Bidirectional Search-Based Optimization of AGV Transportation Paths in Automated Container Terminals

Optimizing the routing and coordination of Automated Guided Vehicles (AGVs) is essential for maximizing productivity at automated container terminals. Existing methods based on reinforcement learning typically encounter difficulties with high computational complexity, slow convergence rates, and no guarantee of achieving a deterministic optimum, which restricts their use in actual time-critical environments. In response to this issue, this paper presents a novel optimization approach to path planning for AGVs used for transporting containers in port container terminals, which is based on a bidirectional search process. Initially, the port environment is represented as a graph where spatial and temporal restrictions are imposed upon the graph. The formulation of an AGV transportation path planning multi-objective optimization problem is described where the objectives are to minimize travel distance, waiting time, and conflict costs. The main contribution of this work is the development of a bidirectional search algorithm to identify optimal transportation paths with fewer than half as many attempts at finding optimal transportation paths by searching both the initial node and the end node at the same time. The use of a conflict aware path refinement method employing time window reservation and node occupancy restrictions ensures that AGVs operate safely by avoiding collisions during navigation. Additionally, to manage the uncertainty associated with real-time operations, a multi-AGV coordination strategy based on priority scheduling and dynamic re-routing has been developed. The proposed framework achieves faster convergence, guaranteed shortest path generation, and improved scalability compared to conventional methods. Experimental evaluation demonstrates significant reductions in terms of Path Length of 1004m, Travel Time of 405s, Waiting Time of 70s respectively.