.png){kind=icon}
.png){kind=icon}
Return
Research on Geomagnetic Perceiving Navigation Method Based on Deep Reinforcement Learning
.png){kind=icon}
PRE
AI
DOI：10.1109/JOE.2025.3596672.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Autonomous underwater vehicle (AUV) navigation poses significant challenges in unknown environments lacking prior information. In this article, a deep Q-network-based geomagnetic sensing navigation method is proposed to solve the navigation problem without prior maps. By constructing a deep Q-network model, the method utilizes the powerful representation capability of deep learning to handle complex geomagnetic sensing data. Meanwhile, an action selection strategy combining heuristic and greedy search is introduced to balance exploration and exploitation and enhance the autonomous navigation capability of underwater vehicles in unknown environments. The strategy dynamically adjusts the exploration probability based on the distance to the target, ensuring effective exploration at an early stage and efficient utilization of the learned knowledge when approaching the target. Moreover, the AUV explores the environment using local geomagnetic data and trains a regression model to predict the global geomagnetic map. Simulation results show that the method is significantly effective in reducing path lengths, improving exploration efficiency, and enhancing geomagnetic map accuracy. The results show that the method significantly improves the robot’s navigation performance in unknown environments and provides a new way to construct geomagnetic maps.
Keywords:
Autonomous underwater vehicle (AUV)
autonomous navigation
construction of geomagnetic map
geomagnetic navigation
path optimization
Journal
IF:
5.3
Papers:
2.6K
Citations:
7.4K