Autonomous port management based AGV path planning and optimization via an ensemble reinforcement learning framework
Abstract
The rapid development of shipping trade pushes automated container terminals toward the direction of intelligence, safety and efficiency. In particular, the formulation of AGV scheduling tasks and the safety and stability of transportation path is an important part of port operation and management, and it is one of the basic tasks to build an intelligent port. Existing research mainly focuses on collaborative operation between port equipment and path optimization under environmental perception, while there is relatively little research on optimization of path smoothness and safety. Therefore, we propose a path optimization model based on the artificial potential field and twin delayed deep deterministic policy gradient (APF-TD3) framework for the port environment. Firstly, we obtain the scheduling task plan of a single AGV by enumeration. Secondly, according to the artificial potential field (APF) algorithm to generate repulsion for obstacles in the harbor and attraction for container storage at the target point with the position information of the AGV as the input data of the reinforcement learning algorithm is inputted into the twin delayed deep deterministic policy gradient algorithm (TD3). Then TD3 selects the optimal action strategy for the AGV according to the input AGV state information and the designed reward mechanism as well as executes the action. Through repeated execution, the optimal action for the next step is selected at each point to generate a path with start and end points. We validate the model by simulating the scale of containerized cargo in the port i.e. small scale, medium scale and large scale scenes. The experimental results show that the method has the shortest path length of 27.519 m, 270.847 m, and 496.389 m compared to artificial potential field and deep deterministic policy gradient (APF-DDPG), APF, and rapidly-exploring random tree (RRT) algorithms, which also have significant advantages in terms of path security and path smoothness. This framework could respond to the scheduling and transportation tasks of single AGV in different environmental layouts and guarantee the smoothness and safety of the path based on the optimization of the path, which promotes the efficient operation and management of ports.
Citations
-
4 5
CrossRef
-
0
Web of Science
-
4 0
Scopus
Authors (6)
Cite as
Full text
full text is not available in portal
Keywords
Details
- Category:
- Articles
- Type:
- artykuły w czasopismach
- Published in:
-
OCEAN & COASTAL MANAGEMENT
no. 251,
ISSN: 0964-5691 - Language:
- English
- Publication year:
- 2024
- Bibliographic description:
- Chen X., Liu S., Zhao J., Wu H., Xian J., Montewka J.: Autonomous port management based AGV path planning and optimization via an ensemble reinforcement learning framework// OCEAN & COASTAL MANAGEMENT -Vol. 251, (2024), s.107087-
- DOI:
- Digital Object Identifier (open in new tab) 10.1016/j.ocecoaman.2024.107087
- Sources of funding:
-
- Free publication
- Verified by:
- Gdańsk University of Technology
seen 77 times