Decentralized Link Adaptation for Multi-link MIMO Interference System Mirza Tayyab Mehmood, Abdelwaheb Marzouki and Djamal Zeghlache Département Réseaux et Services Multimédia Mobiles TELECOM & Management SudParis Evry, France {tayyab.mirza, abdelwaheb.marzouki, djamal.zeghlache}@it-sudparis.eu Abstract—A decentralized link adaptation algorithm designed for multi-link MIMO interference systems to jointly optimize the power and the number of links is presented in this contribution. Game theory is adapted and used for both allocating resources and admitting users (i.e. with embedded congestion control). The MIMO interference system is mapped into a multi-player game. The solution of the proposed game is provided by a new gradient based decentralized link adaptation algorithm. The new algorithm allocates adaptively the power and the modulation scheme of the transmitter and maintains the optimum number of links using soft decision criteria based on the link metric. Each link decision depends on the type of traffic and the QoS requirements. An analytical framework is provided along with simulations and an analysis of the algorithm behavior under different system conditions. This type of algorithm is applicable to spectrum sharing and power management to reduce interference and energy consumption. I. INTRODUCTION Efficient resource allocation is essential in wireless communication to save bandwidth and power. Different users may have conflicting interests regarding resources; excessive allocation of resources to some users affects other users and may prevent fulfilling quality-of-service (QoS) requirements. A fair and efficient resource allocation is desirable. State of the art in the field of resource allocation for communication networks can be found in [1-5]. To achieve distributed control for resource allocation, tools beyond traditional optimization such as game theory, fuzzy logic, and neural networks have proven sufficiently useful to justify further investigation. This paper considers applying game theory for efficient power and bandwidth allocation in MIMO systems. Game theory is a powerful tool to solve resource allocation problems with no centralized control among users. Game theory has been applied to solve several problems in telecommunications [3] and more recently used to solve resource allocation problem. A survey of recent results is available in [4-5]. Wireless communications resources comprise power, bandwidth, spectrum, energy, physical channels (time slots, codes, sub-bands over single or multiple carriers). These resources need to be allocated among users depending on their QoS requirements. Different scenarios can be considered such as allocating different users non- overlapping fractions of time frames while keeping in view the QoS requirements of every user [6], or allocating each time slot to one user depending how much she/he pays for it [7]. Games can be arranged so every user can transmit using a common spectrum without affecting others [8], or transmitting data while keeping power constraints in mind [9-11]. TABLE I. QOS PARAMETERS FOR CLASS I AND CLASS II TRAFFIC Multiple-Input Multiple-Output (MIMO) transmission techniques can increase the efficiency in using radio band- width or spectrum. Transmission power control can also re- duce multiple access interference as well as improve energy efficiency [12]. In a multiple-link MIMO interfering system, links compete for the transmission bandwidth. In particular, signals transmitted on different links interfere with one another. The power allocation problem for a multi-link MIMO interfering system can be mapped into a game in which a player is a link and the utility function is the mutual information of the link. To solve such a game, decentralized algorithms have been proposed in the literature to find a Nash equilibrium, e.g. best response process and gradient play process in [13]. The work in [13] provides the power allocation to maximize the link mutual information. However, it does not yet map this power allocation to practical modulation methods. Neither pays any attention to QoS constraints nor optimizes the number of links in the system, i.e. link adaptation and congestion control. The objective of this paper is to develop a decentralized link adaptation algorithm for wireless network supporting different types of traffic with different QoS constraints. Table I lists an example of QoS parameter values used to assess performance. In this algorithm, we are jointly optimizing the power and the number of links in the system while meeting QoS constraints. We adaptively choose the modulation scheme based on the bandwidth and bit rate required to satisfy the service. Using the power allocation obtained from the gradient play process in [13], we propose a decentralized link adaptation algorithm that decides whether or not to support the traffic on each link, i.e., optimize the number of links. Each link decision depends on the type of traffic (class I or class II) and the QoS requirements. Traffic Type Bit Rate Tolerable BER Bandwidth (B) Class I traffic Class II traffic 700 kbps 500 kbps 10 -3 10 -5 100 kHz 100 kHz 978-1-4244-3375-9/09/$25.00 ©2009 IEEE