Automated scenario analysis of reinforcement learning controlled line-less assembly systems
Amon Göppert, Jonas Rachner, Robert Schmitt
Abstract
The concept of line-less matrix-structured assembly systems with dynamical job routes is an alternative concept to line configurations with a promising performance for complex production scenarios. Due to the dynamical environment, fast online decisions for controlling the job routes are required. A reinforcement-learning algorithm based on Monte Carlo tree search for time-efficient online decision-making is proposed. For the performance comparison of matrix assembly systems with line configurations, a seamlessly automated scenario analysis tool generates, simulates and processes the results for a large set of assembly scenarios. The results show a significant improvement in the utilization of work stations especially for a high number of variants.