Hikaru Sasaki

About

Hikaru Sasaki is a robotics and machine learning researcher whose work centers on reinforcement learning, imitation learning, and autonomous control in challenging real-world environments. Sasaki's research addresses some of the most pressing practical obstacles in deploying intelligent systems outside of controlled laboratory settings, particularly in industrial and weakly observable scenarios. Among Sasaki's notable contributions is the development of disturbance-injected robust imitation learning, which tackles a fundamental limitation of conventional imitation learning methods: the assumption that demonstrations are both optimal and sufficiently varied. By injecting disturbances during training, Sasaki's approach enables learned policies to generalize more robustly — a critical advancement for real-world deployment. This work has garnered 9 citations since its 2022 publication, reflecting growing interest in practical imitation learning. Sasaki has also pioneered Gaussian Process self-triggered policy search tailored for weakly observable environments, such as waste cranes in industrial incineration plants — settings where sensor limitations make state estimation inherently difficult. This work demonstrates a commitment to translating advanced machine learning theory into meaningful solutions for large-scale industrial automation. With a research agenda bridging theoretical rigor and real-world applicability, Sasaki represents an emerging voice in intelligent autonomous systems research.

Research Focus

Key Achievements

Top Papers

1. 1
   Disturbance-injected Robust Imitation Learning with Task Achievement

   9 citations · 2022
2. 2
   Gaussian Process Self-triggered Policy Search in Weakly Observable Environments

   2 citations · 2022

Key Collaborators

Contact & Links