An autonomous mobile robot path planning method fusing dynamic obstacle detection and tracking

Abstract

This invention discloses a path planning method for autonomous mobile robots that integrates dynamic obstacle detection and tracking, belonging to the field of computer path planning algorithm technology. The method includes collecting environmental perception data, identifying static obstacle information and the position and velocity information of dynamic obstacles, using a Kalman filter algorithm to generate predicted position points of dynamic obstacles based on their position and velocity information, constructing a cost map function that integrates static obstacle information and predicted position points of dynamic obstacles, designing a reward function, and embedding the reward function and cost map function into a reinforcement learning framework. The reinforcement learning framework is trained to convergence in a virtual environment to obtain an initial path planning model. The initial path planning model is deployed in a real-world application environment and fine-tuned using data collected from the real environment to obtain the final path planning model.

Images