Kazuhiro Sasabuchi

About

Kazuhiro Sasabuchi is a leading researcher at the intersection of robotics and artificial intelligence, specializing in humanoid robot design, learning-from-observation (LfO), and large language model (LLM)-driven robot control. His work bridges the gap between natural human instruction and complex robotic execution. Sasabuchi’s most influential contributions include pioneering the use of ChatGPT for long-step robot control (92 citations) and GPT-4V for multimodal task planning from human demonstration (60 citations), enabling robots to translate natural language and video into executable actions. He also made foundational advances in biologically-inspired musculoskeletal humanoids (63 citations), designing robots that mimic human anatomy for more natural movement. His research on the Seednoid robot platform (14 citations) demonstrates practical multipurpose design through continuous competition evaluation. Sasabuchi has further advanced multimodal LfO frameworks that integrate language, vision, and affordance reasoning for household robots, addressing common-sense semantic constraints. His work on task-grasping and grasp-type recognition leverages object affordances to improve robot teaching. With over 275 total citations, Sasabuchi’s research is shaping the future of intuitive, human-like robot learning and control.

Research Focus

Key Achievements

Top Papers

1. 1
   ChatGPT Empowered Long-Step Robot Control in Various Environments: A Case Application

   92 citations · 2023
2. 2
   Design Approach of Biologically-Inspired Musculoskeletal Humanoids

   63 citations · 2013
3. 3
   GPT-4V(ision) for Robotics: Multimodal Task Planning From Human Demonstration

   60 citations · 2024
4. 4
   The Seednoid Robot Platform: Designing a Multipurpose Compact Robot From Continuous Evaluation and Lessons From Competitions

   14 citations · 2018
5. 5
   Semantic constraints to represent common sense required in household actions for multimodal learning-from-observation robot

   11 citations · 2023
6. 6
   Task-grasping from a demonstrated human strategy

   11 citations · 2022
7. 7
   A Multimodal Learning-from-Observation Towards All-at-once Robot Teaching using Task Cohesion

   10 citations · 2022
8. 8
   Agreeing to Interact

   9 citations · 2018
9. 9
   Grasp-type Recognition Leveraging Object Affordance

   8 citations · 2020
10. 10
    ChatGPT Empowered Long-Step Robot Control in Various Environments: A Case Application

    7 citations · 2023

Key Collaborators

Contact & Links