Reinforcement Learning for Energy Harvesting in Spatial Modulated MIMO Systems

To enhance the lifetime of a power constrained relay network, Energy Harvesting (EH) is a promising solution. This paper introduces Reinforcement Learning (RL) algorithm for high data rate bidirectional relay network with the merits of Spatial Modulation (SM) and Decode and Forward (DF) relay protocol. As the relay network is energy constrained, the required power for forwarding the data is harvested from the received radio frequency signals. Based on real-time scenarios, it is assumed that relay has knowledge only about the past and the current state of the EH process. The proposed system is modelled as Markov Decision Process (MDP) and power allocation policy is formulated using RL algorithm. This policy maximizes the overall throughput of the system and reduces the outage. Further, the concept of linear function approximation is proposed to handle the real-time scenarios.