Applications of RL for Continuous Problems in RIS-Assisted Communication Systems

DRL has evolved as an effective approach to optimize the performance of various wireless communication systems. DRL is considered a potential candidate to optimize the RIS phase shifts without the need for tuned mathematical relaxations, as in alternating optimization (AO) techniques, or offline training with a labeled dataset, as in supervised machine learning based techniques. To this end, DRL can optimize the RIS phase shifts by combating the implementation limits of the AO techniques. In this chapter, the applications of DDPG to optimize continuous RIS-assisted wireless communications problems are addressed. DDPG is widely used for continuous and large-scale action and state spaces. This chapter explains how it can be leveraged to optimize the continuous transmit power, beamformers, and RIS phase shifts of various RIS-assisted wireless systems.