Research on Navigation and Control Technologies for Mobile Robots in Complex Environments Based on Deep Learning

To address the issue of poor navigation performance of mobile robots in complex environments, a navigation control method based on deep learning is proposed. This method adopts a hierarchical model framework, which consists of two parts: a low-level part and a high-level part. The low-level part includes obstacle avoidance and target-driven control models, responsible for implementing obstacle avoidance and target approaching behaviors, respectively. The behavior selection model in the high-level part can automatically learn stable and reliable behavior selection strategies, reducing the dependency on manually designed control rules. Furthermore, through optimized training of the obstacle avoidance control model, the learned obstacle avoidance strategies become more adaptable to navigation requirements in complex environments. Experimental results demonstrate that this method can significantly improve navigation performance in complex environments and possess good generalization ability. Additionally, preliminary navigation tests in real environments have also verified the practical application prospects of this method.