Deep Reinforcement Learning for Dynamic Berth Allocation with Random Ship Arrivals

With the growth of global trade volume and advancements in information technology, the intelligent transformation of port systems has become a trend in the transportation industry. To address the unpredictable factors of dynamic ship arrivals, this paper utilizes a deep reinforcement learning (DRL) approach to solve the dynamic berth allocation problem (DBAP). The scheduling model aims to minimize the weighted waiting time of ships. The state space is constructed by extracting information about berths and ships in a dynamic environment. This paper proposes ship operation task buffers to map the flexible action space, and the optimization objective is decomposed by decision intervals to design the reward function. A double dueling deep Q-network (D3QN) algorithm, which integrates the advantages of DDQN and Dueling DQN, is used to solve the scheduling scheme. Finally, the network is trained with data to enable the agent to choose the optimal action based on the current state of the harbor berth system. The experimental results show that this method can effectively reduce ship waiting times in a dynamic environment, proving to be more advantageous than methods based on traditional dispatching rules.