put
119860120601119894(119904 119909)
=120597120601119894
120597119904+ [(1 minus ) 119903119909 + 119887119909]
120597120601119894
120597119909+1
2119909221205902 1205972120601119894
1205971199092
+ intR0
120601119894(119904 119909 + 119909120574 (119911)) minus 120601
119894(119904 119909) minus 119909120574 (119911)
120597120601119894
120597119909
times ] (119889119911)
(42)
Thus we may reduce game (4) and (8) to the problem ofsolving a family of nonlinear variational-integro inequalitiesWe summarize as follows
Theorem 4 Suppose there exist satisfying (41) and two 1198621-functions 120601
119894 119894 = 1 2 such that
(1)
119863 = (119904 119909) 1206012(119904 119909) gt 119890
minus1205751199041198802(119909 minus 119862 (119909))
= (119904 119909) 1206011(119904 119909) gt 119890
minus1205751199041198801(119862 (119909))
(43)
(2) 120601119894isin 1198622(119863) 119894 = 1 2
(3) 1206012(119904 119909) = 119890
minus1205751199041198802(119909 minus 119862(119909)) and 120601
1(119904 119909) =
119890minus1205751199041198801(119862(119909)) for all (119904 119909) isin S 119863
(4) 1198601199061206011(119904 119909) minus ℎ(119904 119909 119906) le 0 and 119860119906120601
2(119904 119909) le 0 for all
(119904 119909) isin S 119863 and for all 119906 isin A(5) 119860120601
1(119904 119909) minus ℎ(119904 119909 ) = 0 for all (119904 119909) isin 119863 where
1198601206011is given by (42)
(6) 1198601206012(119904 119909) = 0 for all (119904 119909) isin 119863 where 119860120601
2is given
by (42)
Then the pair ( 120591) is a Nash equilibrium of the stochasticdifferential game (4) and (8) where
120591 = inf 119905 gt 0 119884(119905) notin 119863 (44)
6 Abstract and Applied Analysis
Moreover the corresponding equilibrium performances are
1206011(119904 119909) = J
(119904119909)
1( 120591)
1206012(119904 119909) = J
(119904119909)
2( 120591)
(45)
In this paper we will not discuss general solutions of thisfamily of nonlinear variational-integro inequalities Insteadwe discuss a solution in special case when
120574 (119905 119911) = 0 ℎ (119904 119909 119906) =1199062
2 (46)
Let us try the functions 120601119894 119894 = 1 2 of the form
120601119894(119904 119909) = 119890
minus120575119904120595119894(119909) (47)
and a continuation region119863 of the form
119863 = (119904 119909) 119909 lt 1199090 for some 119909
0gt 0 (48)
Then we have
119860119906120601119894(119904 119909) = 119890
minus120575119904119860119906120595119894(119909) (49)
where119860119906120595119894(119909) = minus 120575120595
119894(119909) + [(1 minus 119906) 119903119909 + 119906119887119909] 120595
1015840
119894(119909)
+1
211990921199062120590212059510158401015840
119894(119909)
(50)
By conditions (1) and (3) in Theorem 4 we get
1205951(119909) = 119880
1(119862 (119909)) 119909 ge 119909
0
1205951(119909) gt 119880
1(119862 (119909)) 0 lt 119909 lt 119909
0
1205952(119909) = 119880
2(119909 minus 119862 (119909)) 119909 ge 119909
0
1205952(119909) gt 119880
2(119909 minus 119862 (119909)) 0 lt 119909 lt 119909
0
(51)
From conditions (4) (5) and (6) ofTheorem 4 we get thecandidate for the optimal control as follows
= Argmax119906isinA
1198601199061205951(119909) minus
1199062
2
= Argmax119906isinA
minus 1205751205951(119909) + [(1 minus 119906) 119903119909 + 119906119887119909] 120595
1015840
1(119909)
+1
211990921199062120590212059510158401015840
1(119909) minus
1199062
2
=(119887119909 minus 119903119909) 120595
1015840
1(119909)
1 minus 1199092120590212059510158401015840
1(119909)
(52)
= Argmax119906isinA
1198601199061205952(119909)
= Argmax119906isinA
minus 1205751205952(119909) + [(1 minus 119906) 119903119909 + 119906119887119909] 120595
1015840
2(119909)
+1
211990921199062120590212059510158401015840
2(119909)
=minus (119887119909 minus 119903119909) 120595
1015840
2(119909)
1199092120590212059510158401015840
2(119909)
(53)
Let120595119894(119909) =
119894(119909) on 0 lt 119909 lt 119909
0 119894 = 1 2 By condition (5)
in Theorem 4 we have 1198601(119909) minus
22 = 0 for 0 lt 119909 lt 119909
0
Substituting (52) into 1198601199061(119909) minus 119906
22 = 0 we obtain
minus 1205751(119909) + [119903119909 +
(119887119909 minus 119903119909)21015840
1(119909)
1 minus 1199092120590210158401015840
1(119909)
] 1015840
1(119909)
+
[(119887119909 minus 119903119909) 1015840
1(119909)]2
2(1 minus 1199092120590210158401015840
1(119909))2[1199092120590210158401015840
1(119909) minus 1] = 0
(54)
Similarly we obtain1198602(119909) = 0 for 0 lt 119909 lt 119909
0by condition
(6) in Theorem 4 And we substitute (53) in 1198601199062(119909) = 0 to
get
minus 1205752(119909) + 119903119909
1015840
2(119909) +
[(119887119909 minus 119903119909) 1015840
2(119909)]2
1 minus 1199092120590210158401015840
2(119909)
+
[119909120590 (119887119909 minus 119903119909) 1015840
2(119909)]2
10158401015840
2(119909)
2(1 minus 1199092120590210158401015840
2(119909))2
= 0
(55)
Therefore we conclude that
1205951(119909) =
1198801(119862 (119909)) 119909 ge 119909
0
1(119909) 0 lt 119909 lt 119909
0
(56)
1205952(119909) =
1198802(119909 minus 119862 (119909)) 119909 ge 119909
0
2(119909) 0 lt 119909 lt 119909
0
(57)
where 1(119909) and
2(119909) are the solutions of (56) and (57)
respectivelyAccording to Theorem 4 we use the continuity and
differentiability of 120595119894at 119909 = 119909
0to determine 119909
0 119894 = 1 2 that
is
1205951(1199090) = 1198801(119862 (1199090))
1205951015840
1(1199090) = 119880
1015840
1(119862 (1199090)) 1198621015840(1199090)
1205952(1199090) = 1198802(1199090minus 119862 (119909
0))
1205951015840
2(1199090) = 119880
1015840
2(1199090minus 119862 (119909
0)) (1 minus 119862
1015840(1199090))
(58)
At the end of this section we summarize the above resultsin the following theorem
Theorem 5 Let 120595119894 119894 = 1 2 and let 119909
0be the solutions of
equations (56)ndash(58) Then the pair ( 120591) given by
=(119887119909 minus 119903119909) 120595
1015840
1(119909)
1 minus 1199092120590212059510158401015840
1(119909)
=minus (119887119909 minus 119903119909) 120595
1015840
2(119909)
1199092120590212059510158401015840
2(119909)
120591 = inf 119905 gt 0 119909 ge 1199090
(59)
is a Nash equilibrium of game (4) and (8) The correspondingequilibrium performances are
120601119894(119904 119909) = 119890
minus120575119904120595119894(119909) 119894 = 1 2 (60)
Abstract and Applied Analysis 7
4 Conclusion
A verification theorem is obtained for the general stochasticdifferential game between a controller and a stopper Inthe special case of quadratic cost we use this theorem tocharacterize the Nash equilibrium However the question ofthe existence and uniqueness of Nash equilibrium for thegame remains open It will be considered in our subsequentwork
Acknowledgment
This work was supported by the National Natural ScienceFoundation of China under Grant 61273022 and 11171050
References
[1] J Bertoin Levy Processes Cambridge University Press Cam-bridge UK 1996
[2] D Applebaum Levy Processes and Stochastic Calculus Cam-bridge University Press Cambridge UK 2003
[3] S Mataramvura and B Oslashksendal ldquoRisk minimizing portfoliosand HJBI equations for stochastic differential gamesrdquo Stochas-tics vol 80 no 4 pp 317ndash337 2008
[4] R J Elliott and T K Siu ldquoOn risk minimizing portfolios underaMarkovian regime-switching Black-Scholes economyrdquoAnnalsof Operations Research vol 176 pp 271ndash291 2010
[5] G Wang and Z Wu ldquoMean-variance hedging and forward-backward stochastic differential filtering equationsrdquo Abstractand Applied Analysis vol 2011 Article ID 310910 20 pages 2011
[6] I Karatzas and W Sudderth ldquoStochastic games of control andstopping for a linear diffusionrdquo in Random Walk SequentialAnalysis and Related Topics A Festschrift in Honor of YS ChowA Hsiung Z L Ying and C H Zhang Eds pp 100ndash117WorldScientific Publishers 2006
[7] S Hamadene ldquoMixed zero-sum stochastic differential gameand American game optionsrdquo SIAM Journal on Control andOptimization vol 45 no 2 pp 496ndash518 2006
[8] M K Ghosh M K S Rao and D Sheetal ldquoDifferential gamesof mixed type with control and stopping timesrdquo NonlinearDifferential Equations and Applications vol 16 no 2 pp 143ndash158 2009
[9] M K Ghosh and K S Mallikarjuna Rao ldquoExistence of value instochastic differential games of mixed typerdquo Stochastic Analysisand Applications vol 30 no 5 pp 895ndash905 2012
[10] I Karatzas and Q Li ldquoBSDE approach to non-zero-sumstochastic differential games of control and stoppingrdquo inStochastic Processes Finance and Control pp 105ndash153 WorldScientific Publishers 2012
[11] J-P Lepeltier ldquoOn a general zero-sum stochastic game withstopping strategy for one player and continuous strategy for theotherrdquo Probability and Mathematical Statistics vol 6 no 1 pp43ndash50 1985
[12] I Karatzas and W D Sudderth ldquoThe controller-and-stoppergame for a linear diffusionrdquo The Annals of Probability vol 29no 3 pp 1111ndash1127 2001
[13] AWeerasinghe ldquoA controller and a stopper gamewith degener-ate variance controlrdquo Electronic Communications in Probabilityvol 11 pp 89ndash99 2006
[14] I Karatzas and I-M Zamfirescu ldquoMartingale approach tostochastic differential games of control and stoppingrdquo TheAnnals of Probability vol 36 no 4 pp 1495ndash1527 2008
[15] E Bayraktar I Karatzas and S Yao ldquoOptimal stopping fordynamic convex risk measuresrdquo Illinois Journal of Mathematicsvol 54 no 3 pp 1025ndash1067 2010
[16] E Bayraktar and V R Young ldquoProving regularity of theminimal probability of ruin via a game of stopping and controlrdquoFinance and Stochastics vol 15 no 4 pp 785ndash818 2011
[17] F Bagherya S Haademb B Oslashksendal and I Turpina ldquoOptimalstopping and stochastic control differential games for jumpdiffusionsrdquo Stochastics vol 85 no 1 pp 85ndash97 2013
[18] B Oslashksendal and A Sulem Applied Stochastic Control of JumpDiffusions Springer Berlin Germany 2nd edition 2007
[19] Q Lin R Loxton K L Teo and Y H Wu ldquoA new computa-tional method for a class of free terminal time optimal controlproblemsrdquo Pacific Journal of Optimization vol 7 no 1 pp 63ndash81 2011
[20] Q Lin R Loxton K L Teo and Y H Wu ldquoOptimal controlcomputation for nonlinear systems with state-dependent stop-ping criteriardquo Automatica vol 48 no 9 pp 2116ndash2129 2012
[21] C Jiang Q Lin C Yu K L Teo and G-R Duan ldquoAn exactpenalty method for free terminal time optimal control problemwith continuous inequality constraintsrdquo Journal of OptimizationTheory and Applications vol 154 no 1 pp 30ndash53 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
![Page 2: Research Article Nonzero-Sum Stochastic Differential Game ...downloads.hindawi.com/journals/aaa/2013/761306.pdf · Controller and Stopper for Jump Diffusions YanWang, 1,2 AiminSong,](https://reader033.vdocuments.mx/reader033/viewer/2022042410/5f26fe8f7005a2296769470b/html5/thumbnails/2.jpg)
2 Abstract and Applied Analysis
Assume that an investor hires a portfolio manager tomanage his wealth 119883(119905) from investments The manager(controller) can choose a portfolio 119906(119905) which represents theproportion of the total wealth 119883(119905) invested in the stock attime 119905 And the investor (stopper) can halt the wealth process119883(119905) by selecting a stop-rule 120591 119862[0 119879] rarr [0 119879] Then thedynamics of the corresponding wealth process 119883(119905) = 119883
119906(119905)
is
119889119883119906(119905) = 119883
119906(119905minus) [(1 minus 119906 (119905)) 119903 (119905) + 119906 (119905) 119887 (119905)] 119889119905
+ 119906 (119905) 120590 (119905) 119889119861 (119905)
+119906 (119905minus) intR
120574 (119905 119911) (119889119905 119889119911)
119883119906(0) = 119909 gt 0
(4)
(see eg [3ndash5])We require that 119906(119905minus)120574(119905 119911) gt minus1 for aa 119905 119911as and that
int
119879
0
|(1 minus 119906 (119905)) 119903 (119905)| + |119906 (119905) 119887 (119905)| + 1199062(119905) 1205902(119905)
+1199062(119905) int
R
1205742(119905 119911) ] (119889119911) 119889119905 lt infin as
(5)
At terminal time 120591 the stopper gives the controller apayoff 119862(119883(120591)) where 119862 119862[0 119879] rarr R is a deterministicmapping Therefore the controller aims to maximize hisutility of the following form
J119909
1(119906 120591)=119864
119909[119890minus120575120591
1198801(119862 (119883 (120591)))minusint
120591
0
119890minus120575119905ℎ (119905 119883 (119905) 119906
119905) 119889119905]
(6)
where 120575 gt 0 is the discounting rate ℎ is a cost function and1198801is the controllerrsquos utility We denote 119864119909 the expectation
with respect to119875119909 and119875119909 the probability laws of119883(119905) startingat 119909
Meanwhile it is the stopperrsquos objective to choose thestopping time 120591 such that his own utility
J119909
2(119906 120591) = 119864
119909[119890minus120575120591
1198802119883 (120591) minus 119862 (119883 (120591))] (7)
is maximized where 1198802is the stopperrsquos utility
As this game is typically a nonzero sum we seek a Nashequilibrium namely a pair (119906lowast 120591lowast) such that
J119909
1(119906lowast 120591lowast) ge J
119909
1(119906 120591lowast) forall119906
J119909
2(119906lowast 120591lowast) ge J
119909
2(119906lowast 120591) forall120591
(8)
This means that the choice 119906lowast is optimal for the controllerwhen the stopper uses 120591lowast and vice verse
The game (4) and (8) are a nonzero-sum stochasticdifferential game between a controller and a stopper Theexistence of Nash equilibrium shows that by an appropriatestopping rule design 120591lowast the stopper can induce the controllerto choose the best portfolio he can Similarly by applying asuitable portfolio 119906lowast the controller can force the stopper tostop the employment relationship at a time of the controllerrsquoschoosing
There have been significant advances in the researchof stochastic differential games of control and stoppingFor example in [6ndash9] the authors considered the zero-sumstochastic differential games ofmixed type with both controlsand stopping between two players In these games each of theplayers chooses an optimal strategy which is composed by acontrol 119906(sdot) and a stopping 120591 Under appropriate conditionsthey constructed a saddle pair of optimal strategies For thenonsum case the games of mixed type were discussed in[6 10] The authors presented Nash equilibria rather thansaddle pairs of strategies [10] Moreover the papers [11ndash17]considered a zero-sum stochastic differential game betweencontroller and stopper where one player (controller) choosesa control process 119906(sdot) and the other (stopper) chooses astopping 120591 One player tries to maximize the reward and theother to minimize it They presented a saddle pair for thegame
In this paper we study a nonzero-sum stochastic differen-tial game between a controller and a stopper The controllerand the stopper have different payoffsThe objectives of themare tomaximize their own payoffsThis game is considered ina jumpdiffusion context under theMarkov control conditionWe prove a verification theorem in terms of VIHJB equationsfor the game to characterize Nash equilibrium
Our setup and approach are related to [3 17] Howevertheir games are different fromours In [3] the authors studiedthe stochastic differential games between two controllersThe work in [17] was carried out for a zero-sum stochasticdifferential game between a controller and a stopper
The paper is organized as follows in the next sectionwe formulate the nonzero-sum stochastic differential gamebetween a controller and a stopper and prove a generalverification theorem In Section 3 we apply the generalresults obtained in Section 2 to characterize the solutions of aspecial game Finally we conclude this paper in Section 4
2 A Verification Theorem forNonzero-Sum Stochastic Differential Gamebetween Controller and Stopper
Suppose the state 119884(119905) = 119884119906(119905) at time 119905 is given by the
following stochastic differential equation
119889119884 (119905) = 120572 (119884 (119905) 1199060(119905)) 119889119905 + 120573 (119884 (119905) 119906
0(119905)) 119889119861 (119905)
+ intR1198960
120579 (119884 (119905minus) 1199061(119905minus 119911) 119911) (119889119905 119889119911)
119884 (0) = 119910 isin R119896
(9)
where 120572 R119896 times U rarr R119896 120573 R119896 times U rarr R119896times119896 120579 R119896 times
U times R119896 rarr R119896times119896 are given functions and U denotes a givensubset of R119901
We regard 1199060(119905) = 119906
0(119905 120596) and 119906
1(119905 119911) = 119906
1(119905 119911 120596)
as the control processes assumed to be cadlag F119905-adapted
and with values in 1199060(119905) isin U 119906
1(119905 119911) isin U for aa 119905 119911 120596
And we put 119906(119905) = (1199060(119905) 1199061(119905 119911)) Then 119884(119905) = 119884
(119906)(119905) is
a controlled jump diffusion (see [18] for more informationabout stochastic control of jump diffusion)
Abstract and Applied Analysis 3
Fix an open solvency set S sub R119896 Let
120591S = inf 119905 gt 0 119884 (119905) notin S (10)
be the bankruptcy time 120591S is the first time at which thestochastic process 119884(119905) exits the solvency set S Similaroptimal control problems in which the terminal time isgoverned by a stopping criterion are considered in [19ndash21] inthe deterministic case
Let 119891119894 R119896 times 119870 rarr R and 119892
119894 R119896 rarr R be given
functions for 119894 = 1 2 Let A be a family of admissiblecontrols contained in the set of 119906(sdot) such that (9) has a uniquestrong solution and
119864119910[int
120591S
0
1003816100381610038161003816119891119894 (119884119905 119906119905)1003816100381610038161003816 119889119905] lt infin 119894 = 1 2 (11)
for all 119910 isin S where 119864119910 denotes expectation given that119884(0) =119910 isin R119896 Denote by Γ the set of all stopping times 120591 le 120591SMoreover we assume that
the family 119892minus
119894(119884120591) 120591 isin Γ is uniformly integrable
for 119894 = 1 2
(12)
Then for 119906 isin A and 120591 isin Γ we define the performancefunctionals as follows
J119910
119894(119906 120591) = 119864
119910[int
120591
0
119891119894(119884 (119905) 119906 (119905)) 119889119905 + 119892
119894(119884 (120591))]
119894 = 1 2
(13)
We interpret 119892119894(119884(120591)) as 0 if 120591 = infin We may regardJ
119910
1(119906 120591)
as the payoff to the controller who controls 119906 andJ1199102(119906 120591) as
the payoff to the stopper who decides 120591
Definition 2 (nash equilibrium) A pair (119906lowast 120591lowast) isin A times Γ iscalled a Nash equilibrium for the stochastic differential game(9) and (13) if the following holds
J119910
1(119906lowast 120591lowast) ge J
119910
1(119906 120591lowast) forall119906 isin U 119910 isin S (14)
J119910
2(119906lowast 120591lowast) ge J
119910
2(119906lowast 120591) forall120591 isin Γ 119910 isin S (15)
Condition (14) states that if the stopper chooses 120591lowast it isoptimal for the controller to use the control 119906lowast Similarlycondition (15) states that if the controller uses 119906lowast it is optimalfor the stopper to decide 120591lowast Thus (119906lowast 120591lowast) is an equilibriumpoint in the sense that there is no reason for each individualplayer to deviate from it as long as the other player does not
We restrict ourselves to Markov controls that is weassume that 119906
0(119905) =
0(119884(119905)) 119906
1(119905 119911) =
1(119884(119905) 119911) As
customary we do not distinguish between 1199060and
0 1199061and
1 Then the controls 119906
0and 119906
1can simply be identified with
functions 0(119910) and
1(119910 119911) where 119910 isin S and 119911 isin R119896
When the control 119906 is Markovian the correspondingprocess 119884(119906)(119905) becomes a Markov process with the generator119860119906 of 120601 isin 1198622(R119896) given by
119860120601 (119910) =
119896
sum
119894=1
120572119894(119910 1199060(119910))
120597120601
120597119910119894
(119910)
+1
2
119896
sum
119894119895=1
(120573120573119879)119894119895(119910 1199060(119910))
1205972120601
120597119910119894120597119910119895
(119910)
+
119896
sum
119895=1
intR
120601 (119910 + 120579119895(119910 1199061(119910 119911) 119911)) minus 120601 (119910)
minusnabla120601 (119910) 120579119895(119910 1199061(119910 119911) 119911) ] (119889119911)
(16)
where nabla120601 = (1205971206011205971199101 120597120601120597119910
119896) is the gradient of 120601 and 120579119895
is the 119895th column of the 119896 times 119896matrix 120579Now we can state the main result of this section
Theorem 3 (verification theorem for game (9) and (13))Suppose there exist two functions 120601
119894 S rarr R 119894 = 1 2 such
that
(i) 120601119894isin 1198621(S0)⋂119862(S) 119894 = 1 2 whereS is the closure of
S and S0 is the interior of S(ii) 120601119894ge 119892119894on S 119894 = 1 2
Define the following continuation regions
119863119894= 119910 isin S 120601
119894(119910) gt 119892
119894(119910) 119894 = 1 2 (17)
Suppose 119884(119905) = 119884119906(119905) spends 0 time on 120597119863
119894as 119894 =
1 2 that is(iii) 119864119910[int120591119878
0120594120597119863119894(119884(119905))119889119905] = 0 for all 119910 isin S 119906 isin U 119894 = 1 2
(iv) 120597119863119894is a Lipschitz surface 119894 = 1 2
(v) 120601119894isin 1198622(S120597119863
119894)The second-order derivatives of 120601
119894are
locally bounded near 120597119863119894 respectively 119894 = 1 2
(vi) 1198631= 1198632= 119863
(vii) there exists isin A such that for 119894 = 1 2
119860120601119894(119910) + 119891
119894(119910 (119910))
= sup119906isinA
119860119906120601119894(119910) + 119891
119894(119910 119906 (119910))
= 0 119910 isin 119863
le 0 119910 isin S 119863
(18)
(viii) 119864119910[|120601119894(119884120591)| + int120591
0|119860119906120601119894(119884119906(119905))|119889119905] lt infin 119894 = 1 2 for all
119906 isin U 120591 isin ΓFor 119906 isin A define
120591119863= 120591119906
119863= inf 119905 gt 0 119884
119906(119905) notin 119863 lt infin (19)
and in particular
120591119863= 120591
119863= inf 119905 gt 0 119884
(119905) notin 119863 lt infin (20)
4 Abstract and Applied Analysis
Suppose that(ix) the family 120601
119894(119884(120591)) 120591 isin Γ 120591 le 120591
119863 is uniformly
integrable for all 119906 isin A and 119910 isin S 119894 = 1 2
Then (120591119863 ) isin Γ timesA is a Nash equilibrium for game (9) (13)
and
1206011(119910) = sup
119906isinA
J119906120591119863
1(119910) = J
120591119863
1(119910) (21)
1206012(119910) = sup
120591isinΓ
J120591
2(119910) = J
120591119863
2(119910) (22)
Proof From (i) (iv) and (v) we may assume by an approx-imation theorem (see Theorem 21 in [18]) that 120601
119894isin 1198622(S)
119894 = 1 2For a given 119906 isin A we define with 119884(119905) = 119884
119906(119905)
120591119863= 120591119906
119863= inf 119905 gt 0 119884
119906(119905) notin 119863 (23)
In particular let be as in (vii) Then
120591119863= 120591
119863= inf 119905 gt 0 119884
(119905) notin 119863 (24)
We first prove that (21) holds Let 120591119863isin Γ be as in (24) For
arbitrary 119906 isin A by (vii) and the Dynkinrsquos formula for jumpdiffusion (see Theorem 124 in [18]) we have
1206011(119910) = 119864
119910[int
120591119863and119898
0
minus1198601199061206011(119884 (119905)) 119889119905 + 120601
1(119884 (120591119863and 119898))]
ge 119864119910[int
120591119863and119898
0
1198911(119884 (119905) 119906 (119905)) 119889119905 + 120601
1(119884 (120591119863and 119898))]
(25)
where119898 = 1 2 Therefore by (11) (12) (i) (ii) (viii) andthe Fatou lemma
1206011(119910)
ge lim inf119898rarrinfin
119864119910[int
120591119863and119898
0
1198911(119884 (119905) 119906 (119905)) 119889119905+120601
1(119884 (120591119863and 119898))]
ge 119864119910[int
120591119863
0
1198911(119884 (119905) 119906 (119905)) 119889119905 + 119892
1(119884 (120591119863))]
= J119910
1(119906 120591119863)
(26)
Since this holds for all 119906 isin A we have
1206011(119910) ge sup
119906isinA
J119910
1(119906 120591119863) (27)
In particular applying the Dynkinrsquos formula to 119906 = we getan equality that is
1206011(119910) = 119864
119910[int
120591119863and119898
0
minus1198601206011( (119905)) 119889119905 + 120601
1( (120591119863and 119898))]
= 119864119910[int
120591119863and119898
0
1198911( (119905) (119905)) 119889119905 + 120601
1( (120591119863and 119898))]
(28)
where (119905) = 119884(119905) and119898 = 1 2 Hence we deduce that
1206011(119910) = J
119910
1( 120591119863) (29)
Since we always have
J119910
1( 120591119863) le sup119906isinA
J119910
1(119906 120591119863) (30)
we conclude by combining (27) (29) and (30) that
1206011(119910) = J
119910
1( 120591119863) = sup119906isinA
J119910
1(119906 120591119863) (31)
which is (21)Next we prove that (22) holds Let isin A be as in (vii)
For 120591 isin Γ by the Dynkinrsquos formula and (vii) we have
119864119910[1206012( (120591119898))] = 120601
2(119910) + 119864
119910[int
120591119898
0
1198601206012( (119905)) 119889119905]
le 1206012(119910) minus 119864
119910[int
120591119898
0
1198912( (119905) (119905)) 119889119905]
(32)
where 120591119898= 120591 and 119898119898 = 1 2
Letting119898 rarr infin gives by (11) (12) (i) (ii) (viii) and theFatou Lemma
1206012(119910) ge lim inf
119898rarrinfin119864119910[int
120591and119898
0
1198912( (119905) (119905)) 119889119905 + 120601
2( (120591119898))]
ge 119864119910[int
120591
0
1198912( (119905) (119905)) 119889119905 + 119892
2( (120591)) 120594
120591ltinfin]
= J119910
2( 120591)
(33)
The inequality (33) holds for all 120591 isin Γ Then we have
1206012(119910) ge sup
120591isinΓ
J119910
2( 120591) (34)
Similarly applying the above argument to the pair ( 120591119863) we
get an equality in (34) that is
1206012(119910) = J
119910
2( 120591119863) (35)
We always have
J119910
2( 120591119863) le sup120591isinΓ
J119910
2( 120591) (36)
Therefore combining (34) (35) and (36) we get
1206012(119910) = J
119910
2( 120591119863) = sup120591isinΓ
J119910
2( 120591) (37)
which is (22) The proof is completed
Abstract and Applied Analysis 5
3 An Example
In this section we come back to Example 1 and useTheorem 3to study the solutions of game (4) and (8) Here and in thefollowing all the processes are assumed to be one-dimensionfor simplicity
To put game (4) and (8) into the framework of Section 2we define the process 119884(119905) = (119884
0(119905) 1198841(119905)) 119884(0) = 119910 = (119904 119910
1)
by
1198891198840(119905) = 119889119905 119884
0(0) = 119904 isin R
1198891198841(119905) = 119889119883 (119905)
= 1198841(119905) [(1 minus 119906 (119905)) 119903 (119905) + 119906 (119905) 119887 (119905)] 119889119905
+ 119906 (119905) 120590 (119905) 119889119861 (119905)
+119906 (119905minus) intR
120574 (119905 119911) (119889119905 119889119911)
1198841(0) = 119909 = 119910
1
(38)
Then the performance functionals to the controller (6) andthe stopper (7) can be formulated as follows
J119910
1(119906 120591) = 119864
1199041199101 [119890minus120575(119904+120591)
1198801(119862 (1198841(120591)))
minusint
120591
0
119890minus120575(119904+119905)
ℎ (1198840(119905) 1198841(119905) 119906 (119905)) 119889119905]
J119910
2(119906 120591) = 119864
1199041199101 [119890minus120575(119904+120591)
1198802(1198841(120591) minus 119862 (119884
1(120591)))]
(39)
In this case the generator 119860119906 in (16) has the form
119860119906120601 (119904 119909)
=120597120601
120597119904+ [(1 minus 119906) 119903119909 + 119906119887119909]
120597120601
120597119909+1
2119909211990621205902 1205972120601
1205971199092
+ intR0
120601 (119904 119909 + 119909119906120574 (119911)) minus 120601 (119904 119909) minus 119909119906120574 (119911)120597120601
120597119909
times ] (119889119911)
(40)
To obtain a possible Nash equilibrium ( 120591) isin A times Γ forgame (4) and (8) according to Theorem 3 it is necessary tofind a subset 119863 of S = R2
+= [0infin)
2 and 120601119894(119904 119909) 119894 = 1 2
such that
(i) 1206011(119904 119909) = 119890
minus1205751199041198801(119862(119909)) and 120601
2(119904 119909) = 119890
minus1205751199041198802(119909 minus
119862(119909)) for all (119904 119909) isin 119863(ii) 1206011(119904 119909) ge 119890
minus1205751199041198801(119862(119909)) and 120601
2(119904 119909) ge 119890
minus1205751199041198802(119909 minus
119862(119909)) for all (119904 119909) isin S(iii) 119860119906120601
1(119904 119909) minus ℎ(119904 119909 119906) le 0 and 119860119906120601
2(119904 119909) le 0 for all
(119904 119909) isin S 119863 and 119906(iv) there exists such that 119860120601
1(119904 119909) minus ℎ(119904 119909 ) =
1198601206012(119904 119909) = 0 for all (119904 119909) isin 119863
Imposing the first-order condition on 1198601199061206011(119904 119909) minus
ℎ(119904 119909 119906) and 1198601199061206012(119904 119909) we get the following equations for
the optimal control processes
(119887119909 minus 119903119909)1205971206011
120597119909(119904 119909) + 119909
2120590212059721206011
1205971199092(119904 119909)
+ intR
119903120574 (119911) [1205971206011
120597119909(119904 119909 + 119903120574 (119911)) minus
1205971206011
120597119909(119904 119909)]
minus120597ℎ
120597119906(119904 119909 ) = 0
(119887119909 minus 119903119909)1205971206012
120597119909(119904 119909) + 119909
2120590212059721206012
1205971199092(119904 119909)
+ intR
119903120574 (119911) [1205971206012
120597119909(119904 119909 + 119903120574 (119911)) minus
1205971206012
120597119909(119904 119909)] = 0
(41)
With as in (41) we put
119860120601119894(119904 119909)
=120597120601119894
120597119904+ [(1 minus ) 119903119909 + 119887119909]
120597120601119894
120597119909+1
2119909221205902 1205972120601119894
1205971199092
+ intR0
120601119894(119904 119909 + 119909120574 (119911)) minus 120601
119894(119904 119909) minus 119909120574 (119911)
120597120601119894
120597119909
times ] (119889119911)
(42)
Thus we may reduce game (4) and (8) to the problem ofsolving a family of nonlinear variational-integro inequalitiesWe summarize as follows
Theorem 4 Suppose there exist satisfying (41) and two 1198621-functions 120601
119894 119894 = 1 2 such that
(1)
119863 = (119904 119909) 1206012(119904 119909) gt 119890
minus1205751199041198802(119909 minus 119862 (119909))
= (119904 119909) 1206011(119904 119909) gt 119890
minus1205751199041198801(119862 (119909))
(43)
(2) 120601119894isin 1198622(119863) 119894 = 1 2
(3) 1206012(119904 119909) = 119890
minus1205751199041198802(119909 minus 119862(119909)) and 120601
1(119904 119909) =
119890minus1205751199041198801(119862(119909)) for all (119904 119909) isin S 119863
(4) 1198601199061206011(119904 119909) minus ℎ(119904 119909 119906) le 0 and 119860119906120601
2(119904 119909) le 0 for all
(119904 119909) isin S 119863 and for all 119906 isin A(5) 119860120601
1(119904 119909) minus ℎ(119904 119909 ) = 0 for all (119904 119909) isin 119863 where
1198601206011is given by (42)
(6) 1198601206012(119904 119909) = 0 for all (119904 119909) isin 119863 where 119860120601
2is given
by (42)
Then the pair ( 120591) is a Nash equilibrium of the stochasticdifferential game (4) and (8) where
120591 = inf 119905 gt 0 119884(119905) notin 119863 (44)
6 Abstract and Applied Analysis
Moreover the corresponding equilibrium performances are
1206011(119904 119909) = J
(119904119909)
1( 120591)
1206012(119904 119909) = J
(119904119909)
2( 120591)
(45)
In this paper we will not discuss general solutions of thisfamily of nonlinear variational-integro inequalities Insteadwe discuss a solution in special case when
120574 (119905 119911) = 0 ℎ (119904 119909 119906) =1199062
2 (46)
Let us try the functions 120601119894 119894 = 1 2 of the form
120601119894(119904 119909) = 119890
minus120575119904120595119894(119909) (47)
and a continuation region119863 of the form
119863 = (119904 119909) 119909 lt 1199090 for some 119909
0gt 0 (48)
Then we have
119860119906120601119894(119904 119909) = 119890
minus120575119904119860119906120595119894(119909) (49)
where119860119906120595119894(119909) = minus 120575120595
119894(119909) + [(1 minus 119906) 119903119909 + 119906119887119909] 120595
1015840
119894(119909)
+1
211990921199062120590212059510158401015840
119894(119909)
(50)
By conditions (1) and (3) in Theorem 4 we get
1205951(119909) = 119880
1(119862 (119909)) 119909 ge 119909
0
1205951(119909) gt 119880
1(119862 (119909)) 0 lt 119909 lt 119909
0
1205952(119909) = 119880
2(119909 minus 119862 (119909)) 119909 ge 119909
0
1205952(119909) gt 119880
2(119909 minus 119862 (119909)) 0 lt 119909 lt 119909
0
(51)
From conditions (4) (5) and (6) ofTheorem 4 we get thecandidate for the optimal control as follows
= Argmax119906isinA
1198601199061205951(119909) minus
1199062
2
= Argmax119906isinA
minus 1205751205951(119909) + [(1 minus 119906) 119903119909 + 119906119887119909] 120595
1015840
1(119909)
+1
211990921199062120590212059510158401015840
1(119909) minus
1199062
2
=(119887119909 minus 119903119909) 120595
1015840
1(119909)
1 minus 1199092120590212059510158401015840
1(119909)
(52)
= Argmax119906isinA
1198601199061205952(119909)
= Argmax119906isinA
minus 1205751205952(119909) + [(1 minus 119906) 119903119909 + 119906119887119909] 120595
1015840
2(119909)
+1
211990921199062120590212059510158401015840
2(119909)
=minus (119887119909 minus 119903119909) 120595
1015840
2(119909)
1199092120590212059510158401015840
2(119909)
(53)
Let120595119894(119909) =
119894(119909) on 0 lt 119909 lt 119909
0 119894 = 1 2 By condition (5)
in Theorem 4 we have 1198601(119909) minus
22 = 0 for 0 lt 119909 lt 119909
0
Substituting (52) into 1198601199061(119909) minus 119906
22 = 0 we obtain
minus 1205751(119909) + [119903119909 +
(119887119909 minus 119903119909)21015840
1(119909)
1 minus 1199092120590210158401015840
1(119909)
] 1015840
1(119909)
+
[(119887119909 minus 119903119909) 1015840
1(119909)]2
2(1 minus 1199092120590210158401015840
1(119909))2[1199092120590210158401015840
1(119909) minus 1] = 0
(54)
Similarly we obtain1198602(119909) = 0 for 0 lt 119909 lt 119909
0by condition
(6) in Theorem 4 And we substitute (53) in 1198601199062(119909) = 0 to
get
minus 1205752(119909) + 119903119909
1015840
2(119909) +
[(119887119909 minus 119903119909) 1015840
2(119909)]2
1 minus 1199092120590210158401015840
2(119909)
+
[119909120590 (119887119909 minus 119903119909) 1015840
2(119909)]2
10158401015840
2(119909)
2(1 minus 1199092120590210158401015840
2(119909))2
= 0
(55)
Therefore we conclude that
1205951(119909) =
1198801(119862 (119909)) 119909 ge 119909
0
1(119909) 0 lt 119909 lt 119909
0
(56)
1205952(119909) =
1198802(119909 minus 119862 (119909)) 119909 ge 119909
0
2(119909) 0 lt 119909 lt 119909
0
(57)
where 1(119909) and
2(119909) are the solutions of (56) and (57)
respectivelyAccording to Theorem 4 we use the continuity and
differentiability of 120595119894at 119909 = 119909
0to determine 119909
0 119894 = 1 2 that
is
1205951(1199090) = 1198801(119862 (1199090))
1205951015840
1(1199090) = 119880
1015840
1(119862 (1199090)) 1198621015840(1199090)
1205952(1199090) = 1198802(1199090minus 119862 (119909
0))
1205951015840
2(1199090) = 119880
1015840
2(1199090minus 119862 (119909
0)) (1 minus 119862
1015840(1199090))
(58)
At the end of this section we summarize the above resultsin the following theorem
Theorem 5 Let 120595119894 119894 = 1 2 and let 119909
0be the solutions of
equations (56)ndash(58) Then the pair ( 120591) given by
=(119887119909 minus 119903119909) 120595
1015840
1(119909)
1 minus 1199092120590212059510158401015840
1(119909)
=minus (119887119909 minus 119903119909) 120595
1015840
2(119909)
1199092120590212059510158401015840
2(119909)
120591 = inf 119905 gt 0 119909 ge 1199090
(59)
is a Nash equilibrium of game (4) and (8) The correspondingequilibrium performances are
120601119894(119904 119909) = 119890
minus120575119904120595119894(119909) 119894 = 1 2 (60)
Abstract and Applied Analysis 7
4 Conclusion
A verification theorem is obtained for the general stochasticdifferential game between a controller and a stopper Inthe special case of quadratic cost we use this theorem tocharacterize the Nash equilibrium However the question ofthe existence and uniqueness of Nash equilibrium for thegame remains open It will be considered in our subsequentwork
Acknowledgment
This work was supported by the National Natural ScienceFoundation of China under Grant 61273022 and 11171050
References
[1] J Bertoin Levy Processes Cambridge University Press Cam-bridge UK 1996
[2] D Applebaum Levy Processes and Stochastic Calculus Cam-bridge University Press Cambridge UK 2003
[3] S Mataramvura and B Oslashksendal ldquoRisk minimizing portfoliosand HJBI equations for stochastic differential gamesrdquo Stochas-tics vol 80 no 4 pp 317ndash337 2008
[4] R J Elliott and T K Siu ldquoOn risk minimizing portfolios underaMarkovian regime-switching Black-Scholes economyrdquoAnnalsof Operations Research vol 176 pp 271ndash291 2010
[5] G Wang and Z Wu ldquoMean-variance hedging and forward-backward stochastic differential filtering equationsrdquo Abstractand Applied Analysis vol 2011 Article ID 310910 20 pages 2011
[6] I Karatzas and W Sudderth ldquoStochastic games of control andstopping for a linear diffusionrdquo in Random Walk SequentialAnalysis and Related Topics A Festschrift in Honor of YS ChowA Hsiung Z L Ying and C H Zhang Eds pp 100ndash117WorldScientific Publishers 2006
[7] S Hamadene ldquoMixed zero-sum stochastic differential gameand American game optionsrdquo SIAM Journal on Control andOptimization vol 45 no 2 pp 496ndash518 2006
[8] M K Ghosh M K S Rao and D Sheetal ldquoDifferential gamesof mixed type with control and stopping timesrdquo NonlinearDifferential Equations and Applications vol 16 no 2 pp 143ndash158 2009
[9] M K Ghosh and K S Mallikarjuna Rao ldquoExistence of value instochastic differential games of mixed typerdquo Stochastic Analysisand Applications vol 30 no 5 pp 895ndash905 2012
[10] I Karatzas and Q Li ldquoBSDE approach to non-zero-sumstochastic differential games of control and stoppingrdquo inStochastic Processes Finance and Control pp 105ndash153 WorldScientific Publishers 2012
[11] J-P Lepeltier ldquoOn a general zero-sum stochastic game withstopping strategy for one player and continuous strategy for theotherrdquo Probability and Mathematical Statistics vol 6 no 1 pp43ndash50 1985
[12] I Karatzas and W D Sudderth ldquoThe controller-and-stoppergame for a linear diffusionrdquo The Annals of Probability vol 29no 3 pp 1111ndash1127 2001
[13] AWeerasinghe ldquoA controller and a stopper gamewith degener-ate variance controlrdquo Electronic Communications in Probabilityvol 11 pp 89ndash99 2006
[14] I Karatzas and I-M Zamfirescu ldquoMartingale approach tostochastic differential games of control and stoppingrdquo TheAnnals of Probability vol 36 no 4 pp 1495ndash1527 2008
[15] E Bayraktar I Karatzas and S Yao ldquoOptimal stopping fordynamic convex risk measuresrdquo Illinois Journal of Mathematicsvol 54 no 3 pp 1025ndash1067 2010
[16] E Bayraktar and V R Young ldquoProving regularity of theminimal probability of ruin via a game of stopping and controlrdquoFinance and Stochastics vol 15 no 4 pp 785ndash818 2011
[17] F Bagherya S Haademb B Oslashksendal and I Turpina ldquoOptimalstopping and stochastic control differential games for jumpdiffusionsrdquo Stochastics vol 85 no 1 pp 85ndash97 2013
[18] B Oslashksendal and A Sulem Applied Stochastic Control of JumpDiffusions Springer Berlin Germany 2nd edition 2007
[19] Q Lin R Loxton K L Teo and Y H Wu ldquoA new computa-tional method for a class of free terminal time optimal controlproblemsrdquo Pacific Journal of Optimization vol 7 no 1 pp 63ndash81 2011
[20] Q Lin R Loxton K L Teo and Y H Wu ldquoOptimal controlcomputation for nonlinear systems with state-dependent stop-ping criteriardquo Automatica vol 48 no 9 pp 2116ndash2129 2012
[21] C Jiang Q Lin C Yu K L Teo and G-R Duan ldquoAn exactpenalty method for free terminal time optimal control problemwith continuous inequality constraintsrdquo Journal of OptimizationTheory and Applications vol 154 no 1 pp 30ndash53 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
![Page 3: Research Article Nonzero-Sum Stochastic Differential Game ...downloads.hindawi.com/journals/aaa/2013/761306.pdf · Controller and Stopper for Jump Diffusions YanWang, 1,2 AiminSong,](https://reader033.vdocuments.mx/reader033/viewer/2022042410/5f26fe8f7005a2296769470b/html5/thumbnails/3.jpg)
Abstract and Applied Analysis 3
Fix an open solvency set S sub R119896 Let
120591S = inf 119905 gt 0 119884 (119905) notin S (10)
be the bankruptcy time 120591S is the first time at which thestochastic process 119884(119905) exits the solvency set S Similaroptimal control problems in which the terminal time isgoverned by a stopping criterion are considered in [19ndash21] inthe deterministic case
Let 119891119894 R119896 times 119870 rarr R and 119892
119894 R119896 rarr R be given
functions for 119894 = 1 2 Let A be a family of admissiblecontrols contained in the set of 119906(sdot) such that (9) has a uniquestrong solution and
119864119910[int
120591S
0
1003816100381610038161003816119891119894 (119884119905 119906119905)1003816100381610038161003816 119889119905] lt infin 119894 = 1 2 (11)
for all 119910 isin S where 119864119910 denotes expectation given that119884(0) =119910 isin R119896 Denote by Γ the set of all stopping times 120591 le 120591SMoreover we assume that
the family 119892minus
119894(119884120591) 120591 isin Γ is uniformly integrable
for 119894 = 1 2
(12)
Then for 119906 isin A and 120591 isin Γ we define the performancefunctionals as follows
J119910
119894(119906 120591) = 119864
119910[int
120591
0
119891119894(119884 (119905) 119906 (119905)) 119889119905 + 119892
119894(119884 (120591))]
119894 = 1 2
(13)
We interpret 119892119894(119884(120591)) as 0 if 120591 = infin We may regardJ
119910
1(119906 120591)
as the payoff to the controller who controls 119906 andJ1199102(119906 120591) as
the payoff to the stopper who decides 120591
Definition 2 (nash equilibrium) A pair (119906lowast 120591lowast) isin A times Γ iscalled a Nash equilibrium for the stochastic differential game(9) and (13) if the following holds
J119910
1(119906lowast 120591lowast) ge J
119910
1(119906 120591lowast) forall119906 isin U 119910 isin S (14)
J119910
2(119906lowast 120591lowast) ge J
119910
2(119906lowast 120591) forall120591 isin Γ 119910 isin S (15)
Condition (14) states that if the stopper chooses 120591lowast it isoptimal for the controller to use the control 119906lowast Similarlycondition (15) states that if the controller uses 119906lowast it is optimalfor the stopper to decide 120591lowast Thus (119906lowast 120591lowast) is an equilibriumpoint in the sense that there is no reason for each individualplayer to deviate from it as long as the other player does not
We restrict ourselves to Markov controls that is weassume that 119906
0(119905) =
0(119884(119905)) 119906
1(119905 119911) =
1(119884(119905) 119911) As
customary we do not distinguish between 1199060and
0 1199061and
1 Then the controls 119906
0and 119906
1can simply be identified with
functions 0(119910) and
1(119910 119911) where 119910 isin S and 119911 isin R119896
When the control 119906 is Markovian the correspondingprocess 119884(119906)(119905) becomes a Markov process with the generator119860119906 of 120601 isin 1198622(R119896) given by
119860120601 (119910) =
119896
sum
119894=1
120572119894(119910 1199060(119910))
120597120601
120597119910119894
(119910)
+1
2
119896
sum
119894119895=1
(120573120573119879)119894119895(119910 1199060(119910))
1205972120601
120597119910119894120597119910119895
(119910)
+
119896
sum
119895=1
intR
120601 (119910 + 120579119895(119910 1199061(119910 119911) 119911)) minus 120601 (119910)
minusnabla120601 (119910) 120579119895(119910 1199061(119910 119911) 119911) ] (119889119911)
(16)
where nabla120601 = (1205971206011205971199101 120597120601120597119910
119896) is the gradient of 120601 and 120579119895
is the 119895th column of the 119896 times 119896matrix 120579Now we can state the main result of this section
Theorem 3 (verification theorem for game (9) and (13))Suppose there exist two functions 120601
119894 S rarr R 119894 = 1 2 such
that
(i) 120601119894isin 1198621(S0)⋂119862(S) 119894 = 1 2 whereS is the closure of
S and S0 is the interior of S(ii) 120601119894ge 119892119894on S 119894 = 1 2
Define the following continuation regions
119863119894= 119910 isin S 120601
119894(119910) gt 119892
119894(119910) 119894 = 1 2 (17)
Suppose 119884(119905) = 119884119906(119905) spends 0 time on 120597119863
119894as 119894 =
1 2 that is(iii) 119864119910[int120591119878
0120594120597119863119894(119884(119905))119889119905] = 0 for all 119910 isin S 119906 isin U 119894 = 1 2
(iv) 120597119863119894is a Lipschitz surface 119894 = 1 2
(v) 120601119894isin 1198622(S120597119863
119894)The second-order derivatives of 120601
119894are
locally bounded near 120597119863119894 respectively 119894 = 1 2
(vi) 1198631= 1198632= 119863
(vii) there exists isin A such that for 119894 = 1 2
119860120601119894(119910) + 119891
119894(119910 (119910))
= sup119906isinA
119860119906120601119894(119910) + 119891
119894(119910 119906 (119910))
= 0 119910 isin 119863
le 0 119910 isin S 119863
(18)
(viii) 119864119910[|120601119894(119884120591)| + int120591
0|119860119906120601119894(119884119906(119905))|119889119905] lt infin 119894 = 1 2 for all
119906 isin U 120591 isin ΓFor 119906 isin A define
120591119863= 120591119906
119863= inf 119905 gt 0 119884
119906(119905) notin 119863 lt infin (19)
and in particular
120591119863= 120591
119863= inf 119905 gt 0 119884
(119905) notin 119863 lt infin (20)
4 Abstract and Applied Analysis
Suppose that(ix) the family 120601
119894(119884(120591)) 120591 isin Γ 120591 le 120591
119863 is uniformly
integrable for all 119906 isin A and 119910 isin S 119894 = 1 2
Then (120591119863 ) isin Γ timesA is a Nash equilibrium for game (9) (13)
and
1206011(119910) = sup
119906isinA
J119906120591119863
1(119910) = J
120591119863
1(119910) (21)
1206012(119910) = sup
120591isinΓ
J120591
2(119910) = J
120591119863
2(119910) (22)
Proof From (i) (iv) and (v) we may assume by an approx-imation theorem (see Theorem 21 in [18]) that 120601
119894isin 1198622(S)
119894 = 1 2For a given 119906 isin A we define with 119884(119905) = 119884
119906(119905)
120591119863= 120591119906
119863= inf 119905 gt 0 119884
119906(119905) notin 119863 (23)
In particular let be as in (vii) Then
120591119863= 120591
119863= inf 119905 gt 0 119884
(119905) notin 119863 (24)
We first prove that (21) holds Let 120591119863isin Γ be as in (24) For
arbitrary 119906 isin A by (vii) and the Dynkinrsquos formula for jumpdiffusion (see Theorem 124 in [18]) we have
1206011(119910) = 119864
119910[int
120591119863and119898
0
minus1198601199061206011(119884 (119905)) 119889119905 + 120601
1(119884 (120591119863and 119898))]
ge 119864119910[int
120591119863and119898
0
1198911(119884 (119905) 119906 (119905)) 119889119905 + 120601
1(119884 (120591119863and 119898))]
(25)
where119898 = 1 2 Therefore by (11) (12) (i) (ii) (viii) andthe Fatou lemma
1206011(119910)
ge lim inf119898rarrinfin
119864119910[int
120591119863and119898
0
1198911(119884 (119905) 119906 (119905)) 119889119905+120601
1(119884 (120591119863and 119898))]
ge 119864119910[int
120591119863
0
1198911(119884 (119905) 119906 (119905)) 119889119905 + 119892
1(119884 (120591119863))]
= J119910
1(119906 120591119863)
(26)
Since this holds for all 119906 isin A we have
1206011(119910) ge sup
119906isinA
J119910
1(119906 120591119863) (27)
In particular applying the Dynkinrsquos formula to 119906 = we getan equality that is
1206011(119910) = 119864
119910[int
120591119863and119898
0
minus1198601206011( (119905)) 119889119905 + 120601
1( (120591119863and 119898))]
= 119864119910[int
120591119863and119898
0
1198911( (119905) (119905)) 119889119905 + 120601
1( (120591119863and 119898))]
(28)
where (119905) = 119884(119905) and119898 = 1 2 Hence we deduce that
1206011(119910) = J
119910
1( 120591119863) (29)
Since we always have
J119910
1( 120591119863) le sup119906isinA
J119910
1(119906 120591119863) (30)
we conclude by combining (27) (29) and (30) that
1206011(119910) = J
119910
1( 120591119863) = sup119906isinA
J119910
1(119906 120591119863) (31)
which is (21)Next we prove that (22) holds Let isin A be as in (vii)
For 120591 isin Γ by the Dynkinrsquos formula and (vii) we have
119864119910[1206012( (120591119898))] = 120601
2(119910) + 119864
119910[int
120591119898
0
1198601206012( (119905)) 119889119905]
le 1206012(119910) minus 119864
119910[int
120591119898
0
1198912( (119905) (119905)) 119889119905]
(32)
where 120591119898= 120591 and 119898119898 = 1 2
Letting119898 rarr infin gives by (11) (12) (i) (ii) (viii) and theFatou Lemma
1206012(119910) ge lim inf
119898rarrinfin119864119910[int
120591and119898
0
1198912( (119905) (119905)) 119889119905 + 120601
2( (120591119898))]
ge 119864119910[int
120591
0
1198912( (119905) (119905)) 119889119905 + 119892
2( (120591)) 120594
120591ltinfin]
= J119910
2( 120591)
(33)
The inequality (33) holds for all 120591 isin Γ Then we have
1206012(119910) ge sup
120591isinΓ
J119910
2( 120591) (34)
Similarly applying the above argument to the pair ( 120591119863) we
get an equality in (34) that is
1206012(119910) = J
119910
2( 120591119863) (35)
We always have
J119910
2( 120591119863) le sup120591isinΓ
J119910
2( 120591) (36)
Therefore combining (34) (35) and (36) we get
1206012(119910) = J
119910
2( 120591119863) = sup120591isinΓ
J119910
2( 120591) (37)
which is (22) The proof is completed
Abstract and Applied Analysis 5
3 An Example
In this section we come back to Example 1 and useTheorem 3to study the solutions of game (4) and (8) Here and in thefollowing all the processes are assumed to be one-dimensionfor simplicity
To put game (4) and (8) into the framework of Section 2we define the process 119884(119905) = (119884
0(119905) 1198841(119905)) 119884(0) = 119910 = (119904 119910
1)
by
1198891198840(119905) = 119889119905 119884
0(0) = 119904 isin R
1198891198841(119905) = 119889119883 (119905)
= 1198841(119905) [(1 minus 119906 (119905)) 119903 (119905) + 119906 (119905) 119887 (119905)] 119889119905
+ 119906 (119905) 120590 (119905) 119889119861 (119905)
+119906 (119905minus) intR
120574 (119905 119911) (119889119905 119889119911)
1198841(0) = 119909 = 119910
1
(38)
Then the performance functionals to the controller (6) andthe stopper (7) can be formulated as follows
J119910
1(119906 120591) = 119864
1199041199101 [119890minus120575(119904+120591)
1198801(119862 (1198841(120591)))
minusint
120591
0
119890minus120575(119904+119905)
ℎ (1198840(119905) 1198841(119905) 119906 (119905)) 119889119905]
J119910
2(119906 120591) = 119864
1199041199101 [119890minus120575(119904+120591)
1198802(1198841(120591) minus 119862 (119884
1(120591)))]
(39)
In this case the generator 119860119906 in (16) has the form
119860119906120601 (119904 119909)
=120597120601
120597119904+ [(1 minus 119906) 119903119909 + 119906119887119909]
120597120601
120597119909+1
2119909211990621205902 1205972120601
1205971199092
+ intR0
120601 (119904 119909 + 119909119906120574 (119911)) minus 120601 (119904 119909) minus 119909119906120574 (119911)120597120601
120597119909
times ] (119889119911)
(40)
To obtain a possible Nash equilibrium ( 120591) isin A times Γ forgame (4) and (8) according to Theorem 3 it is necessary tofind a subset 119863 of S = R2
+= [0infin)
2 and 120601119894(119904 119909) 119894 = 1 2
such that
(i) 1206011(119904 119909) = 119890
minus1205751199041198801(119862(119909)) and 120601
2(119904 119909) = 119890
minus1205751199041198802(119909 minus
119862(119909)) for all (119904 119909) isin 119863(ii) 1206011(119904 119909) ge 119890
minus1205751199041198801(119862(119909)) and 120601
2(119904 119909) ge 119890
minus1205751199041198802(119909 minus
119862(119909)) for all (119904 119909) isin S(iii) 119860119906120601
1(119904 119909) minus ℎ(119904 119909 119906) le 0 and 119860119906120601
2(119904 119909) le 0 for all
(119904 119909) isin S 119863 and 119906(iv) there exists such that 119860120601
1(119904 119909) minus ℎ(119904 119909 ) =
1198601206012(119904 119909) = 0 for all (119904 119909) isin 119863
Imposing the first-order condition on 1198601199061206011(119904 119909) minus
ℎ(119904 119909 119906) and 1198601199061206012(119904 119909) we get the following equations for
the optimal control processes
(119887119909 minus 119903119909)1205971206011
120597119909(119904 119909) + 119909
2120590212059721206011
1205971199092(119904 119909)
+ intR
119903120574 (119911) [1205971206011
120597119909(119904 119909 + 119903120574 (119911)) minus
1205971206011
120597119909(119904 119909)]
minus120597ℎ
120597119906(119904 119909 ) = 0
(119887119909 minus 119903119909)1205971206012
120597119909(119904 119909) + 119909
2120590212059721206012
1205971199092(119904 119909)
+ intR
119903120574 (119911) [1205971206012
120597119909(119904 119909 + 119903120574 (119911)) minus
1205971206012
120597119909(119904 119909)] = 0
(41)
With as in (41) we put
119860120601119894(119904 119909)
=120597120601119894
120597119904+ [(1 minus ) 119903119909 + 119887119909]
120597120601119894
120597119909+1
2119909221205902 1205972120601119894
1205971199092
+ intR0
120601119894(119904 119909 + 119909120574 (119911)) minus 120601
119894(119904 119909) minus 119909120574 (119911)
120597120601119894
120597119909
times ] (119889119911)
(42)
Thus we may reduce game (4) and (8) to the problem ofsolving a family of nonlinear variational-integro inequalitiesWe summarize as follows
Theorem 4 Suppose there exist satisfying (41) and two 1198621-functions 120601
119894 119894 = 1 2 such that
(1)
119863 = (119904 119909) 1206012(119904 119909) gt 119890
minus1205751199041198802(119909 minus 119862 (119909))
= (119904 119909) 1206011(119904 119909) gt 119890
minus1205751199041198801(119862 (119909))
(43)
(2) 120601119894isin 1198622(119863) 119894 = 1 2
(3) 1206012(119904 119909) = 119890
minus1205751199041198802(119909 minus 119862(119909)) and 120601
1(119904 119909) =
119890minus1205751199041198801(119862(119909)) for all (119904 119909) isin S 119863
(4) 1198601199061206011(119904 119909) minus ℎ(119904 119909 119906) le 0 and 119860119906120601
2(119904 119909) le 0 for all
(119904 119909) isin S 119863 and for all 119906 isin A(5) 119860120601
1(119904 119909) minus ℎ(119904 119909 ) = 0 for all (119904 119909) isin 119863 where
1198601206011is given by (42)
(6) 1198601206012(119904 119909) = 0 for all (119904 119909) isin 119863 where 119860120601
2is given
by (42)
Then the pair ( 120591) is a Nash equilibrium of the stochasticdifferential game (4) and (8) where
120591 = inf 119905 gt 0 119884(119905) notin 119863 (44)
6 Abstract and Applied Analysis
Moreover the corresponding equilibrium performances are
1206011(119904 119909) = J
(119904119909)
1( 120591)
1206012(119904 119909) = J
(119904119909)
2( 120591)
(45)
In this paper we will not discuss general solutions of thisfamily of nonlinear variational-integro inequalities Insteadwe discuss a solution in special case when
120574 (119905 119911) = 0 ℎ (119904 119909 119906) =1199062
2 (46)
Let us try the functions 120601119894 119894 = 1 2 of the form
120601119894(119904 119909) = 119890
minus120575119904120595119894(119909) (47)
and a continuation region119863 of the form
119863 = (119904 119909) 119909 lt 1199090 for some 119909
0gt 0 (48)
Then we have
119860119906120601119894(119904 119909) = 119890
minus120575119904119860119906120595119894(119909) (49)
where119860119906120595119894(119909) = minus 120575120595
119894(119909) + [(1 minus 119906) 119903119909 + 119906119887119909] 120595
1015840
119894(119909)
+1
211990921199062120590212059510158401015840
119894(119909)
(50)
By conditions (1) and (3) in Theorem 4 we get
1205951(119909) = 119880
1(119862 (119909)) 119909 ge 119909
0
1205951(119909) gt 119880
1(119862 (119909)) 0 lt 119909 lt 119909
0
1205952(119909) = 119880
2(119909 minus 119862 (119909)) 119909 ge 119909
0
1205952(119909) gt 119880
2(119909 minus 119862 (119909)) 0 lt 119909 lt 119909
0
(51)
From conditions (4) (5) and (6) ofTheorem 4 we get thecandidate for the optimal control as follows
= Argmax119906isinA
1198601199061205951(119909) minus
1199062
2
= Argmax119906isinA
minus 1205751205951(119909) + [(1 minus 119906) 119903119909 + 119906119887119909] 120595
1015840
1(119909)
+1
211990921199062120590212059510158401015840
1(119909) minus
1199062
2
=(119887119909 minus 119903119909) 120595
1015840
1(119909)
1 minus 1199092120590212059510158401015840
1(119909)
(52)
= Argmax119906isinA
1198601199061205952(119909)
= Argmax119906isinA
minus 1205751205952(119909) + [(1 minus 119906) 119903119909 + 119906119887119909] 120595
1015840
2(119909)
+1
211990921199062120590212059510158401015840
2(119909)
=minus (119887119909 minus 119903119909) 120595
1015840
2(119909)
1199092120590212059510158401015840
2(119909)
(53)
Let120595119894(119909) =
119894(119909) on 0 lt 119909 lt 119909
0 119894 = 1 2 By condition (5)
in Theorem 4 we have 1198601(119909) minus
22 = 0 for 0 lt 119909 lt 119909
0
Substituting (52) into 1198601199061(119909) minus 119906
22 = 0 we obtain
minus 1205751(119909) + [119903119909 +
(119887119909 minus 119903119909)21015840
1(119909)
1 minus 1199092120590210158401015840
1(119909)
] 1015840
1(119909)
+
[(119887119909 minus 119903119909) 1015840
1(119909)]2
2(1 minus 1199092120590210158401015840
1(119909))2[1199092120590210158401015840
1(119909) minus 1] = 0
(54)
Similarly we obtain1198602(119909) = 0 for 0 lt 119909 lt 119909
0by condition
(6) in Theorem 4 And we substitute (53) in 1198601199062(119909) = 0 to
get
minus 1205752(119909) + 119903119909
1015840
2(119909) +
[(119887119909 minus 119903119909) 1015840
2(119909)]2
1 minus 1199092120590210158401015840
2(119909)
+
[119909120590 (119887119909 minus 119903119909) 1015840
2(119909)]2
10158401015840
2(119909)
2(1 minus 1199092120590210158401015840
2(119909))2
= 0
(55)
Therefore we conclude that
1205951(119909) =
1198801(119862 (119909)) 119909 ge 119909
0
1(119909) 0 lt 119909 lt 119909
0
(56)
1205952(119909) =
1198802(119909 minus 119862 (119909)) 119909 ge 119909
0
2(119909) 0 lt 119909 lt 119909
0
(57)
where 1(119909) and
2(119909) are the solutions of (56) and (57)
respectivelyAccording to Theorem 4 we use the continuity and
differentiability of 120595119894at 119909 = 119909
0to determine 119909
0 119894 = 1 2 that
is
1205951(1199090) = 1198801(119862 (1199090))
1205951015840
1(1199090) = 119880
1015840
1(119862 (1199090)) 1198621015840(1199090)
1205952(1199090) = 1198802(1199090minus 119862 (119909
0))
1205951015840
2(1199090) = 119880
1015840
2(1199090minus 119862 (119909
0)) (1 minus 119862
1015840(1199090))
(58)
At the end of this section we summarize the above resultsin the following theorem
Theorem 5 Let 120595119894 119894 = 1 2 and let 119909
0be the solutions of
equations (56)ndash(58) Then the pair ( 120591) given by
=(119887119909 minus 119903119909) 120595
1015840
1(119909)
1 minus 1199092120590212059510158401015840
1(119909)
=minus (119887119909 minus 119903119909) 120595
1015840
2(119909)
1199092120590212059510158401015840
2(119909)
120591 = inf 119905 gt 0 119909 ge 1199090
(59)
is a Nash equilibrium of game (4) and (8) The correspondingequilibrium performances are
120601119894(119904 119909) = 119890
minus120575119904120595119894(119909) 119894 = 1 2 (60)
Abstract and Applied Analysis 7
4 Conclusion
A verification theorem is obtained for the general stochasticdifferential game between a controller and a stopper Inthe special case of quadratic cost we use this theorem tocharacterize the Nash equilibrium However the question ofthe existence and uniqueness of Nash equilibrium for thegame remains open It will be considered in our subsequentwork
Acknowledgment
This work was supported by the National Natural ScienceFoundation of China under Grant 61273022 and 11171050
References
[1] J Bertoin Levy Processes Cambridge University Press Cam-bridge UK 1996
[2] D Applebaum Levy Processes and Stochastic Calculus Cam-bridge University Press Cambridge UK 2003
[3] S Mataramvura and B Oslashksendal ldquoRisk minimizing portfoliosand HJBI equations for stochastic differential gamesrdquo Stochas-tics vol 80 no 4 pp 317ndash337 2008
[4] R J Elliott and T K Siu ldquoOn risk minimizing portfolios underaMarkovian regime-switching Black-Scholes economyrdquoAnnalsof Operations Research vol 176 pp 271ndash291 2010
[5] G Wang and Z Wu ldquoMean-variance hedging and forward-backward stochastic differential filtering equationsrdquo Abstractand Applied Analysis vol 2011 Article ID 310910 20 pages 2011
[6] I Karatzas and W Sudderth ldquoStochastic games of control andstopping for a linear diffusionrdquo in Random Walk SequentialAnalysis and Related Topics A Festschrift in Honor of YS ChowA Hsiung Z L Ying and C H Zhang Eds pp 100ndash117WorldScientific Publishers 2006
[7] S Hamadene ldquoMixed zero-sum stochastic differential gameand American game optionsrdquo SIAM Journal on Control andOptimization vol 45 no 2 pp 496ndash518 2006
[8] M K Ghosh M K S Rao and D Sheetal ldquoDifferential gamesof mixed type with control and stopping timesrdquo NonlinearDifferential Equations and Applications vol 16 no 2 pp 143ndash158 2009
[9] M K Ghosh and K S Mallikarjuna Rao ldquoExistence of value instochastic differential games of mixed typerdquo Stochastic Analysisand Applications vol 30 no 5 pp 895ndash905 2012
[10] I Karatzas and Q Li ldquoBSDE approach to non-zero-sumstochastic differential games of control and stoppingrdquo inStochastic Processes Finance and Control pp 105ndash153 WorldScientific Publishers 2012
[11] J-P Lepeltier ldquoOn a general zero-sum stochastic game withstopping strategy for one player and continuous strategy for theotherrdquo Probability and Mathematical Statistics vol 6 no 1 pp43ndash50 1985
[12] I Karatzas and W D Sudderth ldquoThe controller-and-stoppergame for a linear diffusionrdquo The Annals of Probability vol 29no 3 pp 1111ndash1127 2001
[13] AWeerasinghe ldquoA controller and a stopper gamewith degener-ate variance controlrdquo Electronic Communications in Probabilityvol 11 pp 89ndash99 2006
[14] I Karatzas and I-M Zamfirescu ldquoMartingale approach tostochastic differential games of control and stoppingrdquo TheAnnals of Probability vol 36 no 4 pp 1495ndash1527 2008
[15] E Bayraktar I Karatzas and S Yao ldquoOptimal stopping fordynamic convex risk measuresrdquo Illinois Journal of Mathematicsvol 54 no 3 pp 1025ndash1067 2010
[16] E Bayraktar and V R Young ldquoProving regularity of theminimal probability of ruin via a game of stopping and controlrdquoFinance and Stochastics vol 15 no 4 pp 785ndash818 2011
[17] F Bagherya S Haademb B Oslashksendal and I Turpina ldquoOptimalstopping and stochastic control differential games for jumpdiffusionsrdquo Stochastics vol 85 no 1 pp 85ndash97 2013
[18] B Oslashksendal and A Sulem Applied Stochastic Control of JumpDiffusions Springer Berlin Germany 2nd edition 2007
[19] Q Lin R Loxton K L Teo and Y H Wu ldquoA new computa-tional method for a class of free terminal time optimal controlproblemsrdquo Pacific Journal of Optimization vol 7 no 1 pp 63ndash81 2011
[20] Q Lin R Loxton K L Teo and Y H Wu ldquoOptimal controlcomputation for nonlinear systems with state-dependent stop-ping criteriardquo Automatica vol 48 no 9 pp 2116ndash2129 2012
[21] C Jiang Q Lin C Yu K L Teo and G-R Duan ldquoAn exactpenalty method for free terminal time optimal control problemwith continuous inequality constraintsrdquo Journal of OptimizationTheory and Applications vol 154 no 1 pp 30ndash53 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
![Page 4: Research Article Nonzero-Sum Stochastic Differential Game ...downloads.hindawi.com/journals/aaa/2013/761306.pdf · Controller and Stopper for Jump Diffusions YanWang, 1,2 AiminSong,](https://reader033.vdocuments.mx/reader033/viewer/2022042410/5f26fe8f7005a2296769470b/html5/thumbnails/4.jpg)
4 Abstract and Applied Analysis
Suppose that(ix) the family 120601
119894(119884(120591)) 120591 isin Γ 120591 le 120591
119863 is uniformly
integrable for all 119906 isin A and 119910 isin S 119894 = 1 2
Then (120591119863 ) isin Γ timesA is a Nash equilibrium for game (9) (13)
and
1206011(119910) = sup
119906isinA
J119906120591119863
1(119910) = J
120591119863
1(119910) (21)
1206012(119910) = sup
120591isinΓ
J120591
2(119910) = J
120591119863
2(119910) (22)
Proof From (i) (iv) and (v) we may assume by an approx-imation theorem (see Theorem 21 in [18]) that 120601
119894isin 1198622(S)
119894 = 1 2For a given 119906 isin A we define with 119884(119905) = 119884
119906(119905)
120591119863= 120591119906
119863= inf 119905 gt 0 119884
119906(119905) notin 119863 (23)
In particular let be as in (vii) Then
120591119863= 120591
119863= inf 119905 gt 0 119884
(119905) notin 119863 (24)
We first prove that (21) holds Let 120591119863isin Γ be as in (24) For
arbitrary 119906 isin A by (vii) and the Dynkinrsquos formula for jumpdiffusion (see Theorem 124 in [18]) we have
1206011(119910) = 119864
119910[int
120591119863and119898
0
minus1198601199061206011(119884 (119905)) 119889119905 + 120601
1(119884 (120591119863and 119898))]
ge 119864119910[int
120591119863and119898
0
1198911(119884 (119905) 119906 (119905)) 119889119905 + 120601
1(119884 (120591119863and 119898))]
(25)
where119898 = 1 2 Therefore by (11) (12) (i) (ii) (viii) andthe Fatou lemma
1206011(119910)
ge lim inf119898rarrinfin
119864119910[int
120591119863and119898
0
1198911(119884 (119905) 119906 (119905)) 119889119905+120601
1(119884 (120591119863and 119898))]
ge 119864119910[int
120591119863
0
1198911(119884 (119905) 119906 (119905)) 119889119905 + 119892
1(119884 (120591119863))]
= J119910
1(119906 120591119863)
(26)
Since this holds for all 119906 isin A we have
1206011(119910) ge sup
119906isinA
J119910
1(119906 120591119863) (27)
In particular applying the Dynkinrsquos formula to 119906 = we getan equality that is
1206011(119910) = 119864
119910[int
120591119863and119898
0
minus1198601206011( (119905)) 119889119905 + 120601
1( (120591119863and 119898))]
= 119864119910[int
120591119863and119898
0
1198911( (119905) (119905)) 119889119905 + 120601
1( (120591119863and 119898))]
(28)
where (119905) = 119884(119905) and119898 = 1 2 Hence we deduce that
1206011(119910) = J
119910
1( 120591119863) (29)
Since we always have
J119910
1( 120591119863) le sup119906isinA
J119910
1(119906 120591119863) (30)
we conclude by combining (27) (29) and (30) that
1206011(119910) = J
119910
1( 120591119863) = sup119906isinA
J119910
1(119906 120591119863) (31)
which is (21)Next we prove that (22) holds Let isin A be as in (vii)
For 120591 isin Γ by the Dynkinrsquos formula and (vii) we have
119864119910[1206012( (120591119898))] = 120601
2(119910) + 119864
119910[int
120591119898
0
1198601206012( (119905)) 119889119905]
le 1206012(119910) minus 119864
119910[int
120591119898
0
1198912( (119905) (119905)) 119889119905]
(32)
where 120591119898= 120591 and 119898119898 = 1 2
Letting119898 rarr infin gives by (11) (12) (i) (ii) (viii) and theFatou Lemma
1206012(119910) ge lim inf
119898rarrinfin119864119910[int
120591and119898
0
1198912( (119905) (119905)) 119889119905 + 120601
2( (120591119898))]
ge 119864119910[int
120591
0
1198912( (119905) (119905)) 119889119905 + 119892
2( (120591)) 120594
120591ltinfin]
= J119910
2( 120591)
(33)
The inequality (33) holds for all 120591 isin Γ Then we have
1206012(119910) ge sup
120591isinΓ
J119910
2( 120591) (34)
Similarly applying the above argument to the pair ( 120591119863) we
get an equality in (34) that is
1206012(119910) = J
119910
2( 120591119863) (35)
We always have
J119910
2( 120591119863) le sup120591isinΓ
J119910
2( 120591) (36)
Therefore combining (34) (35) and (36) we get
1206012(119910) = J
119910
2( 120591119863) = sup120591isinΓ
J119910
2( 120591) (37)
which is (22) The proof is completed
Abstract and Applied Analysis 5
3 An Example
In this section we come back to Example 1 and useTheorem 3to study the solutions of game (4) and (8) Here and in thefollowing all the processes are assumed to be one-dimensionfor simplicity
To put game (4) and (8) into the framework of Section 2we define the process 119884(119905) = (119884
0(119905) 1198841(119905)) 119884(0) = 119910 = (119904 119910
1)
by
1198891198840(119905) = 119889119905 119884
0(0) = 119904 isin R
1198891198841(119905) = 119889119883 (119905)
= 1198841(119905) [(1 minus 119906 (119905)) 119903 (119905) + 119906 (119905) 119887 (119905)] 119889119905
+ 119906 (119905) 120590 (119905) 119889119861 (119905)
+119906 (119905minus) intR
120574 (119905 119911) (119889119905 119889119911)
1198841(0) = 119909 = 119910
1
(38)
Then the performance functionals to the controller (6) andthe stopper (7) can be formulated as follows
J119910
1(119906 120591) = 119864
1199041199101 [119890minus120575(119904+120591)
1198801(119862 (1198841(120591)))
minusint
120591
0
119890minus120575(119904+119905)
ℎ (1198840(119905) 1198841(119905) 119906 (119905)) 119889119905]
J119910
2(119906 120591) = 119864
1199041199101 [119890minus120575(119904+120591)
1198802(1198841(120591) minus 119862 (119884
1(120591)))]
(39)
In this case the generator 119860119906 in (16) has the form
119860119906120601 (119904 119909)
=120597120601
120597119904+ [(1 minus 119906) 119903119909 + 119906119887119909]
120597120601
120597119909+1
2119909211990621205902 1205972120601
1205971199092
+ intR0
120601 (119904 119909 + 119909119906120574 (119911)) minus 120601 (119904 119909) minus 119909119906120574 (119911)120597120601
120597119909
times ] (119889119911)
(40)
To obtain a possible Nash equilibrium ( 120591) isin A times Γ forgame (4) and (8) according to Theorem 3 it is necessary tofind a subset 119863 of S = R2
+= [0infin)
2 and 120601119894(119904 119909) 119894 = 1 2
such that
(i) 1206011(119904 119909) = 119890
minus1205751199041198801(119862(119909)) and 120601
2(119904 119909) = 119890
minus1205751199041198802(119909 minus
119862(119909)) for all (119904 119909) isin 119863(ii) 1206011(119904 119909) ge 119890
minus1205751199041198801(119862(119909)) and 120601
2(119904 119909) ge 119890
minus1205751199041198802(119909 minus
119862(119909)) for all (119904 119909) isin S(iii) 119860119906120601
1(119904 119909) minus ℎ(119904 119909 119906) le 0 and 119860119906120601
2(119904 119909) le 0 for all
(119904 119909) isin S 119863 and 119906(iv) there exists such that 119860120601
1(119904 119909) minus ℎ(119904 119909 ) =
1198601206012(119904 119909) = 0 for all (119904 119909) isin 119863
Imposing the first-order condition on 1198601199061206011(119904 119909) minus
ℎ(119904 119909 119906) and 1198601199061206012(119904 119909) we get the following equations for
the optimal control processes
(119887119909 minus 119903119909)1205971206011
120597119909(119904 119909) + 119909
2120590212059721206011
1205971199092(119904 119909)
+ intR
119903120574 (119911) [1205971206011
120597119909(119904 119909 + 119903120574 (119911)) minus
1205971206011
120597119909(119904 119909)]
minus120597ℎ
120597119906(119904 119909 ) = 0
(119887119909 minus 119903119909)1205971206012
120597119909(119904 119909) + 119909
2120590212059721206012
1205971199092(119904 119909)
+ intR
119903120574 (119911) [1205971206012
120597119909(119904 119909 + 119903120574 (119911)) minus
1205971206012
120597119909(119904 119909)] = 0
(41)
With as in (41) we put
119860120601119894(119904 119909)
=120597120601119894
120597119904+ [(1 minus ) 119903119909 + 119887119909]
120597120601119894
120597119909+1
2119909221205902 1205972120601119894
1205971199092
+ intR0
120601119894(119904 119909 + 119909120574 (119911)) minus 120601
119894(119904 119909) minus 119909120574 (119911)
120597120601119894
120597119909
times ] (119889119911)
(42)
Thus we may reduce game (4) and (8) to the problem ofsolving a family of nonlinear variational-integro inequalitiesWe summarize as follows
Theorem 4 Suppose there exist satisfying (41) and two 1198621-functions 120601
119894 119894 = 1 2 such that
(1)
119863 = (119904 119909) 1206012(119904 119909) gt 119890
minus1205751199041198802(119909 minus 119862 (119909))
= (119904 119909) 1206011(119904 119909) gt 119890
minus1205751199041198801(119862 (119909))
(43)
(2) 120601119894isin 1198622(119863) 119894 = 1 2
(3) 1206012(119904 119909) = 119890
minus1205751199041198802(119909 minus 119862(119909)) and 120601
1(119904 119909) =
119890minus1205751199041198801(119862(119909)) for all (119904 119909) isin S 119863
(4) 1198601199061206011(119904 119909) minus ℎ(119904 119909 119906) le 0 and 119860119906120601
2(119904 119909) le 0 for all
(119904 119909) isin S 119863 and for all 119906 isin A(5) 119860120601
1(119904 119909) minus ℎ(119904 119909 ) = 0 for all (119904 119909) isin 119863 where
1198601206011is given by (42)
(6) 1198601206012(119904 119909) = 0 for all (119904 119909) isin 119863 where 119860120601
2is given
by (42)
Then the pair ( 120591) is a Nash equilibrium of the stochasticdifferential game (4) and (8) where
120591 = inf 119905 gt 0 119884(119905) notin 119863 (44)
6 Abstract and Applied Analysis
Moreover the corresponding equilibrium performances are
1206011(119904 119909) = J
(119904119909)
1( 120591)
1206012(119904 119909) = J
(119904119909)
2( 120591)
(45)
In this paper we will not discuss general solutions of thisfamily of nonlinear variational-integro inequalities Insteadwe discuss a solution in special case when
120574 (119905 119911) = 0 ℎ (119904 119909 119906) =1199062
2 (46)
Let us try the functions 120601119894 119894 = 1 2 of the form
120601119894(119904 119909) = 119890
minus120575119904120595119894(119909) (47)
and a continuation region119863 of the form
119863 = (119904 119909) 119909 lt 1199090 for some 119909
0gt 0 (48)
Then we have
119860119906120601119894(119904 119909) = 119890
minus120575119904119860119906120595119894(119909) (49)
where119860119906120595119894(119909) = minus 120575120595
119894(119909) + [(1 minus 119906) 119903119909 + 119906119887119909] 120595
1015840
119894(119909)
+1
211990921199062120590212059510158401015840
119894(119909)
(50)
By conditions (1) and (3) in Theorem 4 we get
1205951(119909) = 119880
1(119862 (119909)) 119909 ge 119909
0
1205951(119909) gt 119880
1(119862 (119909)) 0 lt 119909 lt 119909
0
1205952(119909) = 119880
2(119909 minus 119862 (119909)) 119909 ge 119909
0
1205952(119909) gt 119880
2(119909 minus 119862 (119909)) 0 lt 119909 lt 119909
0
(51)
From conditions (4) (5) and (6) ofTheorem 4 we get thecandidate for the optimal control as follows
= Argmax119906isinA
1198601199061205951(119909) minus
1199062
2
= Argmax119906isinA
minus 1205751205951(119909) + [(1 minus 119906) 119903119909 + 119906119887119909] 120595
1015840
1(119909)
+1
211990921199062120590212059510158401015840
1(119909) minus
1199062
2
=(119887119909 minus 119903119909) 120595
1015840
1(119909)
1 minus 1199092120590212059510158401015840
1(119909)
(52)
= Argmax119906isinA
1198601199061205952(119909)
= Argmax119906isinA
minus 1205751205952(119909) + [(1 minus 119906) 119903119909 + 119906119887119909] 120595
1015840
2(119909)
+1
211990921199062120590212059510158401015840
2(119909)
=minus (119887119909 minus 119903119909) 120595
1015840
2(119909)
1199092120590212059510158401015840
2(119909)
(53)
Let120595119894(119909) =
119894(119909) on 0 lt 119909 lt 119909
0 119894 = 1 2 By condition (5)
in Theorem 4 we have 1198601(119909) minus
22 = 0 for 0 lt 119909 lt 119909
0
Substituting (52) into 1198601199061(119909) minus 119906
22 = 0 we obtain
minus 1205751(119909) + [119903119909 +
(119887119909 minus 119903119909)21015840
1(119909)
1 minus 1199092120590210158401015840
1(119909)
] 1015840
1(119909)
+
[(119887119909 minus 119903119909) 1015840
1(119909)]2
2(1 minus 1199092120590210158401015840
1(119909))2[1199092120590210158401015840
1(119909) minus 1] = 0
(54)
Similarly we obtain1198602(119909) = 0 for 0 lt 119909 lt 119909
0by condition
(6) in Theorem 4 And we substitute (53) in 1198601199062(119909) = 0 to
get
minus 1205752(119909) + 119903119909
1015840
2(119909) +
[(119887119909 minus 119903119909) 1015840
2(119909)]2
1 minus 1199092120590210158401015840
2(119909)
+
[119909120590 (119887119909 minus 119903119909) 1015840
2(119909)]2
10158401015840
2(119909)
2(1 minus 1199092120590210158401015840
2(119909))2
= 0
(55)
Therefore we conclude that
1205951(119909) =
1198801(119862 (119909)) 119909 ge 119909
0
1(119909) 0 lt 119909 lt 119909
0
(56)
1205952(119909) =
1198802(119909 minus 119862 (119909)) 119909 ge 119909
0
2(119909) 0 lt 119909 lt 119909
0
(57)
where 1(119909) and
2(119909) are the solutions of (56) and (57)
respectivelyAccording to Theorem 4 we use the continuity and
differentiability of 120595119894at 119909 = 119909
0to determine 119909
0 119894 = 1 2 that
is
1205951(1199090) = 1198801(119862 (1199090))
1205951015840
1(1199090) = 119880
1015840
1(119862 (1199090)) 1198621015840(1199090)
1205952(1199090) = 1198802(1199090minus 119862 (119909
0))
1205951015840
2(1199090) = 119880
1015840
2(1199090minus 119862 (119909
0)) (1 minus 119862
1015840(1199090))
(58)
At the end of this section we summarize the above resultsin the following theorem
Theorem 5 Let 120595119894 119894 = 1 2 and let 119909
0be the solutions of
equations (56)ndash(58) Then the pair ( 120591) given by
=(119887119909 minus 119903119909) 120595
1015840
1(119909)
1 minus 1199092120590212059510158401015840
1(119909)
=minus (119887119909 minus 119903119909) 120595
1015840
2(119909)
1199092120590212059510158401015840
2(119909)
120591 = inf 119905 gt 0 119909 ge 1199090
(59)
is a Nash equilibrium of game (4) and (8) The correspondingequilibrium performances are
120601119894(119904 119909) = 119890
minus120575119904120595119894(119909) 119894 = 1 2 (60)
Abstract and Applied Analysis 7
4 Conclusion
A verification theorem is obtained for the general stochasticdifferential game between a controller and a stopper Inthe special case of quadratic cost we use this theorem tocharacterize the Nash equilibrium However the question ofthe existence and uniqueness of Nash equilibrium for thegame remains open It will be considered in our subsequentwork
Acknowledgment
This work was supported by the National Natural ScienceFoundation of China under Grant 61273022 and 11171050
References
[1] J Bertoin Levy Processes Cambridge University Press Cam-bridge UK 1996
[2] D Applebaum Levy Processes and Stochastic Calculus Cam-bridge University Press Cambridge UK 2003
[3] S Mataramvura and B Oslashksendal ldquoRisk minimizing portfoliosand HJBI equations for stochastic differential gamesrdquo Stochas-tics vol 80 no 4 pp 317ndash337 2008
[4] R J Elliott and T K Siu ldquoOn risk minimizing portfolios underaMarkovian regime-switching Black-Scholes economyrdquoAnnalsof Operations Research vol 176 pp 271ndash291 2010
[5] G Wang and Z Wu ldquoMean-variance hedging and forward-backward stochastic differential filtering equationsrdquo Abstractand Applied Analysis vol 2011 Article ID 310910 20 pages 2011
[6] I Karatzas and W Sudderth ldquoStochastic games of control andstopping for a linear diffusionrdquo in Random Walk SequentialAnalysis and Related Topics A Festschrift in Honor of YS ChowA Hsiung Z L Ying and C H Zhang Eds pp 100ndash117WorldScientific Publishers 2006
[7] S Hamadene ldquoMixed zero-sum stochastic differential gameand American game optionsrdquo SIAM Journal on Control andOptimization vol 45 no 2 pp 496ndash518 2006
[8] M K Ghosh M K S Rao and D Sheetal ldquoDifferential gamesof mixed type with control and stopping timesrdquo NonlinearDifferential Equations and Applications vol 16 no 2 pp 143ndash158 2009
[9] M K Ghosh and K S Mallikarjuna Rao ldquoExistence of value instochastic differential games of mixed typerdquo Stochastic Analysisand Applications vol 30 no 5 pp 895ndash905 2012
[10] I Karatzas and Q Li ldquoBSDE approach to non-zero-sumstochastic differential games of control and stoppingrdquo inStochastic Processes Finance and Control pp 105ndash153 WorldScientific Publishers 2012
[11] J-P Lepeltier ldquoOn a general zero-sum stochastic game withstopping strategy for one player and continuous strategy for theotherrdquo Probability and Mathematical Statistics vol 6 no 1 pp43ndash50 1985
[12] I Karatzas and W D Sudderth ldquoThe controller-and-stoppergame for a linear diffusionrdquo The Annals of Probability vol 29no 3 pp 1111ndash1127 2001
[13] AWeerasinghe ldquoA controller and a stopper gamewith degener-ate variance controlrdquo Electronic Communications in Probabilityvol 11 pp 89ndash99 2006
[14] I Karatzas and I-M Zamfirescu ldquoMartingale approach tostochastic differential games of control and stoppingrdquo TheAnnals of Probability vol 36 no 4 pp 1495ndash1527 2008
[15] E Bayraktar I Karatzas and S Yao ldquoOptimal stopping fordynamic convex risk measuresrdquo Illinois Journal of Mathematicsvol 54 no 3 pp 1025ndash1067 2010
[16] E Bayraktar and V R Young ldquoProving regularity of theminimal probability of ruin via a game of stopping and controlrdquoFinance and Stochastics vol 15 no 4 pp 785ndash818 2011
[17] F Bagherya S Haademb B Oslashksendal and I Turpina ldquoOptimalstopping and stochastic control differential games for jumpdiffusionsrdquo Stochastics vol 85 no 1 pp 85ndash97 2013
[18] B Oslashksendal and A Sulem Applied Stochastic Control of JumpDiffusions Springer Berlin Germany 2nd edition 2007
[19] Q Lin R Loxton K L Teo and Y H Wu ldquoA new computa-tional method for a class of free terminal time optimal controlproblemsrdquo Pacific Journal of Optimization vol 7 no 1 pp 63ndash81 2011
[20] Q Lin R Loxton K L Teo and Y H Wu ldquoOptimal controlcomputation for nonlinear systems with state-dependent stop-ping criteriardquo Automatica vol 48 no 9 pp 2116ndash2129 2012
[21] C Jiang Q Lin C Yu K L Teo and G-R Duan ldquoAn exactpenalty method for free terminal time optimal control problemwith continuous inequality constraintsrdquo Journal of OptimizationTheory and Applications vol 154 no 1 pp 30ndash53 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
![Page 5: Research Article Nonzero-Sum Stochastic Differential Game ...downloads.hindawi.com/journals/aaa/2013/761306.pdf · Controller and Stopper for Jump Diffusions YanWang, 1,2 AiminSong,](https://reader033.vdocuments.mx/reader033/viewer/2022042410/5f26fe8f7005a2296769470b/html5/thumbnails/5.jpg)
Abstract and Applied Analysis 5
3 An Example
In this section we come back to Example 1 and useTheorem 3to study the solutions of game (4) and (8) Here and in thefollowing all the processes are assumed to be one-dimensionfor simplicity
To put game (4) and (8) into the framework of Section 2we define the process 119884(119905) = (119884
0(119905) 1198841(119905)) 119884(0) = 119910 = (119904 119910
1)
by
1198891198840(119905) = 119889119905 119884
0(0) = 119904 isin R
1198891198841(119905) = 119889119883 (119905)
= 1198841(119905) [(1 minus 119906 (119905)) 119903 (119905) + 119906 (119905) 119887 (119905)] 119889119905
+ 119906 (119905) 120590 (119905) 119889119861 (119905)
+119906 (119905minus) intR
120574 (119905 119911) (119889119905 119889119911)
1198841(0) = 119909 = 119910
1
(38)
Then the performance functionals to the controller (6) andthe stopper (7) can be formulated as follows
J119910
1(119906 120591) = 119864
1199041199101 [119890minus120575(119904+120591)
1198801(119862 (1198841(120591)))
minusint
120591
0
119890minus120575(119904+119905)
ℎ (1198840(119905) 1198841(119905) 119906 (119905)) 119889119905]
J119910
2(119906 120591) = 119864
1199041199101 [119890minus120575(119904+120591)
1198802(1198841(120591) minus 119862 (119884
1(120591)))]
(39)
In this case the generator 119860119906 in (16) has the form
119860119906120601 (119904 119909)
=120597120601
120597119904+ [(1 minus 119906) 119903119909 + 119906119887119909]
120597120601
120597119909+1
2119909211990621205902 1205972120601
1205971199092
+ intR0
120601 (119904 119909 + 119909119906120574 (119911)) minus 120601 (119904 119909) minus 119909119906120574 (119911)120597120601
120597119909
times ] (119889119911)
(40)
To obtain a possible Nash equilibrium ( 120591) isin A times Γ forgame (4) and (8) according to Theorem 3 it is necessary tofind a subset 119863 of S = R2
+= [0infin)
2 and 120601119894(119904 119909) 119894 = 1 2
such that
(i) 1206011(119904 119909) = 119890
minus1205751199041198801(119862(119909)) and 120601
2(119904 119909) = 119890
minus1205751199041198802(119909 minus
119862(119909)) for all (119904 119909) isin 119863(ii) 1206011(119904 119909) ge 119890
minus1205751199041198801(119862(119909)) and 120601
2(119904 119909) ge 119890
minus1205751199041198802(119909 minus
119862(119909)) for all (119904 119909) isin S(iii) 119860119906120601
1(119904 119909) minus ℎ(119904 119909 119906) le 0 and 119860119906120601
2(119904 119909) le 0 for all
(119904 119909) isin S 119863 and 119906(iv) there exists such that 119860120601
1(119904 119909) minus ℎ(119904 119909 ) =
1198601206012(119904 119909) = 0 for all (119904 119909) isin 119863
Imposing the first-order condition on 1198601199061206011(119904 119909) minus
ℎ(119904 119909 119906) and 1198601199061206012(119904 119909) we get the following equations for
the optimal control processes
(119887119909 minus 119903119909)1205971206011
120597119909(119904 119909) + 119909
2120590212059721206011
1205971199092(119904 119909)
+ intR
119903120574 (119911) [1205971206011
120597119909(119904 119909 + 119903120574 (119911)) minus
1205971206011
120597119909(119904 119909)]
minus120597ℎ
120597119906(119904 119909 ) = 0
(119887119909 minus 119903119909)1205971206012
120597119909(119904 119909) + 119909
2120590212059721206012
1205971199092(119904 119909)
+ intR
119903120574 (119911) [1205971206012
120597119909(119904 119909 + 119903120574 (119911)) minus
1205971206012
120597119909(119904 119909)] = 0
(41)
With as in (41) we put
119860120601119894(119904 119909)
=120597120601119894
120597119904+ [(1 minus ) 119903119909 + 119887119909]
120597120601119894
120597119909+1
2119909221205902 1205972120601119894
1205971199092
+ intR0
120601119894(119904 119909 + 119909120574 (119911)) minus 120601
119894(119904 119909) minus 119909120574 (119911)
120597120601119894
120597119909
times ] (119889119911)
(42)
Thus we may reduce game (4) and (8) to the problem ofsolving a family of nonlinear variational-integro inequalitiesWe summarize as follows
Theorem 4 Suppose there exist satisfying (41) and two 1198621-functions 120601
119894 119894 = 1 2 such that
(1)
119863 = (119904 119909) 1206012(119904 119909) gt 119890
minus1205751199041198802(119909 minus 119862 (119909))
= (119904 119909) 1206011(119904 119909) gt 119890
minus1205751199041198801(119862 (119909))
(43)
(2) 120601119894isin 1198622(119863) 119894 = 1 2
(3) 1206012(119904 119909) = 119890
minus1205751199041198802(119909 minus 119862(119909)) and 120601
1(119904 119909) =
119890minus1205751199041198801(119862(119909)) for all (119904 119909) isin S 119863
(4) 1198601199061206011(119904 119909) minus ℎ(119904 119909 119906) le 0 and 119860119906120601
2(119904 119909) le 0 for all
(119904 119909) isin S 119863 and for all 119906 isin A(5) 119860120601
1(119904 119909) minus ℎ(119904 119909 ) = 0 for all (119904 119909) isin 119863 where
1198601206011is given by (42)
(6) 1198601206012(119904 119909) = 0 for all (119904 119909) isin 119863 where 119860120601
2is given
by (42)
Then the pair ( 120591) is a Nash equilibrium of the stochasticdifferential game (4) and (8) where
120591 = inf 119905 gt 0 119884(119905) notin 119863 (44)
6 Abstract and Applied Analysis
Moreover the corresponding equilibrium performances are
1206011(119904 119909) = J
(119904119909)
1( 120591)
1206012(119904 119909) = J
(119904119909)
2( 120591)
(45)
In this paper we will not discuss general solutions of thisfamily of nonlinear variational-integro inequalities Insteadwe discuss a solution in special case when
120574 (119905 119911) = 0 ℎ (119904 119909 119906) =1199062
2 (46)
Let us try the functions 120601119894 119894 = 1 2 of the form
120601119894(119904 119909) = 119890
minus120575119904120595119894(119909) (47)
and a continuation region119863 of the form
119863 = (119904 119909) 119909 lt 1199090 for some 119909
0gt 0 (48)
Then we have
119860119906120601119894(119904 119909) = 119890
minus120575119904119860119906120595119894(119909) (49)
where119860119906120595119894(119909) = minus 120575120595
119894(119909) + [(1 minus 119906) 119903119909 + 119906119887119909] 120595
1015840
119894(119909)
+1
211990921199062120590212059510158401015840
119894(119909)
(50)
By conditions (1) and (3) in Theorem 4 we get
1205951(119909) = 119880
1(119862 (119909)) 119909 ge 119909
0
1205951(119909) gt 119880
1(119862 (119909)) 0 lt 119909 lt 119909
0
1205952(119909) = 119880
2(119909 minus 119862 (119909)) 119909 ge 119909
0
1205952(119909) gt 119880
2(119909 minus 119862 (119909)) 0 lt 119909 lt 119909
0
(51)
From conditions (4) (5) and (6) ofTheorem 4 we get thecandidate for the optimal control as follows
= Argmax119906isinA
1198601199061205951(119909) minus
1199062
2
= Argmax119906isinA
minus 1205751205951(119909) + [(1 minus 119906) 119903119909 + 119906119887119909] 120595
1015840
1(119909)
+1
211990921199062120590212059510158401015840
1(119909) minus
1199062
2
=(119887119909 minus 119903119909) 120595
1015840
1(119909)
1 minus 1199092120590212059510158401015840
1(119909)
(52)
= Argmax119906isinA
1198601199061205952(119909)
= Argmax119906isinA
minus 1205751205952(119909) + [(1 minus 119906) 119903119909 + 119906119887119909] 120595
1015840
2(119909)
+1
211990921199062120590212059510158401015840
2(119909)
=minus (119887119909 minus 119903119909) 120595
1015840
2(119909)
1199092120590212059510158401015840
2(119909)
(53)
Let120595119894(119909) =
119894(119909) on 0 lt 119909 lt 119909
0 119894 = 1 2 By condition (5)
in Theorem 4 we have 1198601(119909) minus
22 = 0 for 0 lt 119909 lt 119909
0
Substituting (52) into 1198601199061(119909) minus 119906
22 = 0 we obtain
minus 1205751(119909) + [119903119909 +
(119887119909 minus 119903119909)21015840
1(119909)
1 minus 1199092120590210158401015840
1(119909)
] 1015840
1(119909)
+
[(119887119909 minus 119903119909) 1015840
1(119909)]2
2(1 minus 1199092120590210158401015840
1(119909))2[1199092120590210158401015840
1(119909) minus 1] = 0
(54)
Similarly we obtain1198602(119909) = 0 for 0 lt 119909 lt 119909
0by condition
(6) in Theorem 4 And we substitute (53) in 1198601199062(119909) = 0 to
get
minus 1205752(119909) + 119903119909
1015840
2(119909) +
[(119887119909 minus 119903119909) 1015840
2(119909)]2
1 minus 1199092120590210158401015840
2(119909)
+
[119909120590 (119887119909 minus 119903119909) 1015840
2(119909)]2
10158401015840
2(119909)
2(1 minus 1199092120590210158401015840
2(119909))2
= 0
(55)
Therefore we conclude that
1205951(119909) =
1198801(119862 (119909)) 119909 ge 119909
0
1(119909) 0 lt 119909 lt 119909
0
(56)
1205952(119909) =
1198802(119909 minus 119862 (119909)) 119909 ge 119909
0
2(119909) 0 lt 119909 lt 119909
0
(57)
where 1(119909) and
2(119909) are the solutions of (56) and (57)
respectivelyAccording to Theorem 4 we use the continuity and
differentiability of 120595119894at 119909 = 119909
0to determine 119909
0 119894 = 1 2 that
is
1205951(1199090) = 1198801(119862 (1199090))
1205951015840
1(1199090) = 119880
1015840
1(119862 (1199090)) 1198621015840(1199090)
1205952(1199090) = 1198802(1199090minus 119862 (119909
0))
1205951015840
2(1199090) = 119880
1015840
2(1199090minus 119862 (119909
0)) (1 minus 119862
1015840(1199090))
(58)
At the end of this section we summarize the above resultsin the following theorem
Theorem 5 Let 120595119894 119894 = 1 2 and let 119909
0be the solutions of
equations (56)ndash(58) Then the pair ( 120591) given by
=(119887119909 minus 119903119909) 120595
1015840
1(119909)
1 minus 1199092120590212059510158401015840
1(119909)
=minus (119887119909 minus 119903119909) 120595
1015840
2(119909)
1199092120590212059510158401015840
2(119909)
120591 = inf 119905 gt 0 119909 ge 1199090
(59)
is a Nash equilibrium of game (4) and (8) The correspondingequilibrium performances are
120601119894(119904 119909) = 119890
minus120575119904120595119894(119909) 119894 = 1 2 (60)
Abstract and Applied Analysis 7
4 Conclusion
A verification theorem is obtained for the general stochasticdifferential game between a controller and a stopper Inthe special case of quadratic cost we use this theorem tocharacterize the Nash equilibrium However the question ofthe existence and uniqueness of Nash equilibrium for thegame remains open It will be considered in our subsequentwork
Acknowledgment
This work was supported by the National Natural ScienceFoundation of China under Grant 61273022 and 11171050
References
[1] J Bertoin Levy Processes Cambridge University Press Cam-bridge UK 1996
[2] D Applebaum Levy Processes and Stochastic Calculus Cam-bridge University Press Cambridge UK 2003
[3] S Mataramvura and B Oslashksendal ldquoRisk minimizing portfoliosand HJBI equations for stochastic differential gamesrdquo Stochas-tics vol 80 no 4 pp 317ndash337 2008
[4] R J Elliott and T K Siu ldquoOn risk minimizing portfolios underaMarkovian regime-switching Black-Scholes economyrdquoAnnalsof Operations Research vol 176 pp 271ndash291 2010
[5] G Wang and Z Wu ldquoMean-variance hedging and forward-backward stochastic differential filtering equationsrdquo Abstractand Applied Analysis vol 2011 Article ID 310910 20 pages 2011
[6] I Karatzas and W Sudderth ldquoStochastic games of control andstopping for a linear diffusionrdquo in Random Walk SequentialAnalysis and Related Topics A Festschrift in Honor of YS ChowA Hsiung Z L Ying and C H Zhang Eds pp 100ndash117WorldScientific Publishers 2006
[7] S Hamadene ldquoMixed zero-sum stochastic differential gameand American game optionsrdquo SIAM Journal on Control andOptimization vol 45 no 2 pp 496ndash518 2006
[8] M K Ghosh M K S Rao and D Sheetal ldquoDifferential gamesof mixed type with control and stopping timesrdquo NonlinearDifferential Equations and Applications vol 16 no 2 pp 143ndash158 2009
[9] M K Ghosh and K S Mallikarjuna Rao ldquoExistence of value instochastic differential games of mixed typerdquo Stochastic Analysisand Applications vol 30 no 5 pp 895ndash905 2012
[10] I Karatzas and Q Li ldquoBSDE approach to non-zero-sumstochastic differential games of control and stoppingrdquo inStochastic Processes Finance and Control pp 105ndash153 WorldScientific Publishers 2012
[11] J-P Lepeltier ldquoOn a general zero-sum stochastic game withstopping strategy for one player and continuous strategy for theotherrdquo Probability and Mathematical Statistics vol 6 no 1 pp43ndash50 1985
[12] I Karatzas and W D Sudderth ldquoThe controller-and-stoppergame for a linear diffusionrdquo The Annals of Probability vol 29no 3 pp 1111ndash1127 2001
[13] AWeerasinghe ldquoA controller and a stopper gamewith degener-ate variance controlrdquo Electronic Communications in Probabilityvol 11 pp 89ndash99 2006
[14] I Karatzas and I-M Zamfirescu ldquoMartingale approach tostochastic differential games of control and stoppingrdquo TheAnnals of Probability vol 36 no 4 pp 1495ndash1527 2008
[15] E Bayraktar I Karatzas and S Yao ldquoOptimal stopping fordynamic convex risk measuresrdquo Illinois Journal of Mathematicsvol 54 no 3 pp 1025ndash1067 2010
[16] E Bayraktar and V R Young ldquoProving regularity of theminimal probability of ruin via a game of stopping and controlrdquoFinance and Stochastics vol 15 no 4 pp 785ndash818 2011
[17] F Bagherya S Haademb B Oslashksendal and I Turpina ldquoOptimalstopping and stochastic control differential games for jumpdiffusionsrdquo Stochastics vol 85 no 1 pp 85ndash97 2013
[18] B Oslashksendal and A Sulem Applied Stochastic Control of JumpDiffusions Springer Berlin Germany 2nd edition 2007
[19] Q Lin R Loxton K L Teo and Y H Wu ldquoA new computa-tional method for a class of free terminal time optimal controlproblemsrdquo Pacific Journal of Optimization vol 7 no 1 pp 63ndash81 2011
[20] Q Lin R Loxton K L Teo and Y H Wu ldquoOptimal controlcomputation for nonlinear systems with state-dependent stop-ping criteriardquo Automatica vol 48 no 9 pp 2116ndash2129 2012
[21] C Jiang Q Lin C Yu K L Teo and G-R Duan ldquoAn exactpenalty method for free terminal time optimal control problemwith continuous inequality constraintsrdquo Journal of OptimizationTheory and Applications vol 154 no 1 pp 30ndash53 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
![Page 6: Research Article Nonzero-Sum Stochastic Differential Game ...downloads.hindawi.com/journals/aaa/2013/761306.pdf · Controller and Stopper for Jump Diffusions YanWang, 1,2 AiminSong,](https://reader033.vdocuments.mx/reader033/viewer/2022042410/5f26fe8f7005a2296769470b/html5/thumbnails/6.jpg)
6 Abstract and Applied Analysis
Moreover the corresponding equilibrium performances are
1206011(119904 119909) = J
(119904119909)
1( 120591)
1206012(119904 119909) = J
(119904119909)
2( 120591)
(45)
In this paper we will not discuss general solutions of thisfamily of nonlinear variational-integro inequalities Insteadwe discuss a solution in special case when
120574 (119905 119911) = 0 ℎ (119904 119909 119906) =1199062
2 (46)
Let us try the functions 120601119894 119894 = 1 2 of the form
120601119894(119904 119909) = 119890
minus120575119904120595119894(119909) (47)
and a continuation region119863 of the form
119863 = (119904 119909) 119909 lt 1199090 for some 119909
0gt 0 (48)
Then we have
119860119906120601119894(119904 119909) = 119890
minus120575119904119860119906120595119894(119909) (49)
where119860119906120595119894(119909) = minus 120575120595
119894(119909) + [(1 minus 119906) 119903119909 + 119906119887119909] 120595
1015840
119894(119909)
+1
211990921199062120590212059510158401015840
119894(119909)
(50)
By conditions (1) and (3) in Theorem 4 we get
1205951(119909) = 119880
1(119862 (119909)) 119909 ge 119909
0
1205951(119909) gt 119880
1(119862 (119909)) 0 lt 119909 lt 119909
0
1205952(119909) = 119880
2(119909 minus 119862 (119909)) 119909 ge 119909
0
1205952(119909) gt 119880
2(119909 minus 119862 (119909)) 0 lt 119909 lt 119909
0
(51)
From conditions (4) (5) and (6) ofTheorem 4 we get thecandidate for the optimal control as follows
= Argmax119906isinA
1198601199061205951(119909) minus
1199062
2
= Argmax119906isinA
minus 1205751205951(119909) + [(1 minus 119906) 119903119909 + 119906119887119909] 120595
1015840
1(119909)
+1
211990921199062120590212059510158401015840
1(119909) minus
1199062
2
=(119887119909 minus 119903119909) 120595
1015840
1(119909)
1 minus 1199092120590212059510158401015840
1(119909)
(52)
= Argmax119906isinA
1198601199061205952(119909)
= Argmax119906isinA
minus 1205751205952(119909) + [(1 minus 119906) 119903119909 + 119906119887119909] 120595
1015840
2(119909)
+1
211990921199062120590212059510158401015840
2(119909)
=minus (119887119909 minus 119903119909) 120595
1015840
2(119909)
1199092120590212059510158401015840
2(119909)
(53)
Let120595119894(119909) =
119894(119909) on 0 lt 119909 lt 119909
0 119894 = 1 2 By condition (5)
in Theorem 4 we have 1198601(119909) minus
22 = 0 for 0 lt 119909 lt 119909
0
Substituting (52) into 1198601199061(119909) minus 119906
22 = 0 we obtain
minus 1205751(119909) + [119903119909 +
(119887119909 minus 119903119909)21015840
1(119909)
1 minus 1199092120590210158401015840
1(119909)
] 1015840
1(119909)
+
[(119887119909 minus 119903119909) 1015840
1(119909)]2
2(1 minus 1199092120590210158401015840
1(119909))2[1199092120590210158401015840
1(119909) minus 1] = 0
(54)
Similarly we obtain1198602(119909) = 0 for 0 lt 119909 lt 119909
0by condition
(6) in Theorem 4 And we substitute (53) in 1198601199062(119909) = 0 to
get
minus 1205752(119909) + 119903119909
1015840
2(119909) +
[(119887119909 minus 119903119909) 1015840
2(119909)]2
1 minus 1199092120590210158401015840
2(119909)
+
[119909120590 (119887119909 minus 119903119909) 1015840
2(119909)]2
10158401015840
2(119909)
2(1 minus 1199092120590210158401015840
2(119909))2
= 0
(55)
Therefore we conclude that
1205951(119909) =
1198801(119862 (119909)) 119909 ge 119909
0
1(119909) 0 lt 119909 lt 119909
0
(56)
1205952(119909) =
1198802(119909 minus 119862 (119909)) 119909 ge 119909
0
2(119909) 0 lt 119909 lt 119909
0
(57)
where 1(119909) and
2(119909) are the solutions of (56) and (57)
respectivelyAccording to Theorem 4 we use the continuity and
differentiability of 120595119894at 119909 = 119909
0to determine 119909
0 119894 = 1 2 that
is
1205951(1199090) = 1198801(119862 (1199090))
1205951015840
1(1199090) = 119880
1015840
1(119862 (1199090)) 1198621015840(1199090)
1205952(1199090) = 1198802(1199090minus 119862 (119909
0))
1205951015840
2(1199090) = 119880
1015840
2(1199090minus 119862 (119909
0)) (1 minus 119862
1015840(1199090))
(58)
At the end of this section we summarize the above resultsin the following theorem
Theorem 5 Let 120595119894 119894 = 1 2 and let 119909
0be the solutions of
equations (56)ndash(58) Then the pair ( 120591) given by
=(119887119909 minus 119903119909) 120595
1015840
1(119909)
1 minus 1199092120590212059510158401015840
1(119909)
=minus (119887119909 minus 119903119909) 120595
1015840
2(119909)
1199092120590212059510158401015840
2(119909)
120591 = inf 119905 gt 0 119909 ge 1199090
(59)
is a Nash equilibrium of game (4) and (8) The correspondingequilibrium performances are
120601119894(119904 119909) = 119890
minus120575119904120595119894(119909) 119894 = 1 2 (60)
Abstract and Applied Analysis 7
4 Conclusion
A verification theorem is obtained for the general stochasticdifferential game between a controller and a stopper Inthe special case of quadratic cost we use this theorem tocharacterize the Nash equilibrium However the question ofthe existence and uniqueness of Nash equilibrium for thegame remains open It will be considered in our subsequentwork
Acknowledgment
This work was supported by the National Natural ScienceFoundation of China under Grant 61273022 and 11171050
References
[1] J Bertoin Levy Processes Cambridge University Press Cam-bridge UK 1996
[2] D Applebaum Levy Processes and Stochastic Calculus Cam-bridge University Press Cambridge UK 2003
[3] S Mataramvura and B Oslashksendal ldquoRisk minimizing portfoliosand HJBI equations for stochastic differential gamesrdquo Stochas-tics vol 80 no 4 pp 317ndash337 2008
[4] R J Elliott and T K Siu ldquoOn risk minimizing portfolios underaMarkovian regime-switching Black-Scholes economyrdquoAnnalsof Operations Research vol 176 pp 271ndash291 2010
[5] G Wang and Z Wu ldquoMean-variance hedging and forward-backward stochastic differential filtering equationsrdquo Abstractand Applied Analysis vol 2011 Article ID 310910 20 pages 2011
[6] I Karatzas and W Sudderth ldquoStochastic games of control andstopping for a linear diffusionrdquo in Random Walk SequentialAnalysis and Related Topics A Festschrift in Honor of YS ChowA Hsiung Z L Ying and C H Zhang Eds pp 100ndash117WorldScientific Publishers 2006
[7] S Hamadene ldquoMixed zero-sum stochastic differential gameand American game optionsrdquo SIAM Journal on Control andOptimization vol 45 no 2 pp 496ndash518 2006
[8] M K Ghosh M K S Rao and D Sheetal ldquoDifferential gamesof mixed type with control and stopping timesrdquo NonlinearDifferential Equations and Applications vol 16 no 2 pp 143ndash158 2009
[9] M K Ghosh and K S Mallikarjuna Rao ldquoExistence of value instochastic differential games of mixed typerdquo Stochastic Analysisand Applications vol 30 no 5 pp 895ndash905 2012
[10] I Karatzas and Q Li ldquoBSDE approach to non-zero-sumstochastic differential games of control and stoppingrdquo inStochastic Processes Finance and Control pp 105ndash153 WorldScientific Publishers 2012
[11] J-P Lepeltier ldquoOn a general zero-sum stochastic game withstopping strategy for one player and continuous strategy for theotherrdquo Probability and Mathematical Statistics vol 6 no 1 pp43ndash50 1985
[12] I Karatzas and W D Sudderth ldquoThe controller-and-stoppergame for a linear diffusionrdquo The Annals of Probability vol 29no 3 pp 1111ndash1127 2001
[13] AWeerasinghe ldquoA controller and a stopper gamewith degener-ate variance controlrdquo Electronic Communications in Probabilityvol 11 pp 89ndash99 2006
[14] I Karatzas and I-M Zamfirescu ldquoMartingale approach tostochastic differential games of control and stoppingrdquo TheAnnals of Probability vol 36 no 4 pp 1495ndash1527 2008
[15] E Bayraktar I Karatzas and S Yao ldquoOptimal stopping fordynamic convex risk measuresrdquo Illinois Journal of Mathematicsvol 54 no 3 pp 1025ndash1067 2010
[16] E Bayraktar and V R Young ldquoProving regularity of theminimal probability of ruin via a game of stopping and controlrdquoFinance and Stochastics vol 15 no 4 pp 785ndash818 2011
[17] F Bagherya S Haademb B Oslashksendal and I Turpina ldquoOptimalstopping and stochastic control differential games for jumpdiffusionsrdquo Stochastics vol 85 no 1 pp 85ndash97 2013
[18] B Oslashksendal and A Sulem Applied Stochastic Control of JumpDiffusions Springer Berlin Germany 2nd edition 2007
[19] Q Lin R Loxton K L Teo and Y H Wu ldquoA new computa-tional method for a class of free terminal time optimal controlproblemsrdquo Pacific Journal of Optimization vol 7 no 1 pp 63ndash81 2011
[20] Q Lin R Loxton K L Teo and Y H Wu ldquoOptimal controlcomputation for nonlinear systems with state-dependent stop-ping criteriardquo Automatica vol 48 no 9 pp 2116ndash2129 2012
[21] C Jiang Q Lin C Yu K L Teo and G-R Duan ldquoAn exactpenalty method for free terminal time optimal control problemwith continuous inequality constraintsrdquo Journal of OptimizationTheory and Applications vol 154 no 1 pp 30ndash53 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
![Page 7: Research Article Nonzero-Sum Stochastic Differential Game ...downloads.hindawi.com/journals/aaa/2013/761306.pdf · Controller and Stopper for Jump Diffusions YanWang, 1,2 AiminSong,](https://reader033.vdocuments.mx/reader033/viewer/2022042410/5f26fe8f7005a2296769470b/html5/thumbnails/7.jpg)
Abstract and Applied Analysis 7
4 Conclusion
A verification theorem is obtained for the general stochasticdifferential game between a controller and a stopper Inthe special case of quadratic cost we use this theorem tocharacterize the Nash equilibrium However the question ofthe existence and uniqueness of Nash equilibrium for thegame remains open It will be considered in our subsequentwork
Acknowledgment
This work was supported by the National Natural ScienceFoundation of China under Grant 61273022 and 11171050
References
[1] J Bertoin Levy Processes Cambridge University Press Cam-bridge UK 1996
[2] D Applebaum Levy Processes and Stochastic Calculus Cam-bridge University Press Cambridge UK 2003
[3] S Mataramvura and B Oslashksendal ldquoRisk minimizing portfoliosand HJBI equations for stochastic differential gamesrdquo Stochas-tics vol 80 no 4 pp 317ndash337 2008
[4] R J Elliott and T K Siu ldquoOn risk minimizing portfolios underaMarkovian regime-switching Black-Scholes economyrdquoAnnalsof Operations Research vol 176 pp 271ndash291 2010
[5] G Wang and Z Wu ldquoMean-variance hedging and forward-backward stochastic differential filtering equationsrdquo Abstractand Applied Analysis vol 2011 Article ID 310910 20 pages 2011
[6] I Karatzas and W Sudderth ldquoStochastic games of control andstopping for a linear diffusionrdquo in Random Walk SequentialAnalysis and Related Topics A Festschrift in Honor of YS ChowA Hsiung Z L Ying and C H Zhang Eds pp 100ndash117WorldScientific Publishers 2006
[7] S Hamadene ldquoMixed zero-sum stochastic differential gameand American game optionsrdquo SIAM Journal on Control andOptimization vol 45 no 2 pp 496ndash518 2006
[8] M K Ghosh M K S Rao and D Sheetal ldquoDifferential gamesof mixed type with control and stopping timesrdquo NonlinearDifferential Equations and Applications vol 16 no 2 pp 143ndash158 2009
[9] M K Ghosh and K S Mallikarjuna Rao ldquoExistence of value instochastic differential games of mixed typerdquo Stochastic Analysisand Applications vol 30 no 5 pp 895ndash905 2012
[10] I Karatzas and Q Li ldquoBSDE approach to non-zero-sumstochastic differential games of control and stoppingrdquo inStochastic Processes Finance and Control pp 105ndash153 WorldScientific Publishers 2012
[11] J-P Lepeltier ldquoOn a general zero-sum stochastic game withstopping strategy for one player and continuous strategy for theotherrdquo Probability and Mathematical Statistics vol 6 no 1 pp43ndash50 1985
[12] I Karatzas and W D Sudderth ldquoThe controller-and-stoppergame for a linear diffusionrdquo The Annals of Probability vol 29no 3 pp 1111ndash1127 2001
[13] AWeerasinghe ldquoA controller and a stopper gamewith degener-ate variance controlrdquo Electronic Communications in Probabilityvol 11 pp 89ndash99 2006
[14] I Karatzas and I-M Zamfirescu ldquoMartingale approach tostochastic differential games of control and stoppingrdquo TheAnnals of Probability vol 36 no 4 pp 1495ndash1527 2008
[15] E Bayraktar I Karatzas and S Yao ldquoOptimal stopping fordynamic convex risk measuresrdquo Illinois Journal of Mathematicsvol 54 no 3 pp 1025ndash1067 2010
[16] E Bayraktar and V R Young ldquoProving regularity of theminimal probability of ruin via a game of stopping and controlrdquoFinance and Stochastics vol 15 no 4 pp 785ndash818 2011
[17] F Bagherya S Haademb B Oslashksendal and I Turpina ldquoOptimalstopping and stochastic control differential games for jumpdiffusionsrdquo Stochastics vol 85 no 1 pp 85ndash97 2013
[18] B Oslashksendal and A Sulem Applied Stochastic Control of JumpDiffusions Springer Berlin Germany 2nd edition 2007
[19] Q Lin R Loxton K L Teo and Y H Wu ldquoA new computa-tional method for a class of free terminal time optimal controlproblemsrdquo Pacific Journal of Optimization vol 7 no 1 pp 63ndash81 2011
[20] Q Lin R Loxton K L Teo and Y H Wu ldquoOptimal controlcomputation for nonlinear systems with state-dependent stop-ping criteriardquo Automatica vol 48 no 9 pp 2116ndash2129 2012
[21] C Jiang Q Lin C Yu K L Teo and G-R Duan ldquoAn exactpenalty method for free terminal time optimal control problemwith continuous inequality constraintsrdquo Journal of OptimizationTheory and Applications vol 154 no 1 pp 30ndash53 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
![Page 8: Research Article Nonzero-Sum Stochastic Differential Game ...downloads.hindawi.com/journals/aaa/2013/761306.pdf · Controller and Stopper for Jump Diffusions YanWang, 1,2 AiminSong,](https://reader033.vdocuments.mx/reader033/viewer/2022042410/5f26fe8f7005a2296769470b/html5/thumbnails/8.jpg)
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of