Multi-objective optimization of the textile manufacturing process using deep-Q-network based multi-agent reinforcement learning

Abstract

Multi-objective optimization, such as quality, productivity, and cost, of the textile manufacturing process is increasingly challenging because of the growing complexity involved in the development of textile industry in the upcoming big data era. It is hard for traditional methods to deal with high-dimension decision space in this issue, and prior experts’ knowledge is required as well as human intervention. This paper proposed a novel framework that transformed the textile process optimization problem into a stochastic game, and introduced deep Q-networks algorithm instead of current methods to approach it in a multi-agent system. The developed multi-agent reinforcement learning system applied a utilitarian selection mechanism to maximize the sum of all agents’ rewards (obeying the increasing ε-greedy policy) in each state, to avoid the interruption of multiple equilibria and achieve the correlated equilibrium optimal solutions of the textile process. The case study result reflects that the proposed MARL system can achieve the optimal solutions for the textile ozonation process, and it performs better than the traditional approaches.