A IHQL-LDTPSA to Solve Distributed Heterogeneous Permutation Flow Shop Problem with Lot Streaming

This paper addresses the distributed heterogeneous permutation flow shop scheduling problem with lot streaming (DHPFSP-LS). The problem involves determining sub-batch quantity, batch size, job allocation, and processing sequence across factories. To tackle its large solution space, we decompose it into a lot streaming subproblem and a scheduling subproblem, proposing a two-stage algorithm, IHQL-LDTPSA. In Stage 1, Improved Hybrid Q-learning (IHQL) is used to solve for the optimal batching strategy for each job in the factory. Stage 2 employs the LDTPSA algorithm for job allocation and sequencing, integrating Stage 1's batching results during decoding. Experiments on 10 benchmarks compare IHQL-LDTPSA against four algorithms (traditional/improved heuristics and novel metaheuristics), demonstrating its superior effectiveness and robustness in solving DHPFSP-LS.