Kanghui Du

Kanghui Du

I am currently a postdoctoral researcher at Kyoto University. My current research focuses on designing robot interaction systems built on top of VLMs and LLMs, with an emphasis on how robots can communicate intent and behave more naturally in public environments.

I completed my Ph.D. in Human-Robot Interaction at Kyoto University in September 2025, after earning an M.A. in Social Informatics from Kyoto University and a B.Sc. in Computer Science from China University of Geosciences, Beijing. My work has explored suggestive avoidance, repeated obstruction prevention, and human-inspired responses to sudden robot stops.

I have also worked as a Research Assistant in the HRI Lab at Kyoto University and collaborated with Advanced Telecommunications Research Institute International (ATR). Most of my work is grounded in real-world deployments and experiments rather than simulation-only settings. I am broadly interested in human-centered AI, social robotics, and interaction design for embodied systems, and I am currently seeking academic positions.

news

Presented my HRI 2025 work in Melbourne at the ACM/IEEE International Conference on Human-Robot Interaction.

Completed Ph.D. in Human-Robot Interaction at Kyoto University.

Presented poster on human-inspired solutions for sudden robot stops at MoonShot 2025.

Served as Research Assistant at the HRI Lab, Kyoto University.

Began collaboration with ATR as an assisting international researcher.

Passed the Japanese-Language Proficiency Test N1.

selected publications

  1. Don’t Just Stop Here! Human-Inspired Solutions for Sudden Robot Stops

    Du, Kanghui, D. Brscić, and T. Kanda. Proceedings of the 2025 ACM/IEEE International Conference on Human-Robot Interaction, 2025, pp. 293-302.

    Examines how robots should recover from abrupt stops in shared spaces and proposes socially legible behaviors that better align with human expectations.

  2. Can’t You See I Am Bothered? Human-inspired Suggestive Avoidance for Robots

    Du, Kanghui, D. Brscić, Y. Liu, and T. Kanda. Proceedings of the 2024 ACM/IEEE International Conference on Human-Robot Interaction, 2024, pp. 184-193.

    Studies how robots can express intent through motion to encourage people to give way in a natural and effective manner.

  3. Recognition of Rare Low-Moral Actions Using Depth Data

    Du, Kanghui, T. Kaczmarek, D. Brscić, and T. Kanda. Sensors, 20, 2020.

    Uses depth sensing to recognize rare but socially meaningful actions in public environments.

  4. Replicating the Suggestive Avoidance Model for Robot Repeated Obstruction Prevention in a Spanish University

    Du, Kanghui, D. Brscić, A. G. Anaís, T. Kanda, and A. Sanfeliu. Submitted to ACM THRI, 2025.

    Extends prior suggestive avoidance work by examining repeated obstruction prevention in a new university setting.

teaching

At Kyoto University, I worked as a Teaching Assistant for Introduction to Artificial Intelligence, Human-Robot Interaction, Practice of Information Systems, Design Principles for Information and Communication Technology, and ILAS Seminar.

Teaching has been an important part of my academic work, helping me support students from diverse backgrounds while refining how I explain research ideas and technical systems.

Awards & Honors

  • JSPS DC1 Research Fellow
  • JSPS Grant, 220,000 JPY, Role: PI
  • JLPT N1
  • ACM-ICPC Asia Beijing Regional Bronze Medal
contact

If you are interested in human-robot interaction, socially aware navigation, or collaboration on embodied AI research, feel free to reach out.

du.kanghui.2n@kyoto-u.ac.jp +81-70-3847-7996 github.com/TerransForce