Deep Reinforcement Learning-Based Joint Phase Shift and Reflection Coefficient Optimization for RIS-Assisted Broadcast-Passive IoT Symbiotic Systems

Broadcast-passive Internet of Things (IoT) symbiotic systems exploit ambient broadcast signals to simultaneously power passive devices and support their information transmission through backscatter communication. However, these systems still suffer from severe double-hop fading, limited coverage, and complex interference coordination between broadcast and passive IoT communications. This paper proposes a reconfigurable intelligent surface (RIS)-assisted system architecture and formulates a joint optimization problem of RIS phase shifts and passive device reflection coefficient to maximize passive IoT throughput, subject to broadcast receiver quality-of-service (QoS) and passive device energy harvesting constraints. Given the non-convex and coupled nature of the formulated problem, we develop a deep reinforcement learning (DRL) framework based on proximal policy optimization (PPO), incorporating a power-domain state space representation and a piecewise reward function to balance throughput maximization with multi-constraint satisfaction. Simulation results demonstrate that the proposed scheme achieves significant throughput gains with increasing RIS elements, while the PPO agent learns to push the operating point toward active constraint boundaries, achieving an effective balance among throughput maximization, broadcast QoS protection, and energy sustainability.