Socially Conscious Navigation of Mobile Robots Based on Deep Reinforcement Learning

In a human–robot coexisting environment, mobile robots need to navigate between humans and other obstacles in a way that conforms to social norms. The high dynamism and uncertainty of human–robot coexisting environments pose new challenges to robot motion planning. To solve this problem, this study proposes a navigation strategy based on deep reinforcement learning (DRL) for mobile robots in crowded and complex environments. In this work, an EmoHarmony social distance mapping function (ESDMF) is proposed to describe the relationship between psychological comfort and distance in real life. Thus, different social distances will generate different punishment intensities accordingly. To handle obstacles other than humans, we also added the modeling of obstacles. In the experiment, the performance of the algorithm is evaluated by the success rate, collision rate, timeout rate, navigation time, trajectory length, and danger frequency. The experimental results show that the performance of this algorithm is better than state-of-the-art algorithms in the human–robot coexisting environment.