A Hierarchical Path Planning Approach with Multi-SARSA Based on Topological Map.

Area of Science:

• Robotics
• Artificial Intelligence
• Computational Geometry

Background:

• Path planning is crucial for autonomous systems.
• Existing methods may struggle with complex environments or convergence speed.

Purpose of the Study:

• To propose a novel, efficient path planning algorithm.
• To leverage topological maps and Reinforcement Learning for optimal path discovery.

Main Methods:

• A two-level approach combining region dynamic growth for topological area generation and a two-layer Multi-SARSA Reinforcement Learning method.
• Initialization of Q-tables using artificial potential fields and topological map information for accelerated learning.

Main Results:

• The proposed algorithm successfully finds optimal paths.
• Simulations show a reduction in path length compared to existing methods.
• The combined approach enhances convergence speed and stability.

Conclusions:

• The novel path planning algorithm effectively utilizes Reinforcement Learning and topological maps.
• The method demonstrates superior performance in finding shorter, optimal paths.
• This approach offers a promising solution for complex robotic navigation challenges.