AUTONOMOUS ROBOTICS FOR SMART MANUFACTURING USING DEEP REINFORCEMENT LEARNING

Abstract

The intensive development of intelligent production requires autonomous robots that can make decisions in real time and adapt to changes, as well as operate robustly in a changing production setting. Deep Reinforcement Learning (DRL) has become a new paradigm of facilitating intelligent autonomy in industrial robotics by empowering agents to learn optimal policies of control by engaging with a complex manufacturing system. This paper explores the problems of applying DRL-controlled autonomous robotics to smart factories, paid attention to adaptive tasks, optimization of processes, and decision-making under uncertainty. The framework proposed integrates the state representation based on the perception, a reward-based learning, and policy optimization, such that robots can carry out scheduling, manipulation, navigation, and quality-conscious tasks autonomously, without any reference to some predefined rule-based control. In order to deal with the issues of real-world deployment, the framework adopts simulation-to-real transfer strategies, training with digital twins and safety-aware reward shaping in order to achieve stable convergence and operational stability. Empirical tests in typical manufacturing conditions show that DRL-based robotic agents are capable of massive performance gains in terms of task efficiency, flexibility and resource consumption in contrast to traditional heuristic and model based controllers. The outcomes also suggest the increased resilience to the environmental fluctuations, machine disruptions, and reconfigurations of production. One enabling technology in this study is the use of DRL-based autonomous robotics, which can help ensure the next-generation smart factory, enabling flexible production, minimizing human intervention, and maximizing performance continuously. The results are very insightful into scalability deployment strategies and lead to the direction of completely autonomous, intelligent, and resilient manufacturing systems in compliance with the Industry 4.0 and Industry 5.0 visions.