DISCRETE EVENT SIMULATION WITH REINFORCEMENT LEARNING FOR ADVANCED MANUFACTURING DECISION-MAKING

In recent years, the integration of Artificial Intelligence (AI) to support decision-making in the manufacturing sector has grown significantly. Discrete Event Simulation (DES) offers a powerful method for training AI agents to make rapid decisions in manufacturing process flows. This study explores advancements in combining DES with Reinforcement Learning (RL), addressing data scarcity and integration challenges. A literature review highlights the increasing adoption of RL, particularly Deep Q-Network (DQN) and Proximal Policy Optimization (PPO), for complex tasks like job shop scheduling. The research introduces an AGV-based job shop scheduling use case modelled in Siemens Plant Simulation to test various RL algorithms under different complexities. An integration pipeline using Python and MQTT ensures seamless AI-DES communication, enabling dynamic decision-making. Future work will include developing DQN and PPO algorithms and expand this framework to incorporate real-world factors, refining AI's role in adaptive and efficient manufacturing solutions.