Mobile robot navigation in human populated environments has been widely studied in the past two decades. Significant improvements in this technology suggest a promising future for introducing mobile service robots in human workspaces that help with daily activities. State-of-the-art approaches have shown real-world deployments of robots that can safely navigate through dense crowds. Still, most robots lack the ability to navigate in a human-friendly manner, which requires the ability to identify human intentions, especially in collision-risky situations, and to avoid collisions “legibly” Most studies to date do not take human-robot interaction into consideration, resulting in unexpected human behaviors during deployment, and ineffective robot planning due to large prediction errors regarding human responses. In this work, we therefore seek to model human reactions around a robot in collision-dangerous scenarios, with specific study of situations in which two agents cross paths. We model people's collision avoidance behaviors as a function of their underlying intentions and social preferences, and propose to predict human motions based on such intentions and preferences. Preliminary results suggest large variations in different people's crossing path selection. We turn to a psychological model to explain such diverse responses, and propose to categorize people's strategies based on their assumptions about the robot.

PDF, HTML

In Proceedings of the Human-Centered Robotics workshop of the 13th International Conference on Robotics: Science and System (RSS), Cambridge, MA, USA, July 2017.

@inproceedings{RSS17-yunl, title={Iterative Human-Aware Mobile Robot Navigation}, author={Shih-Yun Lo and Benito Fernandez and Peter Stone}, booktitle={Proceedings of the Human-Centered Robotics workshop of the 13th International Conference on Robotics: Science and System (RSS)}, month={July}, address={Cambridge, MA, USA}, url="http://www.cs.utexas.edu/users/ai-lab?RSS17-yunl", year={2017} }