Visual navigation technology for autonomous driving robots based on strategic gradient-REINFORCE algorithm

Abstract

Currently, autonomous driving robots face challenges of insufficient environmental perception and decision delay in visual navigation. To optimize the visual navigation performance of an autonomous driving robot under intricate working conditions, the paper optimizes the REINFORCE algorithm via integrating strategic gradients. A more efficient and precise visual navigation model built on the improved REINFORCE is designed. Experimental results demonstrate that the research algorithm’s accuracy reaches 95.7%, with 91.2% recall, superior to comparative algorithms. The improved algorithm has 92.5% navigation success rate in simulated environments, 15.8% surpassed than traditional methods. In real-world testing, the robot’s navigation decision time is reduced by 20.3%. Research algorithms can strengthen the visual navigation performance and offer a novel navigation strategy for autonomous driving robots, thereby promoting the application of autonomous driving technology.