Efficient Hardware Implementation of Multi-Armed Bandit Algorithms for RIS-Aided Wireless Networks

The potential of Reconfigurable Intelligent Surfaces (RIS) to support next-generation wireless communication systems has recently garnered significant attention. This research ad-dresses the challenge of optimizing RIS-aided wireless networks, where RIS technology enhances signal control and coverage. However, in complex scenarios with multiple RIS units and receivers, selecting the optimal RIS blocks to maximize SNR becomes computationally intensive. Traditional methods relying on complete channel state information face high latency and scalability issues in dynamic environments. To overcome this, we propose a synthesizable Multi-Armed Bandit (MAB) algorithm, specifically the Upper Confidence Bound (UCB) algorithm, for RIS selection in real-time deployments on the PolarFire SoC. The focus is on efficient decision-making within constrained computational environments, leveraging a hardware-software co-design approach to optimize resource utilization. The implementation efficiently maps the algorithm to PolarFire SoC resources, including logic, memory, and I/O components, ensuring real-time processing performance in smart wireless environments.