The Intelligent Balancing Design of Air-Bearing Table's Mass Properties Based on Deep Reinforcement Learning

This paper investigates the intelligent mass-property balancing problem of the air-bearing table in the aerospace field. By establishing the air-bearing table's coordinate system and representing the design variables, a mass properties balancing model is first developed. Then, based on the design of the Genetic Algorithm (GA) and Deep Q-Network (DQN), an autonomous learning genetic algorithm (GA-DRL) based on deep reinforcement learning is proposed to address the issues of low balancing efficiency and poor precision in traditional genetic algorithms. Simulation results show that, compared to the conventional genetic algorithm, GA-DRL exhibits better convergence efficiency and optimization precision, and can theoretically achieve efficient and high-precision intelligent balancing of the air-bearing table's mass properties.