Hidehito Fujiishi

Nara Institute of Science and Technology

Papers

Total Citations

H-Index

About

Hidehito Fujiishi is a researcher advancing the frontier of robot learning, with a primary focus on **imitation learning**, **behavioral cloning from observation (BCO)**, and **safe autonomous policy acquisition**. His major contributions lie in developing methods that allow robots to learn complex tasks by observing human demonstrations—without requiring access to the expert’s action data. In his 2021 work, "Safe and efficient imitation learning by clarification of experienced latent space" (6 citations), Fujiishi addressed a critical limitation of BCO: the risk of robot failures during the few interactions needed to infer expert actions. He proposed a framework that clarifies the latent space of the robot’s experience, enabling safer and more efficient policy learning. His follow-up paper, "Behavioral Cloning from Observation with Bi-directional Dynamics Model" (2 citations), further refined this approach by introducing a bi-directional dynamics model to improve the accuracy of action inference from observations alone. Though early in his career, Fujiishi’s work is significant for its focus on **practical safety and data efficiency** in imitation learning—key challenges for deploying robots in human-centric environments. His research is particularly relevant for students and engineers working on **learning from demonstration**, **robot safety**, and **model-based reinforcement learning**.