Marine Voyage Optimization and Weather Routing with Deep Reinforcement Learning
Marine voyage optimization determines the optimal route and speed to ensure timely arrival. The problem becomes particularly complex when incorporating a dynamic environment, such as future expected weather conditions along the route and unexpected disruptions. This study explores two model-free Dee...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2025-04-01
|
| Series: | Journal of Marine Science and Engineering |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2077-1312/13/5/902 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1849711852940951552 |
|---|---|
| author | Charilaos Latinopoulos Efstathios Zavvos Dimitrios Kaklis Veerle Leemen Aristides Halatsis |
| author_facet | Charilaos Latinopoulos Efstathios Zavvos Dimitrios Kaklis Veerle Leemen Aristides Halatsis |
| author_sort | Charilaos Latinopoulos |
| collection | DOAJ |
| description | Marine voyage optimization determines the optimal route and speed to ensure timely arrival. The problem becomes particularly complex when incorporating a dynamic environment, such as future expected weather conditions along the route and unexpected disruptions. This study explores two model-free Deep Reinforcement Learning (DRL) algorithms: (i) a Double Deep Q Network (DDQN) and (ii) a Deep Deterministic Policy Gradient (DDPG). These algorithms are computationally costly, so we split optimization into an offline phase (costly pre-training for a route) and an online phase where the algorithms are fine-tuned as updated weather data become available. Fine tuning is quick enough for en-route adjustments and for updating the offline planning for different dates where the weather might be very different. The models are compared to classical and heuristic methods: the DDPG achieved a 4% lower fuel consumption than the DDQN and was only outperformed by Tabu Search by 1%. Both DRL models demonstrate high adaptability to dynamic weather updates, achieving up to 12% improvement in fuel consumption compared to the distance-based baseline model. Additionally, they are non-graph-based and self-learning, making them more straightforward to extend and integrate into future digital twin-driven autonomous solutions, compared to traditional approaches. |
| format | Article |
| id | doaj-art-d445a7315a6a4f5d8e53f741f352de60 |
| institution | DOAJ |
| issn | 2077-1312 |
| language | English |
| publishDate | 2025-04-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Journal of Marine Science and Engineering |
| spelling | doaj-art-d445a7315a6a4f5d8e53f741f352de602025-08-20T03:14:31ZengMDPI AGJournal of Marine Science and Engineering2077-13122025-04-0113590210.3390/jmse13050902Marine Voyage Optimization and Weather Routing with Deep Reinforcement LearningCharilaos Latinopoulos0Efstathios Zavvos1Dimitrios Kaklis2Veerle Leemen3Aristides Halatsis4VLTN BV, De Keyserlei 58-60 bus 19, 2018 Antwerp, BelgiumVLTN BV, De Keyserlei 58-60 bus 19, 2018 Antwerp, BelgiumDanaos Shipping Co., Ltd., 14 Akti Kondyli, 18545 Piraeus, GreeceVLTN BV, De Keyserlei 58-60 bus 19, 2018 Antwerp, BelgiumVLTN BV, De Keyserlei 58-60 bus 19, 2018 Antwerp, BelgiumMarine voyage optimization determines the optimal route and speed to ensure timely arrival. The problem becomes particularly complex when incorporating a dynamic environment, such as future expected weather conditions along the route and unexpected disruptions. This study explores two model-free Deep Reinforcement Learning (DRL) algorithms: (i) a Double Deep Q Network (DDQN) and (ii) a Deep Deterministic Policy Gradient (DDPG). These algorithms are computationally costly, so we split optimization into an offline phase (costly pre-training for a route) and an online phase where the algorithms are fine-tuned as updated weather data become available. Fine tuning is quick enough for en-route adjustments and for updating the offline planning for different dates where the weather might be very different. The models are compared to classical and heuristic methods: the DDPG achieved a 4% lower fuel consumption than the DDQN and was only outperformed by Tabu Search by 1%. Both DRL models demonstrate high adaptability to dynamic weather updates, achieving up to 12% improvement in fuel consumption compared to the distance-based baseline model. Additionally, they are non-graph-based and self-learning, making them more straightforward to extend and integrate into future digital twin-driven autonomous solutions, compared to traditional approaches.https://www.mdpi.com/2077-1312/13/5/902voyage optimizationweather routingdeep reinforcement learningmaritime energy efficiency |
| spellingShingle | Charilaos Latinopoulos Efstathios Zavvos Dimitrios Kaklis Veerle Leemen Aristides Halatsis Marine Voyage Optimization and Weather Routing with Deep Reinforcement Learning Journal of Marine Science and Engineering voyage optimization weather routing deep reinforcement learning maritime energy efficiency |
| title | Marine Voyage Optimization and Weather Routing with Deep Reinforcement Learning |
| title_full | Marine Voyage Optimization and Weather Routing with Deep Reinforcement Learning |
| title_fullStr | Marine Voyage Optimization and Weather Routing with Deep Reinforcement Learning |
| title_full_unstemmed | Marine Voyage Optimization and Weather Routing with Deep Reinforcement Learning |
| title_short | Marine Voyage Optimization and Weather Routing with Deep Reinforcement Learning |
| title_sort | marine voyage optimization and weather routing with deep reinforcement learning |
| topic | voyage optimization weather routing deep reinforcement learning maritime energy efficiency |
| url | https://www.mdpi.com/2077-1312/13/5/902 |
| work_keys_str_mv | AT charilaoslatinopoulos marinevoyageoptimizationandweatherroutingwithdeepreinforcementlearning AT efstathioszavvos marinevoyageoptimizationandweatherroutingwithdeepreinforcementlearning AT dimitrioskaklis marinevoyageoptimizationandweatherroutingwithdeepreinforcementlearning AT veerleleemen marinevoyageoptimizationandweatherroutingwithdeepreinforcementlearning AT aristideshalatsis marinevoyageoptimizationandweatherroutingwithdeepreinforcementlearning |