A learning-driven iterative greedy algorithm for distributed no idle flow shop scheduling with transportation phase

Under the background of globalization, economic prosperity promotes the development of intelligent manufacturing, which is of vital importance to the modern industrial system. To improve efficiency, enterprises have shifted from centralized production to distributed production. In view of the complex correlation between production and transportation scheduling, this paper focuses on the distributed no-idle flow-shop scheduling and transportation problem (DNIFSTP). A mathematical model is established to minimize the delivery time, where the production stage schedules the pro-cessing sequence of orders, and the transportation stage considers vehicle load constraints while planning transportation routes. A learning-driven iterative greedy algorithm (LDIG) is proposed to solve this problem. In the LDIG algorithm, the NEH (nawaz-enscoe-ham) heuristic algorithm is adopted to generate high-quality initial solutions and improve the performance of the algorithm. In the adaptive destruction and reconstruction phase, the orders to be removed are dynamically selected based on their size. In the local search phase, six local search operators are designed based on problem characteristics, and the Q-learning algorithm is applied to select the best operator for each iteration. Finally, experimental tests and comparisons with existing algorithms demonstrate that the LDIG algorithm achieves more satisfactory results.