T. Yoo, E. Setton, X. Zhu, Pr. Goldsmith and Pr. Girod Department of Electrical Engineering Stanford University

Transcription

1 Cross-layer design for video streaming over wireless ad hoc networks T. Yoo, E. Setton, X. Zhu, Pr. Goldsmith and Pr. Girod Department of Electrical Engineering Stanford University

2 Outline Cross-layer design framework for multimedia communication over ad hoc networks Wireless network model Optimized allocation of capacity and flow Experimental results

3 Wireless ad hoc networks Collection of wireless nodes with no infrastructure Every node can be a source, a destination or a relay Many applications : search and rescue, home network New challenges, especially for multimedia applications

4 Real-time streaming in traditional networks Source coding Application layer Packet scheduling Routing Capacity allocation Channel coding Power control Transport layer Network layer MAC layer Link layer Physical layer

5 Our cross-layer design for real-time media Traffic flows Link state information Loss-resilient Loss-resilient source source coding coding and and packetization packetization Capacity Capacity assignment assignment for for multiple multiple service service classes classes Adaptive Adaptive link link layer layer techniques techniques Rate-distortion preamble Application layer Congestion-distortion optimized optimized scheduling scheduling Transport layer Congestion-distortion optimized optimized routing Network layer routing Link capacities MAC layer Link layer Yoo, Setton, Zhu, Goldsmith & Girod, Cross-layer design for video over wireless ad hoc networks

6 Ad hoc wireless network model For any transmission scheme k the achievable data rate may be computed: Transmitter i Receiver j Interferer l

7 Capacity region achievable by time sharing Set of rates simultaneously achievable: The capacity region is computed by pruning the transmission schemes Assumption No overhead for scheduling Yoo, Setton, Zhu, Goldsmith & Girod, Cross-layer design for video over wireless ad hoc networks

8 Optimized allocation of capacity and flow Congestion defined as the maximum link utilization over the network So minimizing the congestion results in : Subject to : rate constraints at the source and destination flow conservation, flow positivity capacity region constraints Paths extracted from the solution

9 Our cross-layer design for real-time media Traffic flows Link state information Loss-resilient Loss-resilient source source coding coding and and packetization packetization Capacity Capacity assignment assignment for for multiple multiple service service classes classes Adaptive Adaptive link link layer layer techniques techniques Rate-distortion preamble Application layer Congestion-distortion optimized optimized scheduling scheduling Transport layer Congestion-distortion optimized optimized routing Network layer routing Link capacities MAC layer Link layer Yoo, Setton, Zhu, Goldsmith & Girod, Cross-layer design for video over wireless ad hoc networks

10 Experimental scenario 10-node wireless ad hoc network simulated in NS-2 1% random loss rate on each link Random walk mobility model (5 m/s) Capacity and flow re-assigned every 1 second Competing scheduling scheme Links established between neighboring nodes to form mesh Time sharing between the links is optimized to yield maximum capacity Least-congestion flow assignment is performed Video streaming Foreman, CIF sequence, 30 fps, encoded with H.264 IBBP coding structure, GOP length of ms playout deadline, previous frame concealment Yoo, Setton, Zhu, Goldsmith & Girod, Cross-layer design for video over wireless ad hoc networks

11 Capacity and flow allocation example Cross-layer design Oblivious layers

12 Video streaming performance s 5 db 3-fold increase (logarithmic scale)

13 Comparison of the best operating points 1Mbps 6 db 250 kbps

14 Video Streaming Results: Sequence (1) Comparison of the optimal operating point for different number of paths : Foreman QCIF Sequence Cross-layer 38.2 db Oblivious layers 33.1 db

15 Conclusions Cross-layer design framework for multimedia data Joint optimization of all the layers of the protocol stack Gains illustrated for a 2-layer case: MAC and network Benefits of cross-layer design for video over ad hoc networks More adaptive to traffic patterns Decreases the multimedia distortion Future work Achieve the same gain by using distributed algorithms Consider multiple video senders [Zhu et al. to appear PCS 2004] Combine with congestion-distortion optimized scheduling Yoo, Setton, Zhu, Goldsmith & Girod, Cross-layer design for video over wireless ad hoc networks