Self-adjusting multi-objective scheduling based on Monte Carlo Tree Search for matrix production assembly systems

Abstract As a response to the low utilization of production lines in the context of an increasing number of product variants with significant differences in cycle times, line-less modular assembly concepts known as matrix production have evolved. Whilst the many degrees of freedom a matrix production offers provide different ways to react to disturbances and balance utilization, it also increases the complexity for scheduling exponentially. For real-time scheduling high solution quality and high efficiency are needed. This paper contributes a multi-objective scheduling approach based on Monte-Carlo Tree Search that self-adjusts to the scheduling problem to improve solution quality and execution time.