HarmonyNav: robot navigation in human-robot coexistence environment considering group and pedestrian intentions
Yutao Wang, Fang Fang, Chenyang Liu, Tao Fan, Yi Li · IET conference proceedings. · 2025
Service robots must navigate efficiently and harmoniously in human-robot coexistence environments. Previous studies often treated pedestrians as independent entities, ignoring human intentions and leading to group intersections and perfor- mance issues. We introduce HarmonyNav, a crowd navigation algorithm that enhances human-robot harmony. Our method uses Human-Robot Interaction sub-networks to account for indirect group impacts on pedestrians and predicts pedestrian trajectories to infer intentions. Employing proximal policy optimization (PPO) in reinforcement learning, we develop a harmonious navigation policy. In simulation experiments, our results show improved navigation performance, reducing group intersections and individual intrusions.