Naoto KAWABE

About

Naoto Kawabe is a pioneering researcher in the field of robotic motion acquisition, specializing in the application of reinforcement learning to dynamic, gymnastic-like maneuvers for compact humanoid robots. His major contribution lies in demonstrating that complex, model-free motions—specifically the robotic giant-swing—can be learned solely through environmental interaction, challenging the conventional wisdom that reinforcement learning methods like Q-learning are ill-suited for dynamic tasks where the Markov property is not guaranteed. Kawabe’s most cited work (2009, 14 citations) established a foundational approach for enabling humanoid robots to acquire such motions without pre-existing robotic models. He further advanced this line of inquiry by improving motion repeatability (2011, 6 citations) and analyzing the resulting motion forms (2010, 4 citations). Though his citation counts reflect a specialized niche, his research is notable for pushing the boundaries of model-free learning in robotics, offering a proof-of-concept that could inspire future work in agile, human-like robot locomotion. Kawabe’s studies remain a touchstone for researchers exploring the intersection of reinforcement learning and dynamic physical control.

Research Focus

Key Achievements

Top Papers

1. 1
   Consideration on robotic giant-swing motion generated by reinforcement learning

   14 citations · 2009
2. 2
   Acquisition of a Gymnast-Like Robotic Giant-Swing Motion by Q-Learning and Improvement of the Repeatability

   6 citations · 2011
3. 3
   Acquisition of Robotic Giant-swing Motion Using Reinforcement Learning and Its Consideration of Motion Forms

   4 citations · 2010

Key Collaborators

Contact & Links