Sara Ozeki

Tokyo University of Agriculture and Technology

Papers

Total Citations

H-Index

About

Sara Ozeki’s research sits at the exciting intersection of developmental robotics, cognitive architectures, and embodied AI, with a strong focus on building autonomous systems that learn efficiently and adapt to real-world tasks. Her most impactful contribution is the proposal of a curiosity-driven algorithm for robots, grounded in the Free Energy Principle, which enables sample-efficient data collection for reinforcement learning. This work, published in 2022 and garnering 6 citations, addresses a critical bottleneck in robotics: the inefficiency of random exploration. By mathematically formalizing intrinsic motivation, Ozeki’s algorithm allows agents to actively seek out informative states, dramatically accelerating task learning. Demonstrating a commitment to bridging theory and practice, she also led the development of “Dishflipper,” a fully integrated robotic system for automating dishwashing in a soba noodle stand. This project tackled the messy, real-world challenge of rinsing and flipping food debris, showcasing her ability to deploy sophisticated control systems in practical, high-variability environments. Through her work, Ozeki is shaping a future where robots are not just programmed, but are curious, self-motivated learners capable of mastering complex, unstructured tasks.