Throughput maximization for fading channels and bursty traffic sources in link layer

We consider the problem of adaptive power allocation and modulation mode selection over fading channels and finite buffer size in the link layer. The objective is to maximize the long-term system throughput under the constraint of some average transmission power and BER by varying the transmission power and modulation mode according to the buffer and channel conditions. In this model, we assume that the knowledge of the buffer occupancy and the fading state is available at the transmitter. In link layer, we use ARQ technology to alleviate transmission error and the process of ARQ continues until the packet is successfully received. So maximizing the system throughput is equivalent to minimizing packet loss due to buffer overflow regardless of transmission error. We formulate this optimization problem as a MDP and use dynamic programming techniques to obtain the solution. We also present a simplified scheduling scheme for package transmission and our results show that the package loss of the simplified scheme is very close to that achieved by the optimal policy.