Reinforcement Learning Based Planning and Control of Modular Body-in-White Production Systems, Considering the Plant Intralogistics (el. knyga) (skaityta knyga) | knygos.lt

Aprašymas

The trend toward shorter product life cycles, customised products, and electromobility means that greater flexibility in automotive body-in-white (BIW) production is required. Modular production, which uses workstations connected by automated guided vehicles, allows for more flexibility, but complicates the scheduling. Therefore, new and innovative scheduling approaches must be developed. This dissertation investigates the potential of reinforcement learning (RL) for the scheduling of modular BIW production systems and compares it with mixed integer programming (MIP) and priority rule heuristics. The RL approach employs an agent that learns to optimise its decisions through interactions with a discrete-event simulation model. In twelve experiments, the approaches are compared with regard to calculation time, makespan, and delays due to breakdowns. In these experiments, the model parameters are varied to evaluate the suitability of RL for the scheduling of modular BIW production systems. Based on the experiment results, concrete recommendations are derived for the scheduling of modular BIW production systems with RL.