Multi-AGV Motion Planning Using Greedy Search Algorithms and Learned Heuristics

For productive operations involving multiple Automated Guided Vehicles (AGVs), it is essential to plan motions quickly that minimize the total sum of costs required for each AGV to perform transport tasks (Sum-of-Costs). However, calculating motions that achieve a superior Sum-of-Costs within a short period is challenging. This study introduces a strategy to accelerate motion planning for multiple AGVs utilizing Conflict-based Search (CBS). It incorporates two methodologies: (a) employing greedy search algorithms, and (b) integrating learned heuristics into the A* search algorithm. After implementing and simulating each method, the proposed integrated methodology showed significant efficiency for 15 AGVs. It achieved a 96% reduction in computation time and limited the deterioration of the Sum-of-Costs to 8%, compared to the CBS-based method. Furthermore, even with a strict CPU time limit of 60 seconds, the proposed method successfully planned motions for 20 AGVs, tasks that the baseline method failed to accomplish.