Abstract: To improve the efficiency of ship entry into the lock chambers and intelligently guide the ship into the ship lift chambers safely and efficiently, the Markov decision-making model is used to model the ship entry process based on the reinforcement learning method. The model training results are compared to identify the optimal solution by setting different reward functions and time steps. Subsequently, an intelligent navigation aid system is obtained by software development. Finally, a sea trail is conducted by using a typical ship passing through the lock chamber. The results show that the optimal maneuvering strategy not only reduces the number of operations but also enables faster arrival at the destination. Furthermore, the theoretical entry time planned by the navigation aid system is 124 seconds shorter than the actual entry time, corresponding to a 34.4% reduction in entry time. The research can provide references for the intelligent guidance of ship entry into the lock chambers.

Key words: ship entry into lock chamber, reinforcement learning, intelligent guidance

摘要： 为智能引导船舶安全高效进入升船机船厢,该文基于强化学习方法,采用马尔可夫决策模型对船舶进厢航行过程建模,通过设置不同的奖惩函数和时间步长,对比模型训练的结果并得出最优方案,然后进行软件开发得到智能助航系统,最后选取典型过厢船舶开展实船试验。结果表明：选择最优的操纵策略不仅可以减少操作次数,而且能够更快达到终点,助航系统规划的理论进厢时间比实际进厢时间缩短了34.4%。该研究成果可为船舶进厢智能引导提供一定参考。

关键词: 船舶进厢, 强化学习, 智能引导