lecture lecture – ––– 25 225525 model predictive spread ... · prof. radhakant padhi, ae...

29
Lecture Lecture Lecture Lecture – 25 25 25 25 Model Predictive Spread Control (MPSC) and Model Predictive Spread Control (MPSC) and Model Predictive Spread Control (MPSC) and Model Predictive Spread Control (MPSC) and Generalized MPSP (G Generalized MPSP (G Generalized MPSP (G Generalized MPSP (G-MPSP) Designs MPSP) Designs MPSP) Designs MPSP) Designs Prof. Radhakant Padhi Prof. Radhakant Padhi Prof. Radhakant Padhi Prof. Radhakant Padhi Dept. of Aerospace Engineering Indian Institute of Science - Bangalore Optimal Control, Guidance and Estimation OPTIMAL CONTROL, GUIDANCE AND ESTIMATION Prof. Radhakant Padhi, AE Dept., IISc-Bangalore 2 Outline Motivation MPSC Design Mathematical Development Alignment Angle Constrained Midcourse Guidance of a Tactical Missile G-MPSP Design Mathematical Development Tactical Missile Guidance with 3-D Impact Angle Constraint Concluding Remarks

Upload: others

Post on 10-Mar-2020

2 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Lecture Lecture – ––– 25 225525 Model Predictive Spread ... · Prof. Radhakant Padhi, AE Dept., IISc -Bangalore 11 11 one gets MPSC with Quadratic Parameterization of Control

Lecture Lecture Lecture Lecture –––– 25252525

Model Predictive Spread Control (MPSC) and Model Predictive Spread Control (MPSC) and Model Predictive Spread Control (MPSC) and Model Predictive Spread Control (MPSC) and

Generalized MPSP (GGeneralized MPSP (GGeneralized MPSP (GGeneralized MPSP (G----MPSP) DesignsMPSP) DesignsMPSP) DesignsMPSP) Designs

Prof. Radhakant PadhiProf. Radhakant PadhiProf. Radhakant PadhiProf. Radhakant Padhi

Dept. of Aerospace Engineering

Indian Institute of Science - Bangalore

Optimal Control, Guidance and Estimation

OPTIMAL CONTROL, GUIDANCE AND ESTIMATION

Prof. Radhakant Padhi, AE Dept., IISc-Bangalore2

Outline

� Motivation

� MPSC Design

• Mathematical Development

• Alignment Angle Constrained Midcourse Guidance of a Tactical Missile

� G-MPSP Design

• Mathematical Development

• Tactical Missile Guidance with 3-D Impact Angle Constraint

� Concluding Remarks

Page 2: Lecture Lecture – ––– 25 225525 Model Predictive Spread ... · Prof. Radhakant Padhi, AE Dept., IISc -Bangalore 11 11 one gets MPSC with Quadratic Parameterization of Control

Model Predictive Spread Control (MPSC)Model Predictive Spread Control (MPSC)Model Predictive Spread Control (MPSC)Model Predictive Spread Control (MPSC)

Prof. Radhakant PadhiProf. Radhakant PadhiProf. Radhakant PadhiProf. Radhakant Padhi

Dept. of Aerospace Engineering

Indian Institute of Science - Bangalore

OPTIMAL CONTROL, GUIDANCE AND ESTIMATION

Prof. Radhakant Padhi, AE Dept., IISc-Bangalore4

Motivations

� High computational efficiency: Real-time online solution (better than MPSP??)

� Terminal conditions should be met as “hard constraints” (in missile guidance problems, this leads to high accuracy)

� No approximation of system dynamics

� Minimum control usage (without compromising on output accuracy)

� Control Smoothness (by enforcement)

Page 3: Lecture Lecture – ––– 25 225525 Model Predictive Spread ... · Prof. Radhakant Padhi, AE Dept., IISc -Bangalore 11 11 one gets MPSC with Quadratic Parameterization of Control

OPTIMAL CONTROL, GUIDANCE AND ESTIMATION

Prof. Radhakant Padhi, AE Dept., IISc-Bangalore5

System dynamics:

MPSP Design: An Overview

Discretized

Goal: with additional (optimal) objective(s)*

N NY Y→

( )

( )

,X f X U

Y h X

=

=

ɺ ( )

( )1

,k k k k

k k

X F X U

Y h X

+ =

=

OPTIMAL CONTROL, GUIDANCE AND ESTIMATION

Prof. Radhakant Padhi, AE Dept., IISc-Bangalore6

MPSP Design: An Overview

Philosophy:

• Guess a control history

• Simulate the system dynamics

• Compute the “error in the output” at k = N

• Update the control history optimally utilizing this error information

• Iterate the control history until convergence

( )* 0N N NY Y Y∆ − →Objective : ≜

Page 4: Lecture Lecture – ––– 25 225525 Model Predictive Spread ... · Prof. Radhakant Padhi, AE Dept., IISc -Bangalore 11 11 one gets MPSC with Quadratic Parameterization of Control

OPTIMAL CONTROL, GUIDANCE AND ESTIMATION

Prof. Radhakant Padhi, AE Dept., IISc-Bangalore7

MPSP Design: An Overview

1 1k k N N NB dU B dU dY

− −+ + =⋯

1 1

1 1

1 1

1 2 2 1

2 2

1 2 2 1

N

N N N

N

N N N

N N

N N N

N N N N N N

N N

N N N N N N

YY dY dX

X

Y F FdX dU

X X U

Y F F F Y FdX dU

X X X U X U

− −

− −

− −

− − − −− −

− − − −

∂∆ ≈ =

∂ ∂ ∂ = +

∂ ∂ ∂

∂ ∂ ∂ ∂ ∂ ∂ = + +

∂ ∂ ∂ ∂ ∂ ∂ 1

1 1 1

1

1 1 1

N

N N k N N k N N

k k N

N N k N N k N N

dU

Y F F Y F F Y FdX dU dU

X X X X X U X U

− − −−

− − −

∂ ∂ ∂ ∂ ∂ ∂ ∂ ∂= + + +

∂ ∂ ∂ ∂ ∂ ∂ ∂ ∂

⋯ ⋯ ⋯⋯

0

kB

1NB

(small error approximation)

The sensitivity matrices can be

computed “recursively”.

OPTIMAL CONTROL, GUIDANCE AND ESTIMATION

Prof. Radhakant Padhi, AE Dept., IISc-Bangalore8

� Parameterize control as a linear function of

� Carryout a sensitivity analysis of the output error with respect to the error in the control history

got

MPSC with Linear

Parameterization of Control

( ) ( )� ( )�0 0

0

0 0

0

,k k

k

k go k go

k k k go

a a b b

U a t b U at b

dU U U a t b

− −

= + = +

= − = ∆ + ∆

( ) ( )1 1

1 1 1 1

1 1 1 1

1 1Note: can be computed recursively !

N

yy

N N N

go N go N

DC

y y N

dY B dU B dU

B t B t a B B b

C a D b B B

− −

− −

= + +

= + + ∆ + + + + ∆

= ∆ + ∆

⋯⋯

⋯ ⋯ ⋯����������������

⋯⋯

Page 5: Lecture Lecture – ––– 25 225525 Model Predictive Spread ... · Prof. Radhakant Padhi, AE Dept., IISc -Bangalore 11 11 one gets MPSC with Quadratic Parameterization of Control

OPTIMAL CONTROL, GUIDANCE AND ESTIMATION

Prof. Radhakant Padhi, AE Dept., IISc-Bangalore9

MPSC with Linear

Parameterization of Control

( )

( )

1 2

0 0

1Optimize subject to

2

T T

y y N y y y

J a R a b R b

C a D b dY C a D b K

= +

+ = − + + ≜

� Formulate an optimization problem

� Solve this optimization problem in closed form

( )

1

1

1

2

11 1

1 2where

T

y

T

y

T T

y y y y y

a R C

b R D

C R C D R D K

λ

λ

λ

−− −

= −

= −

= − +

OPTIMAL CONTROL, GUIDANCE AND ESTIMATION

Prof. Radhakant Padhi, AE Dept., IISc-Bangalore1010

Control Parameterization

Error in control

Substituting for dUk for k = 1,.....,N-1 in

MPSC with Quadratic

Parameterization of Control

Page 6: Lecture Lecture – ––– 25 225525 Model Predictive Spread ... · Prof. Radhakant Padhi, AE Dept., IISc -Bangalore 11 11 one gets MPSC with Quadratic Parameterization of Control

OPTIMAL CONTROL, GUIDANCE AND ESTIMATION

Prof. Radhakant Padhi, AE Dept., IISc-Bangalore1111

one gets

MPSC with Quadratic

Parameterization of Control

OPTIMAL CONTROL, GUIDANCE AND ESTIMATION

Prof. Radhakant Padhi, AE Dept., IISc-Bangalore12

� If number of equations is same as number of

unknowns, then

12

� if number of unknowns is greater than the number of

equations, then optimal solution can be obtained by minimizing the following objective (cost) function

MPSC with Quadratic

Parameterization of Control

( )1 2 3

1

2

T T TJ a R a b R b c R c= + +

Page 7: Lecture Lecture – ––– 25 225525 Model Predictive Spread ... · Prof. Radhakant Padhi, AE Dept., IISc -Bangalore 11 11 one gets MPSC with Quadratic Parameterization of Control

OPTIMAL CONTROL, GUIDANCE AND ESTIMATION

Prof. Radhakant Padhi, AE Dept., IISc-Bangalore13

Start

Guess a control history

(in parameter form)

Propagate system dynamics

Compute output

Converged control solution

Update the control history(parameters)

Compute the sensitivitymatrices recursively

Stop

Checkconvergence

Yes

No

MPSC

ALGORITHM

OPTIMAL CONTROL, GUIDANCE AND ESTIMATION

Prof. Radhakant Padhi, AE Dept., IISc-Bangalore14

MPSC Design: Reasons for

Computational Efficiency

� Costate variable becomes “static”; i.e. only one time-independent (constant) costate vector is needed for the entire control history update!

� Dimension of costate vector is same as the dimension of the output vector (which is much lesser than the number of states)

� The costate vector is computed symbolically.

� Leads to closed form control history update.

� The computations needed include sensitivity matrices, which are computed “recursively”.

� If necessary, concepts like “iteration unfolding” can be incorporated to save computational time further.

Page 8: Lecture Lecture – ––– 25 225525 Model Predictive Spread ... · Prof. Radhakant Padhi, AE Dept., IISc -Bangalore 11 11 one gets MPSC with Quadratic Parameterization of Control

Alignment Angle Constrained Alignment Angle Constrained Alignment Angle Constrained Alignment Angle Constrained

Midcourse Guidance using MPSCMidcourse Guidance using MPSCMidcourse Guidance using MPSCMidcourse Guidance using MPSC

Prof. Radhakant PadhiProf. Radhakant PadhiProf. Radhakant PadhiProf. Radhakant Padhi

Dept. of Aerospace Engineering

Indian Institute of Science - Bangalore

OPTIMAL CONTROL, GUIDANCE AND ESTIMATION

Prof. Radhakant Padhi, AE Dept., IISc-Bangalore16

OBJECTIVES

� Interceptor must have sufficient capability and proper

initial condition for terminal guidance phase .

� Mid course guidance to provide proper initial

condition to terminal guidance phase.

� Interceptor spends most of its time during mid

course phase and hence should be energy efficient

� Objective: Interceptor has to reach desired point(xd,

yd,zd) with desired heading angle (Φd) and flight path

angle (γd) using minimum acceleration ηΦ and ηγ.

Page 9: Lecture Lecture – ––– 25 225525 Model Predictive Spread ... · Prof. Radhakant Padhi, AE Dept., IISc -Bangalore 11 11 one gets MPSC with Quadratic Parameterization of Control

OPTIMAL CONTROL, GUIDANCE AND ESTIMATION

Prof. Radhakant Padhi, AE Dept., IISc-Bangalore1717

MID COURSE GUIDANCE WITH MPSC (Mathematical model)

OPTIMAL CONTROL, GUIDANCE AND ESTIMATION

Prof. Radhakant Padhi, AE Dept., IISc-Bangalore1818

System Dynamics with Downrange as Independent Parameter

System Dynamics:

Control Parameterization:

Output Error at Final Time:

Page 10: Lecture Lecture – ––– 25 225525 Model Predictive Spread ... · Prof. Radhakant Padhi, AE Dept., IISc -Bangalore 11 11 one gets MPSC with Quadratic Parameterization of Control

OPTIMAL CONTROL, GUIDANCE AND ESTIMATION

Prof. Radhakant Padhi, AE Dept., IISc-Bangalore1919

RESULTS

OPTIMAL CONTROL, GUIDANCE AND ESTIMATION

Prof. Radhakant Padhi, AE Dept., IISc-Bangalore20

20

Page 11: Lecture Lecture – ––– 25 225525 Model Predictive Spread ... · Prof. Radhakant Padhi, AE Dept., IISc -Bangalore 11 11 one gets MPSC with Quadratic Parameterization of Control

OPTIMAL CONTROL, GUIDANCE AND ESTIMATION

Prof. Radhakant Padhi, AE Dept., IISc-Bangalore21

Improvement with Iterations

OPTIMAL CONTROL, GUIDANCE AND ESTIMATION

Prof. Radhakant Padhi, AE Dept., IISc-Bangalore22

Page 12: Lecture Lecture – ––– 25 225525 Model Predictive Spread ... · Prof. Radhakant Padhi, AE Dept., IISc -Bangalore 11 11 one gets MPSC with Quadratic Parameterization of Control

OPTIMAL CONTROL, GUIDANCE AND ESTIMATION

Prof. Radhakant Padhi, AE Dept., IISc-Bangalore23

OPTIMAL CONTROL, GUIDANCE AND ESTIMATION

Prof. Radhakant Padhi, AE Dept., IISc-Bangalore24

Page 13: Lecture Lecture – ––– 25 225525 Model Predictive Spread ... · Prof. Radhakant Padhi, AE Dept., IISc -Bangalore 11 11 one gets MPSC with Quadratic Parameterization of Control

OPTIMAL CONTROL, GUIDANCE AND ESTIMATION

Prof. Radhakant Padhi, AE Dept., IISc-Bangalore25

Further Results

P. N. Dwivedi, A. Bhattacharya and Radhakant Padhi

Suboptimal Mid-course Guidance of Interceptors for High Speed Targets with Alignment Angle Constraint

AIAA Journal of Guidance, Control and Dynamics, Vol. 34, No. 3, 2011, pp. 860-877.

Reference:

Generalized Model Predictive Static Generalized Model Predictive Static Generalized Model Predictive Static Generalized Model Predictive Static

Programming (GProgramming (GProgramming (GProgramming (G----MPSP)MPSP)MPSP)MPSP)

Prof. Radhakant PadhiProf. Radhakant PadhiProf. Radhakant PadhiProf. Radhakant Padhi

Dept. of Aerospace Engineering

Indian Institute of Science - Bangalore

Page 14: Lecture Lecture – ––– 25 225525 Model Predictive Spread ... · Prof. Radhakant Padhi, AE Dept., IISc -Bangalore 11 11 one gets MPSC with Quadratic Parameterization of Control

OPTIMAL CONTROL, GUIDANCE AND ESTIMATION

Prof. Radhakant Padhi, AE Dept., IISc-Bangalore27

Motivations

� High computational efficiency: Real-time online solution

� Terminal conditions should be met as “hard constraints” (in missile guidance problems, this leads to high accuracy)

� No approximation of system dynamics

� Minimum control usage (without compromising on output accuracy)

� Question: Can the discretized problem formulation be avoided?

OPTIMAL CONTROL, GUIDANCE AND ESTIMATION

Prof. Radhakant Padhi, AE Dept., IISc-Bangalore28

System dynamics:

GMPSP Design: An Overview

Goal: with additional (optimal)

objective(s)( ) ( )*

f fY t Y t→

( )

( )

,X f X U

Y h X

=

=

ɺ

where, , ,n m pX U Y∈ℜ ∈ℜ ∈ℜ

Page 15: Lecture Lecture – ––– 25 225525 Model Predictive Spread ... · Prof. Radhakant Padhi, AE Dept., IISc -Bangalore 11 11 one gets MPSC with Quadratic Parameterization of Control

OPTIMAL CONTROL, GUIDANCE AND ESTIMATION

Prof. Radhakant Padhi, AE Dept., IISc-Bangalore29

GMPSP Design : An Overview

Philosophy:

• Guess a control history

• Simulate the system dynamics

• Compute the “error in the output” at t = tf

• Update the control history optimally utilizing this error information

• Iterate the control history until convergence

( ) ( ) ( )( )* 0f f fY t Y t Y tδ − →Objective : ≜

OPTIMAL CONTROL, GUIDANCE AND ESTIMATION

Prof. Radhakant Padhi, AE Dept., IISc-Bangalore30

GMPSP Design: An Overview

( ) ( ) ( ) ( ) ( )( ),W t X t W t f X t U t=ɺ ( )where, p nW t ×∈ℜ

( ) ( ) ( ) ( ) ( )( )0 0

,f ft t

t tW t X t dt W t f X t U t dt = ∫ ∫ɺ

( )( ) ( )( ) ( ) ( ) ( )( ) ( ) ( )0 0

,f ft t

f ft t

Y X t Y X t W t f X t U t dt W t X t dt = + − ∫ ∫ ɺ

� Multiplying both sides of the system dynamics by the matrix : ( )W t

� Integrating both sides from to : 0

tf

t

� Adding to both sides: ( )( )fY X t

Page 16: Lecture Lecture – ––– 25 225525 Model Predictive Spread ... · Prof. Radhakant Padhi, AE Dept., IISc -Bangalore 11 11 one gets MPSC with Quadratic Parameterization of Control

OPTIMAL CONTROL, GUIDANCE AND ESTIMATION

Prof. Radhakant Padhi, AE Dept., IISc-Bangalore31

GMPSP Design: An Overview

( ) ( ) ( ) ( )( )

( )00 0

f fft tt

tt t

dW tW t X t dt W t X t X t dt

dt

= −

∫ ∫ɺ

( )( ) ( )( ) ( ) ( ) ( ) ( )

( ) ( ) ( )( ) ( ) ( )0

0 0

,f

f f f f

t

t

Y X t Y X t W t X t W t X t

W t f X t U t W t X t dt

= − +

+ + ∫ ɺ

� Integrating by parts of the last term of the right hand side of last equation:

� Substituting above relation in last equation of previous slide:

OPTIMAL CONTROL, GUIDANCE AND ESTIMATION

Prof. Radhakant Padhi, AE Dept., IISc-Bangalore32

GMPSP Design: An Overview

( )( )( )

( )( ) ( ) ( ) ( )

( )( ) ( )( )

( )( ) ( ) ( )

( ) ( )( )( )

( )0

0 0

( )

, ,

f

f

f

t t

t

t

B t

Y X tY t W t X t W t X t

X t

f X t U t f X t U tW t W t X t W t U t dt

X t U t

δ δ δ

δ δ

=

∂= − + ∂

∂ ∂ + + + ∂ ∂

∫ ɺ

����������

0

( ) ( ) ( )0

ft

ft

Y t B t U t dtδ δ= ∫

� Taking the variation of the both sides and re-arranging terms:

0

0

Page 17: Lecture Lecture – ––– 25 225525 Model Predictive Spread ... · Prof. Radhakant Padhi, AE Dept., IISc -Bangalore 11 11 one gets MPSC with Quadratic Parameterization of Control

OPTIMAL CONTROL, GUIDANCE AND ESTIMATION

Prof. Radhakant Padhi, AE Dept., IISc-Bangalore33

Recursive Relation for Computation

of Sensitivity Matrices

� General formula for Recursive Computation:

( ) ( )( ) ( )( )

( )

,f X t U tB t W t

U t

∂=

( )( )( )

( )f

f

f

Y X tW t

X t

∂=

( ) ( )( ) ( )( )

( )

,f X t U tW t W t

X t

∂= −

ɺ

OPTIMAL CONTROL, GUIDANCE AND ESTIMATION

Prof. Radhakant Padhi, AE Dept., IISc-Bangalore34

Augmented Cost Function:

GMPSP Design: Mathematical Formulation

Minimize:

Subject to: ( ) ( ) ( )0

ft

ft

Y t B t U t dtδ δ= ∫

( ) ( )( ) ( ) ( ) ( )( )0

0 01

2

ft T

ct

J U t U t R t U t U t dtδ δ = − − ∫

( ) ( )( ) ( ) ( ) ( )( )

( ) ( ) ( )

0

0

0 01

2

f

f

t T

ct

tT

ft

J U t U t R t U t U t dt

Y t B t U t dt

δ δ

λ δ δ

= − −

+ −

Page 18: Lecture Lecture – ––– 25 225525 Model Predictive Spread ... · Prof. Radhakant Padhi, AE Dept., IISc -Bangalore 11 11 one gets MPSC with Quadratic Parameterization of Control

OPTIMAL CONTROL, GUIDANCE AND ESTIMATION

Prof. Radhakant Padhi, AE Dept., IISc-Bangalore35

Necessary Conditions of Optimality:

GMPSP Design: Mathematical Formulation

( ) ( ) ( )0

0 ft

c

ft

JY t B t U t dtδ δ

λ

∂= ⇒ = ∂ ∫

( )( )( ) ( ) ( )( ) ( )( )0 0

TcJ

R t U t U t B tU t

δ λδ

∂= − − − =

OPTIMAL CONTROL, GUIDANCE AND ESTIMATION

Prof. Radhakant Padhi, AE Dept., IISc-Bangalore36

Control Solution:

GMPSP Design: Mathematical Formulation

( ) ( )( ) ( )( ) ( )1 0T

U t R t B t U tδ λ−

= +

( ) ( ) ( )0

ft

ft

Y t B t U t dtδ δ= ∫

( ) ( )1

fA Y t bλ λλ δ− = −

( ) ( )( ) ( )

( ) ( )

0

0

1

0

f

f

tT

t

t

t

A B t R t B t dt

b B t U t dt

λ

λ

where

Page 19: Lecture Lecture – ––– 25 225525 Model Predictive Spread ... · Prof. Radhakant Padhi, AE Dept., IISc -Bangalore 11 11 one gets MPSC with Quadratic Parameterization of Control

OPTIMAL CONTROL, GUIDANCE AND ESTIMATION

Prof. Radhakant Padhi, AE Dept., IISc-Bangalore37

Control Update:

GMPSP Design: Mathematical Formulation

( ) ( ) ( )

( ) ( )( ) ( )( ) ( ) ( ){ } ( )

( )( ) ( )( ) ( ) ( ){ }

0

1 10 0

1 1

=

=

T

f

T

f

U t U t U t

U t R t B t A Y t b U t

R t B t A Y t b

λ λ

λ λ

δ

δ

δ

− −

− −

= −

− − −

− −

where ( ) ( )1

fA Y t bλ λλ δ− = −

OPTIMAL CONTROL, GUIDANCE AND ESTIMATION

Prof. Radhakant Padhi, AE Dept., IISc-Bangalore38

Start

Guess a control history

Propagate system dynamics

Compute output

Converged control solution

Update the static costateand the control history

Compute the weighting matrixby Backward integration

Stop

Checkconvergence

Yes

No

G-MPSP

ALGORITHM

Page 20: Lecture Lecture – ––– 25 225525 Model Predictive Spread ... · Prof. Radhakant Padhi, AE Dept., IISc -Bangalore 11 11 one gets MPSC with Quadratic Parameterization of Control

Tactical Missile Guidance with 3Tactical Missile Guidance with 3Tactical Missile Guidance with 3Tactical Missile Guidance with 3----D D D D

Impact Angle ConstraintImpact Angle ConstraintImpact Angle ConstraintImpact Angle Constraint

Prof. Radhakant PadhiProf. Radhakant PadhiProf. Radhakant PadhiProf. Radhakant Padhi

Dept. of Aerospace Engineering

Indian Institute of Science - Bangalore

OPTIMAL CONTROL, GUIDANCE AND ESTIMATION

Prof. Radhakant Padhi, AE Dept., IISc-Bangalore40

Motivation

� Is it possible to achieve impact angles in 3D simultaneously in some optimal manner?

� Can terminal constraints in both the angles Azimuth-Angle (Direction of heading), Elevation-Angle (Pitch or Top) be dictated?

� Can the above objective be achieved for stationary, moving and maneuvering ground

targets?

� Can this be achieved with minimum latax demand?

Page 21: Lecture Lecture – ––– 25 225525 Model Predictive Spread ... · Prof. Radhakant Padhi, AE Dept., IISc -Bangalore 11 11 one gets MPSC with Quadratic Parameterization of Control

OPTIMAL CONTROL, GUIDANCE AND ESTIMATION

Prof. Radhakant Padhi, AE Dept., IISc-Bangalore41

Challenges

� Strong (nonlinear) coupling between elevation angle and azimuth angle dynamics should be accounted for

� Zero/Near-zero miss distance is desired

� Impact angle constraints in 3D are desired

� Latax demand has to be as minimum as possible.

OPTIMAL CONTROL, GUIDANCE AND ESTIMATION

Prof. Radhakant Padhi, AE Dept., IISc-Bangalore42

A Typical Engagement Scenario

Page 22: Lecture Lecture – ––– 25 225525 Model Predictive Spread ... · Prof. Radhakant Padhi, AE Dept., IISc -Bangalore 11 11 one gets MPSC with Quadratic Parameterization of Control

OPTIMAL CONTROL, GUIDANCE AND ESTIMATION

Prof. Radhakant Padhi, AE Dept., IISc-Bangalore43

[ ]sin( )

cos( )

cos( )

cos( )cos( )

cos( )sin( )

sin( )

m m

m m

m

z m

m

m

y

m

m m

m m m m

m m m m

m m m

T DV g

m

a g

V

a

V

x V

y V

z V

γ

γγ

ψγ

γ ψ

γ ψ

γ

−= −

− −=

=

=

=

=

ɺ

ɺ

ɺ

ɺ

ɺ

ɺ

[ ]T

m m m m m mX V x y zγ ψ=

Missile System Dynamics

State Vector:Model:

[ ]T

z yU a a=

Control (Guidance Commands):

Note:

• Autopilot delays for both ay and az have also been considered while evaluating the guidance laws.

OPTIMAL CONTROL, GUIDANCE AND ESTIMATION

Prof. Radhakant Padhi, AE Dept., IISc-Bangalore44

Target Model

Assumptions:

• A point mass model is assumed

• Measurement of coordinates (xt, yt) are available.

• Target velocity Vt is constant (includes stationary targets too)

• Moving targets can do one of the followings:

• No maneuver (straight line path)

• Constant g maneuvers

• Sinusoidal maneuvers

cos( )

sin( )

ty

t

t

t t t

t t t

a

V

x V

y V

ψ

ψ

ψ

=

=

=

ɺ

ɺ

ɺ

Page 23: Lecture Lecture – ––– 25 225525 Model Predictive Spread ... · Prof. Radhakant Padhi, AE Dept., IISc -Bangalore 11 11 one gets MPSC with Quadratic Parameterization of Control

OPTIMAL CONTROL, GUIDANCE AND ESTIMATION

Prof. Radhakant Padhi, AE Dept., IISc-Bangalore45

Problem Formulation in G-MPSP

Goal:

( ) ( ) ( ) ( ) ( ) ( )T

f m f m f m f m f m fY t t t x t y t z tγ ψ =

( ) ( )*

f fY t Y t→

Define:

OPTIMAL CONTROL, GUIDANCE AND ESTIMATION

Prof. Radhakant Padhi, AE Dept., IISc-Bangalore46

Guess History: Augmented PN

2

1y z z y x

z x x z y

x y y x z

r r r r

r r r rr

r r r r

σ

σ σ

σ

− = − = −

ɺ ɺ ɺ

ɺ ɺ ɺ ɺ

ɺ ɺ ɺ

Line-of-Sight Rate:

Generation of Yaw and Pitch plane Line of sight rates

sin( ) cos( )

sin( )[cos( ) sin( ) ] cos( )

p m x m y

y m m x m y m z

σ ψ σ ψ σ

σ γ ψ σ ψ σ γ σ

= − +

= − + +

ɺ ɺ ɺ

ɺ ɺ ɺ ɺ

Generation of Yaw and Pitch plane latax commands using closing

velocity Vc

, cos( ),c c

x x y y z z

c z e c z m y e c y

r r r r r rV a N V g a N V

rσ γ σ

+ += − = + =

ɺ ɺ ɺɺ ɺ

Page 24: Lecture Lecture – ––– 25 225525 Model Predictive Spread ... · Prof. Radhakant Padhi, AE Dept., IISc -Bangalore 11 11 one gets MPSC with Quadratic Parameterization of Control

OPTIMAL CONTROL, GUIDANCE AND ESTIMATION

Prof. Radhakant Padhi, AE Dept., IISc-Bangalore47

Stationary Targets:

Same initial conditions & Different

Terminal Constraints

0

0

10

10

o

m

o

m

γ

ψ

=

=

Various Constraints in

both the angles

OPTIMAL CONTROL, GUIDANCE AND ESTIMATION

Prof. Radhakant Padhi, AE Dept., IISc-Bangalore48

Stationary Targets:

Same initial conditions & Different

Terminal Constraints (Latax)

Page 25: Lecture Lecture – ––– 25 225525 Model Predictive Spread ... · Prof. Radhakant Padhi, AE Dept., IISc -Bangalore 11 11 one gets MPSC with Quadratic Parameterization of Control

OPTIMAL CONTROL, GUIDANCE AND ESTIMATION

Prof. Radhakant Padhi, AE Dept., IISc-Bangalore49

Stationary Targets:

Perturbation in initial conditions

20

20

f

f

o

m

o

m

γ

ψ

= −

=

Initial Condition perturbation with same terminal constraint

OPTIMAL CONTROL, GUIDANCE AND ESTIMATION

Prof. Radhakant Padhi, AE Dept., IISc-Bangalore50

Stationary Targets:

Perturbation in initial conditions

(Angle Histories)

Page 26: Lecture Lecture – ––– 25 225525 Model Predictive Spread ... · Prof. Radhakant Padhi, AE Dept., IISc -Bangalore 11 11 one gets MPSC with Quadratic Parameterization of Control

OPTIMAL CONTROL, GUIDANCE AND ESTIMATION

Prof. Radhakant Padhi, AE Dept., IISc-Bangalore51

Stationary Targets:

Perturbation in initial conditions

(Latax)

OPTIMAL CONTROL, GUIDANCE AND ESTIMATION

Prof. Radhakant Padhi, AE Dept., IISc-Bangalore52

Maneuvering Targets

GMPSP Vs APN: A Comparison

0

0

80

20

0

20

f

f

o

m

o

m

o

m

o

m

γ

ψ

γ

ψ

= −

=

=

=

Constraint in

both the angles

Page 27: Lecture Lecture – ––– 25 225525 Model Predictive Spread ... · Prof. Radhakant Padhi, AE Dept., IISc -Bangalore 11 11 one gets MPSC with Quadratic Parameterization of Control

OPTIMAL CONTROL, GUIDANCE AND ESTIMATION

Prof. Radhakant Padhi, AE Dept., IISc-Bangalore53

Maneuvering Targets

GMPSP Vs APN: A Comparison

0

0

80

20

0

20

f

f

o

m

o

m

o

m

o

m

γ

ψ

γ

ψ

= −

=

=

=

Constraint in

both the angles

Time histories of Azimuth and Elevation Angle

OPTIMAL CONTROL, GUIDANCE AND ESTIMATION

Prof. Radhakant Padhi, AE Dept., IISc-Bangalore54

Time histories of 3D Angles

Moving/Maneuvering Targets:

Straight line, Constant g & sinusoidal maneuvers

Page 28: Lecture Lecture – ––– 25 225525 Model Predictive Spread ... · Prof. Radhakant Padhi, AE Dept., IISc -Bangalore 11 11 one gets MPSC with Quadratic Parameterization of Control

OPTIMAL CONTROL, GUIDANCE AND ESTIMATION

Prof. Radhakant Padhi, AE Dept., IISc-Bangalore55

Zero Effort Miss (ZEM) Plot

(Sinusoidal Maneuver)

Modified

Definition of

ZEM:

Vt Non Zero- Target

allowed to maneuver

OPTIMAL CONTROL, GUIDANCE AND ESTIMATION

Prof. Radhakant Padhi, AE Dept., IISc-Bangalore56

Concluding Remarks: G-MPSP

� In G-MPSP formulation, discretization of system dynamics is not required in problem formulation.

� In G-MPSP, any higher-order integration technique can be used (e.g. forth-order Runge-Kutta scheme).

� MPSP is a special case of the G-MPSP

� 3-D impact angle constrained guidance problem has been resolved using G-MPSP.

� Results are pretty similar to MPSP results; i.e. Superior results have been obtained as compared to an “Augmented PN law” (especially for maneuvering targets).

Page 29: Lecture Lecture – ––– 25 225525 Model Predictive Spread ... · Prof. Radhakant Padhi, AE Dept., IISc -Bangalore 11 11 one gets MPSC with Quadratic Parameterization of Control

OPTIMAL CONTROL, GUIDANCE AND ESTIMATION

Prof. Radhakant Padhi, AE Dept., IISc-Bangalore57

Thanks for the Attention….!!