Cooperative Spectrum Sensing in Cognitive RadioNetworksGhurumuruhan Ganesanand Ye (Geoffrey) LiSchool of Electrical and Computer EngineeringGeorgia Institute of Technology, Atlanta, Georgia 303320250AbstractIn cognitive networks, cognitive (unlicensed) usersneed to continuously monitor spectrum to detect the presence ofprimary (licensed) users. In this paper, we illustrate the benefitsof cooperation in cognitive radio. We show that by allowing thecognitive radios operating in the same band to cooperate we canreduce the detection time and thus increasing their agility. Wefirst consider the case of two cognitive users and show how theinherent asymmetry in the network can be exploited to increasethe probability of detection. We then extend our work to multiplecognitive user networks. We also propose a practical algorithmwhich allows cooperation in random networks.Index Termscognitive radio, cooperative diversity, agilitygain, detection time.I. INTRODUCTIONRecently there has been tremendous interest in the field ofsoftware defined radio (SDR) and its relatively newer versioncognitive radio (CR). SDR, which has been introduced in 1,achieves significant improvements over services offered bycurrent wireless networks. With SDR, the software embeddedin a radio cell phone, for example, would define the parametersunder which the phone should operate in real-time as its usermoves from place to place. CR is even smarter than SDR.CR is designed to be aware of and sensitive to the changesin its surroundings. Thus it learns from its environment andperforms functions that best serve its user. This is a verycrucial feature of CR networks since, currently, the FederalCommunications Commission (FCC) is reviewing its policiesregarding the usage of licensed bands by unlicensed users 2.In this context, the greatest advantage of CR is that it can beoperated in licensed bands without a license.Since the cognitive (unlicensed) users are utilizing thelicensed band, they must detect the presence of licensed(primary) users in a very short time and must vacate the bandfor the primary users. Thus one of the major challenges thatconfronts this technology is: how do the cognitive (unlicensed)radios sense the presence of the primary (licensed) user?One may expect this to be trivial but as shown in 3, thereare fundamental limits to the detection capabilities of CRnetworks. In this paper, we show improvement in spectrumsensing capabilities through cooperation between individualcognitive users.Cooperative schemes with orthogonal transmission in aTDMA system have been recently proposed in 4 and 5. Corresponding Author: email : guruece.gatech.edu, phone/fax : 404-385-3012/894-7883It has been shown in 4 that two user single hop networksin which one of the user acts as a relay for the other, resultin lower outage probabilities. In particular, it is shown thatthe amplify-and-forward (AF) protocol 4, in which the relaytransmits the signal obtained from the transmitter without anyprocessing, achieves full diversity. In this paper, we studythe effect of the AF cooperation protocol on the spectrumsensing capabilities of multi-user single-carrier cognitive radionetworks. Analysis of multi-user single carrier networks can befound in 6. In 6, we describe a cooperation scheme wherethe relay user is assumed to have no power constraint. Howeverin practice all relay users have power constraint. In this paper,we describe a power constrained cooperation scheme.The paper is organized as follows: In Section II, we describethe system model and formulate the problem of the primaryuser detection in a simple two user cooperative network.We also propose a practical cooperation scheme to detectthe primary user. To show that cooperation leads to reduceddetection time, in Section III, we consider a simple two-userasymmetric cognitive radio network and derive expressions foragility gain. In Section IV, we analyze multiuser single-carriercognitive networks with and without centralized schedulingand derive precise conditions under which agility gain isachieved. Finally, in Section V, we present our conclusion.We first introduce a few notations we shall be using through-out the paper. Consider any two non-negative functions func-tions f(x) and g(x). The following are standard notations 7:1) We say f(x) = (g(x), if there exists a constant C2 0such that f(x) C2g(x) for sufficiently large x.2) We say f(x) = (g(x) if there exists constants C3,C40 such that C3g(x) f(x) C4g(x) for sufficiently large x.The following are the symbols used throughout the paper:U1,U2: Cognitive (unlicensed) usersP1: Received power at U1due to the primary userP2: Received power at U2due to the primary userG12: Channel gain between U1and U2 : Primary user indicator; = 1 implies presence of theprimary user and = 0 implies its absenceII. COOPERATION SCHEMEIn this section, we describe the channel model that willbe used throughout the paper, formulate the primary userdetection problem and propose a practical