A Reinforcement Learning Approach for Mobile Beamforming
Anastasios Dimas, Konstantinos Diamantaras, Athina P. Petropulu · 2019
We consider the problem of enhancing Quality-of-Service in mobile relay beamforming networks in the presence of a dynamic channel, and propose a novel reinforcement learning approach for optimally controlling relay motion. The network includes a fixed source and destination, and a number of cooperative mobile relays. We assume a time slotted system, where the relays update their positions before the beginning of each time slot. We propose a 2-stage approach for optimally specifying relay positions and beamforming weights, such that the expected signal-to-interference plus noise ratio (SINR) at the destination is maximized, based on causal Channel State Information and under a total relay transmission power budget. In each time slot, the relays optimally beamform to the destination, and then, using a novel reinforcement learning approach, the relays decide the positions to move to, so that they are optimally positioned in the next time slot; the relays determine the latter positions based on measurements of channel magnitudes obtained at places they have visited up to the present time.