Learning Mobile Robot Navigation in the Dense Crowd with Deep Reinforcement Learning

In recent years, the growing demand for more intelligent service robots is pushing the development of mobile robot navigation algorithms. Moving in a dense crowd safely and efficiently is an important yet challenging task for the operation of the mobile robots. Reinforcement learning (RL) approaches have shown superior ability in solving sequential decision making problems, and recent work has explored its potential to learn navigation polices in a socially compliant manner. In our project, we propose to apply an RL framework to develop a smart system for human-aware navigation. We propose to use value function approximation based methods to learn human-aware navigation policies and implement the reward shaping, hindsight experience replay, and curriculum learning techniques for our task. The effectiveness of our methods is validated in a simulated human-aware navigation environment. The video demonstration is at \url{https://mycuhk-my.sharepoint.com/:v:/g/personal/1155131468_link_cuhk_edu_hk/ERJ6SdHXiVZNrT2FiyapPaEBYg0RxWedc2h6_tM-X1q0iQ?e=q2recC}.