Digital-twin-driven green human-robot collaborative flexible job shop scheduling

Industry 4.0 demands manufacturing that is flexible, green, and resilient. We introduce the DT-MOD-HRC-FJSP: a Digital-Twin-driven, multi-objective, dynamic Human–Robot Collaborative Flexible Job-Shop Scheduling Problem that simultaneously minimizes makespan, energy, and tardiness while coping with disruptions like machine failures and rush orders. A real-time Digital Twin fuses physical floor data with a cyber scheduler. Operations can be executed by machines, humans, or human-robot teams, accounting for worker fatigue and skill levels. To solve this NP-hard challenge we propose Adaptive Memetic Algorithm with Reinforcement Learning (AMARL), an Adaptive Memetic Algorithm with Reinforcement Learning. Proactive scheduling uses AMARL to evolve robust baselines: a genetic algorithm explores globally while a variable neighborhood search exploits locally; a Q-learning agent adaptively selects operators to boost both speed and quality. Reactive scheduling, triggered by live DT data, applies right-shift and partial rescheduling to restore feasibility quickly while preserving stability. New benchmark instances show the DT-AMARL framework surpasses state-of-the-art methods, improving Hypervolume by 18.5 % on average and demonstrating strong resilience under dynamic conditions.