A soft actor-critic approach for energy-conscious flexible job shop scheduling incoporating machine usage constraints and job release times

According to the International Energy Agency, manufacturing accounts for 30% of global energy consumption, making job shop scheduling a critical lever for reducing energy consumption. In the flexible job-shop scheduling problem (FJSP), each job consists of a sequence of operations, each of which can be processed on one of several eligible machines, and the scheduler must decide both which machine to assign to each operation and the order in which operations are processed on each machine. The Energy-Conscious Flexible Job-Shop Scheduling Problem with Release Times and Machine Usage constraints (EC-FJSP-RTMU) extends the traditional FJSP by (1) embedding a detailed multi-component energy model that quantifies the energy consumption of processing, setup, idle time, transportation, machine startup, and common facility operations, and (2) explicitly enforcing job release times and machine-usage constraints. We address this enriched scheduling problem using the Soft Actor-Critic (SAC) reinforcement learning algorithm, which learns policies to minimize total energy consumption subject to all timing and usage constraints. The SAC agent is trained in a workshop simulation featuring parallel-machine slots, shutdown policies and machine warmup/cool-down thermal dynamics. Benchmark experiments on eighteen instances show that the SAC agent consistently produces constraint-compliant feasible schedules in under 0.42 seconds.