Adaptive stochastic iterative rate selection for wireless channels

Abstract

A stochastic algorithm for channel adaptive rate selection (modulation and coding scheme or MCS) is proposed. The algorithm learns the optimal policy by iteratively augmenting a rate selection probability vector. While simple to implement, this technique requires no explicit channel estimation phase. The single bit ACK/NACK signal feedback from the data link layer is used as the input to the stochastic algorithm. As shown in the convergence theorems, the algorithm achieves the optimal rate in "static" channels. A time varying channel is seen as a "quasi-static" channel, and adaptively converged to optimal rates as channel state changes.