IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 60, NO. 11, NOVEMBER 2012 3407 Compressed Video Streaming in Cooperative Wireless Networks with Interfering Transmissions Antonios Argyriou, Member, IEEE Abstract—In this paper, we present a video streaming system for wireless networks that employs utility optimization of pre- compressed video at the application layer, together with a novel cooperative wireless physical layer (PHY) that allows interfering transmissions. Our system model considers multiple and independent unicast or multicast transmissions between network nodes while a number of them serve as relays. For this new PHY the average transmission rate that each sender- destination pair can achieve is estimated first. Next, we show that the utility optimization problem can be simplified due to the features of the proposed PHY. Subsequently, we devise a utility optimization algorithm that is executed independently at each sender and it derives the optimal video packets that should be sent in the wireless link given the calculated rate constraint. Simulation results demonstrate that for unicast video streaming scenarios and with the utility optimization framework, allowing packets to interfere is a better choice than employing a typical cooperative PHY. For the case of multicast video delivery the performance of the proposed scheme is significantly improved for good channel conditions while the improvement is minimized in the low channel SNR regime when compared to the unicast scenario. Index Terms—Video streaming, wireless networks, interfer- ence, cooperative systems, physical layer network coding. I. I NTRODUCTION V IDEO transmission through streaming in wireless net- works is still one of the most challenging problems of information transmission [1]. Contrary to elastic video distribution mechanisms (e.g. video download) that have seen huge benefits from the increases in network capacity, video streaming in wireless communication systems still suffers from several problems: Channel errors are difficult to be corrected in real-time in ways that are not observable by the user (e.g. with error concealment [1]). Channel coding mechanisms with forward error correction (FEC) or automatic repeat request (ARQ) introduce significant overheads that cannot solve en- tirely the problem of error resiliency due to the dynamic nature of the wireless channels. Furthermore, asymmetry of multi- hop wireless networks exacerbates the problem of video trans- mission through such a path [2]. Bandwidth fluctuations in Paper approved by B. Sikdar, the Editor for Wireless Packet Access and Cross-Layer Design of the IEEE Communications Society. Manuscript received November 16, 2011; revised May 22, 2012. The material in this paper was presented in part at the Packet Video Workshop, Munich, Germany, May 2012. A. Argyriou is with the Department of Computer and Communications Engineering, University of Thessaly, Volos, 38221, Greece (e-mail: anar- gyr@uth.gr). The authors would like to acknowledge the support from the European Commission through the Marie Curie Intra-European Fellowship WINIE- 273041 and the STREP project CONECT (FP7ICT257616). Digital Object Identifier 10.1109/TCOMM.2012.082712.110772 the wireless case create very frequently the well known video stuttering problems requiring thus bigger playback buffers and larger playback delay. Another important parameter to consider is the inherent characteristic of the video data. Video streams are charac- terized by different rate-distortion (RD) characteristics which means that data units do not have the same importance for the video decoding process [3]. Therefore, existing solutions have to adapt not only to the channel conditions but also to the transmitted video stream. Naturally, the RD tradeoff is part of source coding and transmission system. In imperfect channels joint source-channel coding should be exercised [4], [5]. However, when the source description is already coded, the options regarding the real-time actions that can occur during transmission are limited [6]. There is a need to radically address these problems and we believe that this can be accomplished when the technical progress at the lower layers of the protocol stack are carefully taken into account. One of the most promising approaches for achieving considerably better throughput and lowering delay at the lower layers of the protocol stack, is to allow packet transmissions to interfere for the advantage of all the involved users. Interference exploitation has recently attracted considerable attention from the wireless communication and information theory communities so as to increase the network capacity [7], [8], [9]. Physical layer network coding (PLNC) is another term frequently used for this technique. Allowing packet transmissions to interfere means that wireless signals are transmitted concurrently, reducing thus the total duration of the transmissions. However, a subset of the interfering signals must be known at the final destinations (a-priori information) in order to be able to decode the desired packet. By overhearing signals and with the help of a relay more transmissions can occur per time unit [10]. However, for the particular class of packet-based video communication systems the idea of allowing packets to in- terfere might not be always beneficial. The reason is that with video communications when a certain packet is transmitted the importance of this packet might be completely different from subsequent or previous packets as we explained earlier. A failed attempt to remove interference and decode a PHY packet might result in the inability to deliver an important video packet to the application. This failure might not be a problem for elastic data transfer in wireless networks since the performance is averaged over several channel realizations and every packet erasure is equally important. Therefore, in the case of video payload the problem that has to be solved is to identify the conditions for allowing specific packet transmis- 0090-6778/12$31.00 c  2012 IEEE