Deep Learning Based Integrated Information and Energy Relaying in RF Powered Communication

Energy transfer (ET) in RF powered harvesting as well as information transfer (IT) in end-to-end communication is obstructed by range of transmission in the network under consideration. This can be resolved by employing a cooperative relay in both ET and IT operations. However, involving the composite operations in a practically unknown environment require a learning based algorithms to obtain an optimal policy for efficient energy management and data communication together. To confront it, here, we propose a deep learning algorithm based on deep deterministic policy gradient (DDPG), providing continuous course of actions under optimal online policy for integrated information and energy relaying (i2ER) network. In the designed nonconvex problem, the long-term average net bit rate of the end-to-end communication is maximized in four phases of operations under the given constraints on the harvested energy at relay and source nodes. Via extensive simulations, various insights are obtained on the performance of the proposed algorithm in different used modulation for transmission and learning rate while and after learning. Lastly, the achieved bit rate in the i2ER network is compared with the performance of a greedy benchmark scheme and get an improvement upto 62%.