Improved Artificial Potential Field Algorithm Assisted by Multisource Data for AUV Path Planning.

Area of Science:

• Robotics
• Ocean Engineering
• Artificial Intelligence

Background:

• Ocean exploration relies heavily on autonomous underwater vehicles (AUVs) for missions in complex environments.
• Efficient and safe path planning is crucial for AUVs to reach target points quickly and smoothly.
• Existing path-planning algorithms often combine deep reinforcement learning (DRL) with other methods for obstacle avoidance and path shortening.

Purpose of the Study:

• To address the local minimum problem in artificial potential field (APF) methods for AUV path planning.
• To enhance the efficiency and safety of AUV navigation in challenging underwater conditions.
• To develop a novel path-planning method by improving APF and integrating it with DRL.

Main Methods:

• Proposed an improved artificial potential field (IAPF) method to overcome local minima by introducing a traction force.
• Integrated the IAPF method with deep reinforcement learning (DRL) for autonomous underwater vehicle path planning.
• Optimized the reward function within the DRL algorithm and utilized generated paths to refine future navigation.

Main Results:

• The proposed IAPF-DRL method demonstrated positive effects and advantages in AUV path planning.
• Experimental comparisons showed superior performance of the developed method over existing algorithms.
• The approach proved effective in enabling AUVs to escape local minima and achieve efficient path planning.

Conclusions:

• The developed IAPF-DRL method offers an efficient and safe solution for AUV path planning.
• This approach shows significant potential for improving underwater navigation devices.
• The study contributes a valuable method for enhancing AUV mission success in complex marine environments.