Generalized queue-aware radio resource management: A dynamic programming approach

A general model for queue-aware radio resource management (RRM) and scheduling in a slotted-time communication system employing a finite data buffer is studied under quasi-static fading channel condition. Queue-awareness therefore dominates the RRM and scheduling problem. According to the buffer queue state in each time slot, the system adaptively selects transmission mode or schedules the available radio resources. The resulting queue-aware sub-fading-block scheduling improves resource utilization when long coherence time is experienced, without resorting to physical layer techniques such as opportunistic beamforming which require more sophisticated RF hardware at the transmitter. The scheduling problem is then formulated as an infinite horizon Markov decision process (MDP) which is irreducible under arbitrary resource allocation policy. The dynamic programming (DP) approach is employed to solve the MDP problem. Two DP algorithms, namely relative value iteration (RVI) and policy iteration with relative policy evaluation (PIRE) are investigated. Numerical results for single user adaptive modulation show that both algorithms effectively approach the optimal scheduling policy obtained by exhaustive search in the average queue length minimization and the average packet blocking plus packet retransmission minimization problems. A significant performance improvement over random RRM policy and equal buffer partitioning queue-aware policy is observed.