D3QN-Based Online Sensor Scheduling for Remote State Estimation

This paper addresses the challenge of remote state estimation in wireless sensor networks (WSNs), where the number of available wireless channels is often limited. We propose a novel sensor scheduling strategy using an improved Dueling Double Deep Q-Network (D3QN) algorithm to optimize transmission scheduling for enhanced estimation efficiency at a central gateway. By formulating the sensor scheduling issue as a Markov Decision Process (MDP), our improved D3QN algorithm, which enhances the traditional Deep Q-Network (DQN) framework for better decision-making efficiency, is applied to solve it. Comparative analysis with existing reinforcement learning (RL) algorithms demonstrates the superiority of our approach, as evidenced by simulation results showcasing improvements in estimation accuracy and network efficiency.