MANAGEMENT OF AGV FLEET IN AUTOMATED CONTAINER TERMINAL

Abstract

In an automated container terminal, the handling of containers is accomplished through the coordination of multiple material handling equipment. The transportation of containers between the berth and the storage yard, traditionally done by prime movers, is now carried out by the Electric Automated Guided Vehicles (AGV). To accomplish the transportation of containers, these AGVs have to be routed carefully to avoid collision or deadlocks, and at the same time sent for recharging to maintain sufficient battery charge. An efficient AGV fleet management relies on an optimal charging facility and operational planning, which includes routing and scheduling of the recharging operations. There are however practical challenges that arise mainly due to the limited charging stations, short planning period and the sheer scale of the system. In this thesis, the battery management at the facility and operational planning levels are studied together with a fast dynamic routing for a large AGV fleet, with the aim of optimising the terminal operations. Firstly, the planning of charging facility is studied with the aid of a simulation model under different configurations and scenarios. Then, the recharging scheduling for the AGV fleet under a rolling horizon planning is tackled using the proposed multi-agent reinforcement learning approach. This problem is extended to the dynamic planning context, where job assignment is carried out dynamically. Finally, a fast deadlock-free dynamic routing is examined under capacitated network, particularly for systems with a large AGV fleet. The research presented highlights the success of the proposed approach and the benefits of efficient AGV management on the terminal performance.