As a pivotal research topic in the scheduling domain, the Flexible Job Shop Scheduling Problem (FJSP) has long attracted significant academic attention. However, research on FJSP involving transportation equipment capacity constraints remains relatively scarce. To address this gap, this paper proposes a Two-Layer Reinforcement Learning-Integrated Genetic Algorithm (TRLIGA) that simultaneously minimizes makespan and reduces energy consumption. The algorithmic framework integrates pre-scheduling with dynamic rescheduling constrained by transportation equipment capacity. This paper constructs a four-layer encoding structure and proposes a corresponding initialization method. During iteration, elite solutions are archived into an external non-dominated set for co-evolution, while a two-layer reinforcement learning framework adaptively tunes key parameters. A dedicated repair mechanism handles infeasible solutions. Experimental results demonstrate the effectiveness and superiority of the proposed methodology across comprehensive test scenarios.
Wang et al. (Fri,) studied this question.