Cooperative Layered Wireless Video Multicast

Ozgu Alay, Thanasis Korakis, Yao Wang, Elza Erkip, Shivendra Panwar

Introduction Video multicast over wireless channels

Wireless video applications are emergingMulticast is effectiveWireless video multicast is still a

challenging problem High packet loss rate Bandwidth variations

Cooperation is a natural solutionHigher spatial diversityAdaptive to network conditions

Prior Work: Cooperation for Unicast physical-layer cooperation for point-

to-point video communicationSingle-layer cooperationlayered cooperation

MAC-layer cooperation for point-to-point communication

Each receiver has different channel quality

Conventional Multicast Source transmits based on furthermost

receiver the receivers with a good channel quality

unnecessarily suffer and see a lower quality video .

Why Cooperative Multicasting ?

Why Cooperative Multicasting ?

Cooperative Multicast Divide all the receivers into

two groups such that receivers in Group 1 have better average channel quality than Group 2

Sender targets receivers with good channel quality (Group1)

These receivers relay the video to other receivers (Group2)

It is likely that we achieve a larger coverage area (Extended Group 2).

Both groups see better quality

Relay 1

Relay 4 AP

Relay 3

Relay 2

Group 1

Group 2

Extended Group 2

rrelay

rext

R2

,r1

,r dRd

R1

,r2C

B

A

Received Video Rates

T1 T2 T2

T

Design Variables Number of relays N Sustainable rates (R1, R2) or transmission ranges

(r1, r2) Time partition (T1, T2)

N controls the tradeoff between R2 and T2

How to optimize? Maximize the average quality All users have same quality Group1 has better quality

Approach For a particular (r1, r2) we determine the

optimum (T1, T2) and N in two steps.

1. We first determine the user partition with a minimum number of relays.

2. Then for this user partition, we find the optimum T1 and T2 (time scheduling) that maximizes the system performance index

By repeating the above procedure for all possible (r1, r2) we find the optimum user partition and time scheduling that maximizes the performance criterion.

User Partition Goal: Find minimum

number of relays N that covers all the users

User partition is defined by (r1, r2) and the separation angle where,

N = 2/2

rrelay

rext

R2

,r1

, r d R d

R1

,r2C

B

A

User Partition

We define max as the maximum angle which satisfies the constraints below,

rrelay

rext

R2

,r1

, r d R d

R1

,r2C

B

A

Optimum User Partition

is maximum when

Then, using cosine theorem

r1rd

r2C

B

Amax

Optimum User Partition

Then N is,

And rext can be computed as

rrelay

rext

R2

,r1

, r d R d

R1

,r2C

B

A

Time Scheduling and Performance Metric We use exhaustive search over a

discretizied space of feasible T1 and T2, for each candidate T1 and T2, determine Rv1 and Rv2 and correspondingly D1 and D2.

Here D1(Rv1) is the distortion of Group 1 receivers and D2(Rv2) is the distortion for Group 2 receivers.

Minimum Average Distortion

N1 and N2 are the number of users in Group 1 and Group 2, respectively.

Equal Distortion at all users

We require all the receivers have the same distortion.

In other words, we find the optimum user partition and time scheduling that minimizes D1(Rv1) = D2(Rv2).

Best Quality at Group 1 users Considering that relays are spending

their own resources to help others, We find the optimum user partition

and time scheduling that minimizes D1(Rv1) while guaranteeing Rv2 = Rd

Sustainable Rates vs. Distance with IEEE 802.11b

r1=61m, R1=11 Mbps

r2=72m, R2=5.5 Mbps

r3=100m, R3=1 Mbps

r 1, R 1r 3, R 3

r 2, R 2

Example Scenario 802.11b based WLAN Uniformly distributed

users within 100m radius (r=100m)

Achievable rate with direct transmission to all users,

Rd = 1 Mbps =0.75 Soccer

704x576 resolution 240 frames

r , R

Performance

Visual Quality

750 kbps ( 29.84 dB )

1.178 Mbps ( 30.42 dB ) 3.75 Mbps ( 33.32 dB )

Conclusion

User cooperation can improve the quality of service in video multicastEqual quality at all usersBetter quality at selected usersAll better than direct transmission

Optimization of relay selection, user partition, and transmission scheduling depends on the chosen multicast performance criterion