A Survey on Reinforcement Learning for Reconfigurable Intelligent Surfaces in Wireless Communications

Area of Science:

• Wireless Communication Engineering
• Artificial Intelligence
• Signal Processing

Background:

• Reconfigurable Intelligent Surfaces (RIS) offer power-efficient signal reflection, improving wireless communication quality.
• Machine Learning (ML), particularly Reinforcement Learning (RL), enables autonomous decision-making for complex systems.
• A gap exists in comprehensive studies on applying Deep Reinforcement Learning (DRL) to RIS technology.

Purpose of the Study:

• To provide an overview of RIS technology and its potential in wireless communications.
• To explain the operations and implementation of RL algorithms for RIS parameter optimization.
• To address the need for advanced control strategies in RIS-enabled systems.

Main Methods:

• Review of RIS principles and conventional relay technology.
• Explanation of Reinforcement Learning (RL) and Deep Reinforcement Learning (DRL) algorithms.
• Analysis of RL applications for optimizing RIS parameters.

Main Results:

• Demonstration of RL's capability to optimize RIS parameters for enhanced system performance.
• Identification of benefits such as maximized sum rate and improved energy efficiency.
• Highlighting the potential for minimizing information age in communication systems.

Conclusions:

• RL algorithms are crucial for unlocking the full potential of RIS technology.
• Optimizing RIS parameters via RL leads to significant improvements in wireless communication metrics.
• Future research should focus on addressing implementation challenges of RL for RIS.