The University of British Columbia

In this paper we present re-enforcement learning algorithms for adaptive modulation in flat fading channels for reconfigurable, agile wireless communications devices. We derive the dynamical stochastic control model, convexity properties of the stated optimization problem, learning based feedback control optimization and numerical simulations of the designed system. We show how this technique can be applied independently of channel model, error correction coding, and modulation constellation options. In addition, we demonstrate the algorithm's learning and tracking capabilities