Multicast Recipient Maximization in

IEEE 802.16j WiMAX Relay Networks

Wen-Hsing Kuo † (郭文興 ) & Jeng-Farn Lee ‡ (李正帆 )

† Department of Electrical Engineering, Yuan Ze University, Taiwan‡ Department of CSIE, National Chung Cheng University, Taiwan

IEEE Transactions on Vehicular Technology, vol. 59, no. 1, Jan. 2010

Outline

• Introduction

• Problem & Goal

• System model

• Challenge

• Proposed Resource– Greedy Approach (GD)

– Dynamic Station Selection (DSS)

• Performance

• Conclusion

Introduction

• WiMAX 802.16 networks– better coverage

– higher throughput

• Wireless resources available for each wireless service is inevitably limited.

• As the capacity of wireless devices improves, the multicast applications, including Video conferencing, have been developed.

Introduction

• Resource-management policy– limit the maximum time slots of a single multicast,

e.g., 10% of a TDD super-frame

– to maintain the quality of different services

• With the given resource budget, a BS should serve as many recipients, i.e., SSs, as possible – to maximize user satisfaction

– to maximize resource utilization

Problem & Goal

• How to address the multicast recipient maximization (MRM) problem in the WiMAX 802.16j network ?

• To propose a resource-allocation scheme for multicast service in downlink transmission– To maximize the total number of recipients

– with the given budget (maximal usable resource)

• To the best of authors’ knowledge, this is the first work to study the problem.

System model

• Resource can be distributed to different transmissions– time slots

• This budget is to be distributed among the BS and RSs – since they are in the same interference range

– only one of them can transmit at the same time

• Routing of each SS is assumed to be decided beforehand– SS accesses the BS either directly or through an RS

– it is impractical that the whole multicast tree can dynamically be formed and adjusted as the channel condition of any recipient changes.

System model

• M RSs & N SSs

• Let not only the SSs but also RSs directly served by the BS be classified as group 0.

• The SSs that receive data via the mth RS be placed in group m, where m > 0.

Group 0

Group 1

Group 2

Group m

System model

• Nm : number of nodes in group m

• S(m,n) : the nth node in group m

• r(m,n) : the resource requirement of S(m,n)

• Since SSs have different bit error rates due to heterogeneous channel conditions, they may require different amounts of resource for receiving the same data from the BS.

Group 2

System model

• im : RS’s order in group 0

– RSm = S(0,im) = S(m,0)

• r(0,im) =0= RS’s resource requirement

Group 2

Group 0

RS2 = S(0,2)= S(2,0)

System model

• Nodes in each group are placed in increasing order of r(m,n), i.e., r(m,1) ≤ r(m,2) ≤ · · · ≤ r(m, Nm)

System model

• Δrm(n) represents the additionally required resource of S(m,n) when the last node S(m, n−1) is served.

– Δrm(n) = r(m,n) – r(m, n – 1)

System model

• Binary function Dm(n)

– RSm can receive from the BS when n nodes are served in group 0. Dm(n) is equal to 1 if im ≤ n and 0 otherwise.

• Um(n) is the number of served SSs when serving S(m,n), starting from the BS.

Challenge

• MRM Problem is NP-Complete

• The goal of MRM is to maximize the total number of served SSs; however, the total resource consumed by the RSs and the BS should not be greater than rbudget.

• Likes the integral knapsack problem (NP-hard)– (1) Object’s price and its weight,

(2) the weight limitation

– (1) Group’s nodal amount and the resource requirements,(2) the budget limitation

Challenge

• MRM Problem is NP-Complete

• MRM is also NP, because the a solution (i.e., {n0, n1, ... , nM }) can be validated by calculating

• MRM problem is NP-hard and NP, so that MRM problem is NP-Complete

M

m mm nUnD0 0 )()(

Proposed Algorithm

• Greedy approach (GD)– um(n): allocation utility of including S(m,n) Um(n), ( um(n) )

)(

)1()()(

nr

nUnUnu

m

mmm

Proposed Algorithm

• Greedy approach (GD)– um(n): allocation utility of including S(m,n) Um(n), ( um(n) )

)(

)1()()(

nr

nUnUnu

m

mmm

25.0

01

)2(

)1()2()2(

0

000

r

UUu

Proposed Algorithm

• Greedy approach (GD)– um(n): allocation utility of including S(m,n) Um(n), ( um(n) )

)(

)1()()(

nr

nUnUnu

m

mmm

11

01

)1(

)0()1()1(

1

111

r

UUu

Proposed Algorithm

• Greedy approach (GD)– um(n): allocation utility of including S(m,n) If rbudget = 2

Proposed Algorithm

• Dynamic Station Selection (DSS)– Um

*(n) : the envelope function of Um(n)

– um*(n) : the optimal allocation utility of including S(m,n)

– U0*(0) = U0(0) = 0

Proposed Algorithm

• Dynamic Station Selection (DSS)

Um*(n), ( um

*(n) )

If rbudget = 2

Performance

• BS at (0,0), RS uniformly distributed, SS random deployed

• Required resource for each node = (1/da)– d : distance between sender and receiver

– a : channel attenuation factor, 2 a 4

Simulation I

• DSS: Dynamic station selection

• OP: Optimal solution

• GD: Greedy algorithm

a = 2 a = 3

5 RSs and 100 SSs

Simulation II

• DSS: Dynamic station selection

• OP: Optimal solution

• GD: Greedy algorithm

Resource budget = 20000a = 2

Conclusion & Future Work

• This paper have considered a resource-allocation problem called MRM for multicast over WiMAX relay networks.

• It proposes a dynamic station selection (DSS) to solve the problem based on the proposed envelope function.

• The future research can be extended in (1) relay networks with more than two hops(2) the distributed approach to solve MRM problem

TheENDThanks for your attention !