PF-TEB: Timed Elastic Band-Based Human-Aware Robot Navigation Framework in Crowded Environments

Abstract

To enhance the social navigation performance of mobile robots in crowded environments, we propose a novel framework—Prediction and Fuzzy Timed Elastic Band (PF-TEB) for robot social navigation. Our framework incorporates predicted pedestrian trajectories into pedestrian proxemics modeling as a social layer cost map, allowing the robot to better interpret pedestrian movement intentions and proactively avoid potential conflicts. Additionally, social constraints are integrated into TEB to ensure that both pedestrian comfort and safety are prioritized during trajectory planning. To address the adaptability limitations of conventional methods in highly dynamic, crowded environments, we introduce three type-2 fuzzy logic controllers. These controllers dynamically adjust the optimization weights of the corresponding constraints in the objective function based on real time environmental data, enabling timely modifications to the planned trajectory and significantly enhancing the flexibility and adaptability of the navigation system. The simulation experiments demonstrate that our method performs well in social metrics by respecting pedestrian comfort and its effectiveness has been validated on the Tracer robot in real-world environments.