Robot Navigation in Dynamic and Crowded Environments

Mobile robots are increasingly implemented in various scenarios. However, navigating in crowded environments with moving obstacles and uncertain intentions poses significant challenges for robots. To address this challenge, we propose a novel social-based planning framework, which integrates behavior planning and motion planning to generate safe and efficient paths for robot navigation. In the behavior planning part, we first design an efficient method based on the graph to construct frozen areas to increase safety. Then, we build the cost function and propose the social behavior sampling strategy to adaptively generate behavior paths for dynamic environments. Next, we design the backward optimization algorithm that can directly optimize behavior paths in the continuous space efficiently. In the motion planning part, we develop the forward reachable set (FRS)-based algorithm to sample paths that can satisfy kinematic constraints. Furthermore, we design a collision-risk evaluation method to determine the optimal motion path by considering the interactions. Experiments and comparison studies verify the advantages of the proposed path-planning method in terms of average moving speed and collision risk.