throughput and delay scaling of cognitive radio networks...
TRANSCRIPT
UCLA ENGINEERING Computer Science
Throughput and Delay Scaling of Cognitive Radio Networks with Heterogeneous Mobile Users
Pengyuan Du*, Mario Gerla*, Xinbing Wang†
*Department of Computer Science, UCLA, USA
†Department of Electronic Engineering, Shanghai Jiao Tong University, China
UCLA ENGINEERING Computer Science
Related Work
The research of ad-hoc net scaling law starts with the seminal
work of Gupta and Kumar in 2000 [1].
Consider a static network with users.
A multi-hop transmission scheme achieves per-node
throughput of , with theoretical upper
bound .
Mobility allows to trade-off throughput with delay [8]
[1] Gupta, Piyush, and Panganmala R. Kumar. "The capacity of wireless networks." Information
Theory, IEEE Transactions on 46.2 (2000): 388-404.
[8] Grossglauser M, Tse D N C. Mobility increases the capacity of ad hoc wireless networks[J].
Networking, Ieee/Acm Transactions On, 2002, 10(4): 477-486.
n
(1/ log )n n
(1/ )n
UCLA ENGINEERING Computer Science
Related Work
Shortage of radio spectrum drives the research on Cognitive
Radio Network (CRN).
Primary (PU) and secondary users (SU) operate at the same
time, space and share the spectrum.
SU use spectrum opportunistically.
PU and SU can achieve the same throughput and delay
scaling laws as stand-alone wireless networks [3][4].
[3] Jeon S W, Devroye N, Vu M, et al. Cognitive networks achieve throughput scaling of a homogeneous network[J]. Information Theory, IEEE Transactions on, 2011, 57(8): 5103-5115.
[4] Yin C, Gao L, Cui S. Scaling laws for overlaid wireless networks: a cognitive radio network versus a primary network[J]. IEEE/ACM Transactions on Networking (TON), 2010, 18(4): 1317-1329.
UCLA ENGINEERING Computer Science
Related Work Moreover, recent research showed that Mobility in CRN can
bring even further benefits [5-7].
The movement of secondary users will facilitate possible co-
operations between PU and SU.
Secondary users are willing to relay for primary users.
Throughput and delay scaling is improved to near-optimal for
both PU and SU.
[5] Gao L, Zhang R, Yin C, et al. Throughput and delay scaling in supportive two-tier
networks[J]. Selected Areas in Communications, IEEE Journal on, 2012, 30(2): 415-424.
[6] X. Wang, L. Fu, Y. Li, Z. Cao and X. Gan, “Mobility Reduces the Number of Secondary
Users in Cognitive Radio Network” in Proc. of IEEE GLOBECOM, 2011.
[7] Li Y, Wang X, Tian X, et al. Scaling laws for cognitive radio network with heterogeneous
mobile secondary users[C]//INFOCOM, 2012 Proceedings IEEE. IEEE, 2012: 46-54.
UCLA ENGINEERING Computer Science
Motivation & Objective
The mobility models in previous works are not general and
representative (only SU are mobile while PU are static).
We propose a General Heterogeneous Speed-Restricted Mobility
(GHSM) model.
PU and SU possess heterogeneous mobility patterns.
Exploit the mobility of PU and SU to improve throughput and
delay performance.
UCLA ENGINEERING Computer Science
Outline Introduction
Scaling Laws of Mobile Cognitive Network
Network Model
Routing and Scheduling Protocols
Throughput and Delay Performance
Conclusion
UCLA ENGINEERING Computer Science
PU and SU co-exist in a unit torus
SU are denser than PU.
Source and destination pairs are randomly grouped.
The unit area is divided into primary and secondary cells for
connectivity [3].
Network Model
[3] Jeon S W, Devroye N, Vu M, et al. Cognitive networks achieve throughput scaling of a homogeneous network[J]. Information Theory, IEEE Transactions on, 2011, 57(8): 5103-
5115.
UCLA ENGINEERING Computer Science
GHSM Model
At the beginning, all
users are uniformly
and randomly
distributed over the
network.
Then they would
move within a
circular area
according to a i.i.d.
mobility model.
Network Model
UCLA ENGINEERING Computer Science
Multiple PU grids overlaid on the network A similar setting for SU nodes Movements restricted to cells
UCLA ENGINEERING Computer Science
GHSM Model
To determine the cells in which PU and SU move, we define:
where denotes the i-th cell layer.
Users moving in layer have moving area
and radius , where is a random
variable.
Network Model
{ | 0 }iT i h
iT
0i
hiA n
iT
0
2= ( )
i
i hi
AR n
0
UCLA ENGINEERING Computer Science
where denotes the i-th moving pattern. The moving area
becomes smaller and smaller as i increases.
covers the whole network, covers an area of .
0T hT
0
1
n
GHSM Model
To determine the moving area of PU and SU, we first define:
Network Model
{ | 0 }iT i h
iTMoving patterns of primary users are selected as follows: Moving patterns of secondary users are selected as follows:
UCLA ENGINEERING Computer Science
GHSM Model
For the primary network, we assign nodes to each moving
pattern (ie each cell layer)
For the secondary network, we assign nodes to each
moving pattern.
Network Model
n
n
UCLA ENGINEERING Computer Science
UCLA ENGINEERING Computer Science
Network Model
Comparison
Previous Mobility
Models
GHSM Model in this
work
Number of PU
Number of SU
Moving Pattern of PU Static Mobile
Moving Pattern of SU Mobile Mobile
( 1)h n
n
( )p
( )s
( 1)pN h n
( 1)sM h n
UCLA ENGINEERING Computer Science
Primary Routing Scheme
Primary routing scheme utilizes the heterogeneous mobility of both
PU and SU to propagate packets to the destination.
Packets are delivered hierarchically to the destination.
Network Protocols
Delivered!
UCLA ENGINEERING Computer Science
UCLA ENGINEERING Computer Science
Network Protocols
Secondary Routing Scheme
Secondary routing scheme only utilizes the mobility of SU to
propagate packets to the destination.
UCLA ENGINEERING Computer Science
Network Protocols
Different from Gupta and Kumar’s model, for CRN to work,
the following scheme is introduced from work [2]:
64-TDMA scheduling scheme
Preservation region
On this basis:
PU are scheduled to relay packets from every primary moving
pattern;
SU are scheduled to relay packets from every primary and
secondary pattern.
[2] Yin C, Gao L, Cui S. Scaling laws for overlaid wireless networks: a cognitive radio network versus a primary network[J]. IEEE/ACM Transactions on Networking (TON), 2010, 18(4): 1317-1329.
UCLA ENGINEERING Computer Science
Performance Analysis
Impact of Different Moving Pattern
Achieve near-optimal throughput regardless of different
moving patterns.
Packet delay will decrease when moving pattern increases.
with , which indicates a
similar result as in [8], i.e., where .
Heterogeneous mobility introduces transmission diversity
into the network.
3
0 ( ) ( log )D N n n 0 3
1( ) ( )
logN
n
[8] Grossglauser M, Tse D N C. Mobility increases the capacity of ad hoc wireless
networks[J]. Networking, Ieee/Acm Transactions On, 2002, 10(4): 477-486.
( ) ( )D n n n ( ) 1n
UCLA ENGINEERING Computer Science
Li’s scheme Our scheme
Throughput
of PU
Delay
of PU
Performance Analysis
Optimal Performance
Comparison with Li’s scheme in [7]:
0 3
0
4
0
3log log( log ), 0
log( )
3log log(log ),
log
optimal
nn n
nD n
nn
n
3
1( ) ( )
logN
n
01 3
0
4
0
log log( log ), 0 1
log( )
log log(log ), 1
log
optimal
nn n
nD N
nn
n
1( ) ( )
logn
n
[7] Li Y, Wang X, Tian X, et al. Scaling laws for cognitive radio network with heterogeneous
mobile secondary users[C]//INFOCOM, 2012 Proceedings IEEE. IEEE, 2012: 46-54.
For SU, we have similar results as [7]
UCLA ENGINEERING Computer Science
Outline Introduction
Scaling Laws of Mobile Cognitive Network
Conclusion
UCLA ENGINEERING Computer Science
Propose a GHSM model
Cooperative routing strategy
Better optimal performance
We also prove that the secondary network can still achieve the
same performance as in previous works even though secondary
network may suffer from the moving ability gaps..
Conclusion
This paper studies the throughput and delay scaling laws of a
cognitive radio network with the General Heterogeneous Speed-
restricted Model.
UCLA ENGINEERING Computer Science
Q&A