IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 59, NO. 12, DECEMBER 2011 3485 Cognitive Base Stations in LTE/3GPP Femtocells: A Correlated Equilibrium Game-Theoretic Approach Jane Wei Huang, Student Member, IEEE, and Vikram Krishnamurthy, Fellow, IEEE AbstractThis paper considers downlink spectrum allocation in a long term evolution (LTE) system macrocell which contains multiple femtocells. By incorporating cognitive capabilities into femtocell base stations, the Home evolved Node Bs (HeNBs) can be formulated as secondary base stations seeking to maximize the spectrum utility while minimizing interference to primary base stations (evolved Node-Bs). The competition amongst cognitive HeNBs for spectrum resources is formulated as a non-cooperative game-theoretic learning problem where each agent (HeNB) seeks to adapt its strategy in real time. We formulate the resource block (RB) allocation among HeNBs in the downlink of a LTE system using a game-theoretic framework, where the correlated equi- librium solutions of the formulated game are being investigated. A distributed RB access algorithm is proposed to compute the correlated equilibrium RB allocation policy. AbstractLTE/3GPP system, cognitive base stations, fem- tocells, self-organized network, correlated equilibrium, game- theoretic learning. GLOSSARY 3GPP 3rd Generation Partnership Project 4G 4th Generation AWGN Additive White Gaussian Noise CCI Co-Channel Interference DFP Dynamic Frequency Planning DSL Digital Subscriber Line eNB Evolved Node-B HeNB Home Evolved Node-B LTE Long Term Evolution Mbps Megabit Per Second OFDMA Orthogonal Frequency-Division Multiple Access QoS Quality of Service RB Resource Block SON Self-Organized Network UMTS Universal Mobile Telecommunication System WLAN Wireless Local Area Network I. I NTRODUCTION A N important feature of Long Term Evolution (LTE)/3rd Generation Partnership Project (3GPP) systems [1] is that it allows distributed implementation of femtocells to meet a variety of service requirements. The femtocell access points, denoted as Home evolved Node-B (HeNB) in 3GPP, are low- cost, low-power, plug-and-play cellular base stations. In order Paper approved by E. G. Larsson, the Editor for Game Theory and Communications Systems Optimization of the IEEE Communications Society. Manuscript received November 23, 2010; revised June 6, 2011. The authors are with the Department of Electrical and Computer Engineer- ing, The University of British Columbia, 5500-2332 Main Mall, Vancouver, BC V6T 1Z4, Canada (e-mail: {janeh, vikramk}@ece.ubc.ca). Digital Object Identier 10.1109/TCOMM.2011.093011.100693 to provide broadband connectivity, these HeNBs will need to possess adaptive/cognitive facilities. In the October 2010 release of 3GPP, HeNBs are described as self-optimized nodes in a Self-Organized Network (SON) which need to maintain quality of service (QoS) with minimal intervention from the service operator [2]. HeNBs are equipped with cognitive functionalities for load balancing, interference management, random access channel optimization, capacity and coverage optimization, and handover parameter optimization. With the above motivation, this paper considers spectrum resource allocation in an orthogonal frequency division multi- ple access (OFDMA) LTE downlink system which consists of a macrocell base station (evolved Node-B (eNB)) and multiple femtocell base stations (HeNBs). By incorporating cognitive capabilities into these self-optimized femtocell base stations, the cognitive HeNBs aim to maximize the spectrum utility by utilizing the unoccupied frequencies while minimizing interference to the eNB (primary base station) in a spectrum overlay LTE system. The unit of spectrum resource to be allocated in a LTE system is called a resource block (RB) and it is comprised of 12 subcarriers at a 15 kHz spacing. Given the RB occupancy of the eNB, the competition for the spectrum resources among HeNBs can be formulated in a game-theoretic setting [3]. Instead of computing the Nash equilibrium policy of the formulated game, we seek to characterize and compute the correlated equilibrium policy set [4], [5]. The set of correlated equilibria is a convex polytope. It includes the set of Nash equilibria indeed the Nash equilibria are isolated points at the extrema of this set [6], [7]. The set of correlated equilibria [5] is arguably the most natural attractive set for a decentralized adaptive algorithm such as the one considered here, and describes a condition of competitive optimality between agents (cognitive femtocell base stations). It is more preferable than Nash equilibria since it directly considers the ability of agents to coordinate their actions. This coordination can lead to higher performance than if each agent was required to act in isolation. Indeed, Hart and Mas-Colell observe in [8] that, for most simple adaptive procedures, ... there is a natural coordination device: the common history, observed by all players. It is thus not reasonable to expect that, at the end, independence among players will obtain. Since the set of correlated equilibria is convex, fairness between players can be addressed in this domain. Finally, decentralized, online adaptive procedures naturally converge to the correlated equilibria, whereas the same is not true for Nash equilibria (the so-called law of conservation of coordination [9]). 0090-6778/11$25.00 c ⃝ 2011 IEEE