A Cross-Layer Approach to Heterogeneous Multi-User Diversity with Link Adaptation

Multi-user diversity (MUD) scheduling is based on the idea that access to the channel should be given to users with better channel state. In a system with many users whose channels vary independently, such strategy maximizes the probability that the channel of the scheduled user is near its peak. Consequently, the total ergodic capacity of the system is maximized. However, from a network point of view, performance is usually evaluated in terms of average throughput and average delay rather than ergodic capacity and fairness. Furthermore, information-theoretic rates are not achievable at the PHY layer and more practical solutions must be considered. In this paper the MUD problem is approached from a MAC layer perspective and the throughput maximal MUD scheduling policy, i.e., the policy that maximizes the total average throughput is investigated. We also study the average throughput region and show the suitability of optimal information-theoretic approaches in practical environments. The average throughput region is also contemplated in wireless networks where Heterogeneous multi-user diversity (HMUD) can be exploited. Finally, the average throughput region is related to the stability region and some average delay bounds for different ARQ mechanisms are provided.