Enhanced Evolutionary Multi-Objective Deep Reinforcement Learning for Reliable and Efficient Wireless Rechargeable Sensor Networks

The paper presents an enhanced evolutionary multi-objective deep reinforcement learning algorithm designed to optimize wireless rechargeable sensor networks (WRSNs) by balancing node survival rates and energy efficiency. This approach integrates a long short-term memory (LSTM) policy network for temporal pattern recognition, a multilayer perceptron for state prediction, and a time-varying Pareto evaluation method, addressing the NP-hard optimization problem in dynamic conditions. Simulation results indicate a 25% faster convergence compared to traditional methods, making it a significant advancement for practitioners working on energy-efficient sensor network solutions.