Intelligent robot path planning based on multimodal deep learning algorithm

In order to overcome the problems of poor performance, low obstacle avoidance success rate, and long time in traditional intelligent robot path planning methods, this paper proposes an intelligent robot path planning method based on multimodal deep learning algorithm. By using depth cameras, LiDAR, and IMU to collect multimodal data, and utilising multimodal deep learning algorithms to determine the characteristics of the collected signal and image data, intelligent robot obstacle target detection can be achieved. By combining the gravity and repulsion functions of the artificial potential field method to improve the RRT Connect algorithm, a smooth path trajectory is obtained by smoothing the inflection point of the path through a cubic B-spline curve. The experimental results show that the proposed method has high smoothness and low complexity of intelligent robot paths, shorter paths, an average obstacle avoidance success rate of 97.08%, and an average path planning time of 74.25 ms.