An adaptive energy-efficient link layer protocol using stochastic learning control

We present a computationally simple stochastic learning control framework for an adaptive energy efficient link layer protocol. A stochastic iterative technique is discussed that can produce soft channel state predictions and track slow/rapidly varying bursty, finite state wireless channels. No a priori knowledge about the state transition probabilities is needed for this. Theoretical convergence of the proposed technique is shown. The proposed link layer protocol utilizes the channel state predictions from the stochastic learning algorithm while computing energy efficient transmission policies. This entire process is performed on-line with no pilot (training) symbols, etc., thereby improving the throughput and avoiding energy wastage due to pilot symbols. Simulation results show that up to 50% energy savings can be obtained for some channels when compared with a popular link layer protocol. Energy and delay can be traded-off efficiently using the proposed method.