Abstract
We investigate the problem of streaming multiplevideos over multi-hop cognitive radio (CR) networks.Fine-Granularity-Scalability (FGS) and Medium-Grain-Scalable(MGS) videos are adopted to accommodate the heterogeneityamong channel availabilities and dynamic network conditions.We obtain a mixed integer nonlinear programming (MINLP)problem formulation, with objectives to maximize the overallreceived video quality and to achieve fairness among the videosessions, while bounding the collision rate with primary usersunder the presence of spectrum sensing errors. We first solve theMINLP problem using a centralized sequential fixing algorithm,and derive upper and lower bounds for the objective value.We then apply dual decomposition to develop a distributedalgorithm and prove its optimality and convergence conditions.The proposed algorithms are evaluated with simulations and areshown to be effective in supporting concurrent scalable videosessions in multi-hop CR networks.Index Terms—Cross-layer optimization, dynamic spectrumaccess, distributed algorithm, multi-hop cognitive radio networks,video streaming.
I. INTRODUCTION
A cognitive radio (CR) is an advanced radio device withinterface(s) to sense the radio environment, an intelligentagent for decision-making based on radio environment andpast experience, and a frequency-agile radio that can be tunedto a wide range of frequency bands and operate from there.CR represents a paradigm change in spectrum regulation andaccess, from exclusive use by licensed, or primary, usersto sharing spectrum with, and dynamic spectrum access forunlicensed, or secondary, users. It has profound impact onhow future wireless networks will be designed and operated.The high potential of CRs has attracted substantial interest.The mainstream CR research has focused on developing effectivespectrum sensing and access techniques (see [1] and [2]and reference therein). Although considerable understandingshave been gained on various aspects of CR, the importantproblem of guaranteeing application performance has notbeen the focus of major CR research. To this end, we findspectrum-intensive and rate-adaptive multimedia, or video as a reference application, makes excellent use of the enhancedspectrum efficiency in CR networks. Unlike data, where eachbit should be delivered, video is loss tolerant and rate adaptive.They are highly suited for CR networks, where the availablebandwidth heavily depends on primary user behavior. Gracefuldegradation of video quality can be achieved as spectrumopportunities change over time.In our prior work [3], we investigated the problem ofmulticasting Fine-Granularity-Scalablity (FGS) video in aninfrastructure-based CR network and demonstrated the feasibilityof video over CR networks. In this paper, we studythe more challenging problem of video over multi-hop CRnetworks. As illustrated in Fig. 1, we consider an infrastructurelessCR network co-located with one or more fixedprimary networks. CR users non-intrusively exploit whitespaces in the licensed bands for streaming multiple videos.The objective is two-fold: to maximize the overall videoquality and to achieve fairness among the concurrent videosessions, subject to bounded interference to primary users.We adopt FGS videos to accommodate heterogeneous channelavailabilities and dynamic network conditions [4]. FGSvideo is coded into a base layer (BL) and an enhancementlayer (EL). The EL can be truncated at any bit location, whileall the remaining bits are still useful for decoding. This featuresimplifies the design of video streaming systems. We alsoconsider H.264/SVC Medium-Grain-Scalable (MGS) videosin this paper. MGS is shown to achieve better rate-distortionperformance over MPEG-4 FGS, although MGS only hasNetwork Abstraction Layer (NAL) unit-based granularity

Download full report
http://eng.auburn.edu/~szm0001/twireless...r-2010.pdf