Two-stage unmanned vessels co-scheduling method based on the reinforcement learning algorithm
After the pandemic,there will be economic growth,and the volume of container shipping will significantly increase,but the existing port and land-side facilities are increasingly inadequate to handle the container capacity.Therefore,to improve the efficiency of container transfer between ports,this paper proposes a two-stage scheduling method to drive unmanned vessels for container transportation in the port waterside.In the first stage,taking into ac-count the recharging mode of unmanned vessels,time windows,coordinated berths,and other constraints,a task planning model is established with the goal of minimizing operation delays,reducing the penalties caused by unmanned vessel delays.Another task planning model is then constructed with the objective of minimizing execution costs to obtain a collision-free transport plan.In the second stage,an unmanned vessel path tracking control method is proposed.Consid-ering rudder angle saturation constraints,a model predictive controller is designed to smoothly achieve unmanned vessel path control.Finally,an MARL algorithm is introduced,which,in conjunction with historical training data,quickly obtains optimal solutions to problem-solving.Simulation results show that the proposed scheduling method can obtain low-cost unmanned ves-sel transport plans,and the introduced algorithm still demonstrates good performance in solving large-scale problems.