Kento Kawaharazuka
Papers
77
Total Citations
712
H-Index
15
About
Kento Kawaharazuka is a robotics researcher whose work spans two interconnected frontiers: biomimetic musculoskeletal humanoid systems and AI-driven robot learning. His early career made significant contributions to tendon-driven musculoskeletal humanoids, tackling the formidable challenge of controlling anatomically complex robotic structures where precise geometric modeling is nearly impossible. Landmark papers on antagonist inhibition control (35 citations), online joint-muscle mapping (28 citations), and the Musculoskeletal AutoEncoder framework (28 citations) established him as a leading voice in learning-based control strategies that allow robots to adaptively model their own bodies — a capability critical for human-inspired designs with inherently nonlinear dynamics. More recently, Kawaharazuka has pivoted toward leveraging large language models and vision-language models for robotics, producing highly impactful reviews and novel frameworks. His 2024 review of foundation models in real-world robotics (60 citations) and work on Vision-Language Interpreters for task planning (35 citations) reflect the field's broader shift toward generalist AI systems. His research on humanoid locomotion via adversarial imitation learning (31 citations) further demonstrates versatility across perception, planning, and physical control. With over 300 cumulative citations across a decade of research, Kawaharazuka represents a rare bridge between classical biomechanical robotics and cutting-edge AI integration.
Research Focus
Key Achievements
Top Papers
- 1Real-world robot applications of foundation models: a review60 citations · 2024
- 2Vision-Language Interpreter for Robot Task Planning35 citations · 2024
- 3
- 4
- 5
- 6
- 7
- 8
- 9VQA-based Robotic State Recognition Optimized with Genetic Algorithm22 citations · 2023
- 10