基于深度强化学习的北极最优航线智能规划算法研究  

Study on intelligent planning algorithm for optimal Arctic shipping routes based on deep reinforcement learning

作　　者：胡浩帆 吴阿丹 韩冰[3] 朱小文 陈胜鹏 张瑞 HU Haofan;WU Adan;HAN Bing;ZHU Xiaowen;CHEN Shengpeng;ZHANG Rui(COSCO Shipping Specialized Carriers Co.,Ltd.,Guangzhou 510623,China;State Key Laboratory of Cryospheric Science and Frozen Soil Engineering/Key Laboratory of Remote Sensing of Gansu Province/Heihe Remote Sensing Experimental Research Station,Northwest Institute of Eco-Environment and Resources,Chinese Academy of Sciences,Lanzhou 730000,China;Shanghai Ship and Shipping Research Institute Co.,Ltd.,Shanghai 200135,China;School of Geodesy&Geomatics Engineering,Gansu Forestry Voctech University,Tianshui 741204,Gansu,China;School of Computer Science,China University of Geosciences(Wuhan),Wuhan 430074,China)

机构地区：[1]中远海运特种运输股份有限公司,广东广州510623 [2]中国科学院西北生态环境资源研究院冰冻圈科学与冻土工程全国重点实验室/甘肃省遥感重点实验室/黑河遥感试验研究站,甘肃兰州730000 [3]上海船舶运输科学研究所有限公司,上海200135 [4]甘肃林业职业技术大学测绘工程学院,甘肃天水741204 [5]中国地质大学(武汉)计算机学院,湖北武汉430074

出　　处：《冰川冻土》2025年第2期587-598,共12页Journal of Glaciology and Geocryology

基　　金：国家重点研发计划项目(2022YFC2807004);甘肃省重大科技项目(23ZDFA017);国家自然科学基金项目(42271492);甘肃省杰出青年基金项目(24JRRA165)资助。

摘　　要：在全球气候变化的背景下,北极海冰持续减少,为北极航道的开通提供了更为有利的条件。由于其距离和成本优势,北极东北航道吸引了航运公司的关注。然而,北极地区复杂多变的气候和海冰条件对船舶航行的安全性提出了严峻挑战,迫切需要一种智能化的路径规划方法来优化北极航道的使用。本研究基于极地运行限制风险评估系统(POLARIS),评估商船在北极航道的通航风险,并设计了结合传统A*算法和深度强化学习的最优路径规划实验。结果表明,深度强化学习模型在计算效率上显著优于传统A*算法,效率提升约50倍。总体而言,深度强化学习在搜索策略和计算效率上具有明显优势,更适用于环境复杂多变的北极航线规划,可作为实现北极航道智能规划系统的核心算法。In the context of global climate change,the persistent reduction of Arctic sea ice has created more favorable conditions for the opening of Arctic shipping routes.The Northeast Passage,in particular,has garnered the interest of shipping companies due to its potential distance and cost benefits.However,the passage presents significant navigational safety challenges stemming from its complex and variable climatic and sea ice condi⁃tions.Therefore,there is an urgent demand for an intelligent path planning method to optimize the utilization of Arctic routes.This study introduces an intelligent route planning approach that integrates the Polar Operational Limit Assessment Risk Indexing System(POLARIS)with deep reinforcement learning.POLARIS assesses nav⁃igational risks by evaluating sea ice conditions along potential routes.The researchers integrate POLARIS with traditional A*algorithms and Deep Reinforcement Learning(DRL)to enhance the path planning process.This integration is crucial as it improves the capacity to manage dynamic environments—typical of Arctic conditions—more effectively than static algorithms such as A*.Experimental results indicate that DRL significantly surpass⁃es the traditional A*algorithm in computational efficiency,achieving approximately 50 times faster processing.This efficiency is vital for real-time route planning,necessary to adapt to the rapidly changing Arctic environ⁃ment.The article explores the mechanics of DRL,elucidating its superiority in managing complex and dynamic conditions.Unlike traditional methods that struggle with large state spaces and require predefined heuristic func⁃tions,DRL employs neural networks to learn optimal strategies through trial and error,making it adept at navi⁃gating the unpredictable Arctic environment.The study highlights DRL’s capacity for learning and adaptation,providing a practical solution for real-time decision-making in shipping route planning.Moreover,the study of⁃fers a detailed explanation of the employed methodologies.It

关 键 词：北极航道 深度强化学习 航线规划 通航风险 海冰 

分 类 号：P237[天文地球—摄影测量与遥感] P941.62[天文地球—测绘科学与技术]