1 MULTIPATH VIDEO TRANSPORT OVER AD HOC NETWORKS Shiwen Mao, Shunan Lin, Yao Wang, Panwar, S.S., Yihan Li, Wireless Communications, IEEE Volume 12, Issue 4, Aug Page(s):

2 Outline ABSTRACT INTRODUCTION MULTIPATH MULTIMEDIA TRANSPORT ARCHITECTURE OVERVIEW MULTISTREAM VIDEO CODING FEEDBACK-BASED REFERENCE PICTURE SELECTI ON LAYERED CODING WITH SELECTIVE ARQ MULTIPLE DESCRIPTION CODING COMPARISON AND DISCUSSION PERFORMANCE EVALUATION CONCLUSIONS

3 ABSTRACT Real-time multimedia transport has stringent bandwidt h, delay, and loss requirements. It is a great challenge to support such applications in w ireless ad hoc networks, which are characterized by fr equent link failures and congestion. We describe a framework for multipath video transport over wireless ad hoc networks, and examine its esse ntial components, including multistream video coding, multipath routing, and transport mechanisms. Our experiments show that multipath transport is a pro mising technique for efficient video communications o ver ad hoc networks.

4 INTRODUCTION Ad hoc network A wireless link has a high transmission error rate beca use of shadowing, fading, path loss, and interference from other transmitting users. The path used for the video session should be stable f or most of the video session period. Multipath transport, by which multiple paths are used t o transfer data for an end-to-end session, is highly suit able for ad hoc networks Multipath transport has been combined with multiple description coding for video transport.

5 MULTIPATH MULTIMEDIA TRANSPORT ARCHITECTURE OVERVIEW At the sender, raw video is first compressed by a video encoder into M streams. We call the coder a multistrea m coder. Then the streams are partitioned and assigned to K pa ths by a traffic allocator. These paths are maintained by a multipath routing prot ocol. When the flows arrive at the receiver, they are first put into a resequencing buffer to restore the original orde r. Finally, the video data is extracted from the resequenci ng buffer to be decoded and displayed.

6 The general architecture for the multipath transport of real-time multimedia applications: point to point video communications

7 The general architecture for the multipath transport of realtime multimedia applications: concurrent streamingmulticast using two trees

8 ADVANTAGES AND DESIGN TRADE-OFFS The study in [10] demonstrates that the most significan t performance gain is achieved when K increase s from 1 to 2, with lesser improvements achieved for further increases in K. As a result, a baseline system having M = 2 and K = 2 will provide significant performance gains at a moderat e cost.

9 MULTISTREAM VIDEO CODING One must carefully design the video coder to generate streams so that the loss in one stream does not adver sely affect the decoding of other streams. We illustrate how to adapt a video coder to multipath tr ansport for better performance.

10 FEEDBACK-BASED REFERENCE PICTURE SELECTION Sending coded even and odd frames separately A more network-aware coding method is used, which s elects the reference picture based on feedback and es timated path status. Note that the RPS scheme is only applicable for online coding, because it adapts the encoding operation base d on channel feedback.

11 The RPS scheme The arrow from a frame indicates the reference picture used in coding that frame

12 LAYERED CODING WITH SELECTIVE ARQ (automatic repeat request) A video frame is coded into a base layer and one or m ore enhancement layers. Reception of the base layer can provide low but accep table quality. Reception of the enhancement layer(s) can further imp rove the quality over the base layer alone, but the enh ancement layers cannot be decoded without the base layer. Disjoint paths are used and then base layer packet retransmission using the enhancement layer path coul d have higher success probability and lower delay.

13 LC with ARQ the two-path layered video transmission model with selective ARQ for base layer packets

14 MULTIPLE DESCRIPTION CODING(MDC) MDC is a technique that generates multiple equally im portant descriptions. Two descriptions are generated by sending even pictu res as one description and odd pictures as the oth er. When both descriptions are received, the decoder can reconstruct a picture When only one description is received, the decoder ca n only generate the side prediction. Nor does it require any receiver feedback.

15 COMPARISON AND DISCUSSION LC with ARQ has the best performance for medium an d high loss rates. RPS outperforms the other two when the loss rate is v ery low [6]. However, because RPS adapts its encoder based on receiver feedback,

16 Comparison of the three schemes

17 MD (multiple description using two paths) vs. SD (single description) MD vs. SD distortion as the number of joint links between two paths varies.

18 The way of using multiple paths is by using Stream Co ntrol Transmission Protocol (SCTP) Multiflow Real-Time Transport Protocol (MRTP) Multiflow Real-Time Transport Control Protocol (MRT CP)

19 The protocol stack using MRTP

20 PERFORMANCE EVALUATION We found that each of these three techniques is best suited for a particular environment, depending on the availability of feedback channels, the end-to-end delay constraint, and the error characteristics of the paths. we present a comparison study of a multipath transport (MP T) system (using two paths) with a single path transport (SP T) system for video streaming in an ad hoc network For MPT, a multipath routing extension of the DSR algorith m or MDSR [6], and MRTP/MRTCP were used. SPT consist s of the DSR routing protocol and RTP/RTCP. For both syst ems, the MDMC codec was used to generate two descriptio ns, With the MPT system, the two descriptions are sent over tw o maximally disjoint paths maintained by MDSR. For the SP T system, the two descriptions are multiplexed onto a single path established by DSR.

21 Comparison of single path transport and multipath transport (two paths used)

22 Comparison of single path transport and multipath transport The average PSNR is a) dB; b) dB. peak signal-to-noise ratio (PSNR) the deepest drop occurs when a large burst of losses i n one description overlaps with a loss burst of the othe r flow. It can be seen that SPT has higher loss rates than MP T. Furthermore, a careful examination shows that the t wo loss traces of SPT are highly correlated. Therefore, the PSNR curve in Fig. 5a has more frequent and sev ere degradations than that in Fig. 5b. MPT achieves a significant 1.26 dB gain in average PSNR over SPT in this experiment.

23 CONCLUSIONS In this article whey describe a framework for multipath video transport over wireless ad hoc networks, and examine its essential components, including mult istream video coding, multipath routing, and transp ort mechanisms. We present example solutions for each component an d the performance achievable with a system integr ating these components. We demonstrate that multipath transport combined wit h appropriate video coding techniques can lead to substantial gain over a system using a single path. It is worth noting that these schemes also apply to wire d mesh networks (e.g., the Internet).