Kazutsugu Fujihara

About

Kazutsugu Fujihara is a pioneering researcher in the field of robotics and reinforcement learning, with a primary focus on enabling real-world robotic systems to learn autonomously in dynamic, physical environments. His most significant contribution lies in bridging the gap between theoretical reinforcement learning algorithms and practical robotic applications, addressing the critical challenges of computational cost and time constraints inherent in real-environment learning. His seminal 2002 paper, "Realtime reinforcement learning for a real robot in the real environment," which has garnered 18 citations, demonstrates a novel approach to allowing a physical robot to acquire behaviors through real-time learning, moving beyond the simulation-based paradigms that dominated the field. Additionally, his 1997 work on "Accelerating Reinforcement Learning for a Real Robot with Automated Abstract Sub-Rewards Generation" introduced an innovative method to speed up learning by automatically generating sub-rewards, further enhancing the efficiency of robotic skill acquisition. Fujihara's research has been instrumental in advancing the feasibility of deploying reinforcement learning in real-world robotics, making him a notable figure in the intersection of machine learning and autonomous systems.

Research Focus

Key Achievements

Top Papers

1. 1
   Realtime reinforcement learning for a real robot in the real environment

   18 citations · 2002
2. 2
   Accelerating Reinforcement Learning for a Real Robot with Automated Abstract Sub-Rewards Generation

   2 citations · 1997

Key Collaborators

Contact & Links