Max Robotics — Global Service Robot Directory + US Certification

About

Ryo Okumura is a robotics researcher whose work bridges the gap between perception and control for complex industrial manipulation. His primary research areas include imitation learning, state representation learning, and tactile-based control, with a strong focus on enabling robots to perform high-precision tasks. Okumura’s most notable contribution is the development of the Tactile-Sensitive NewtonianVAE, a world model that achieves submillimeter positioning accuracy for industrial connector insertion—a task requiring delicate grasp pose compensation. This work, which has garnered 14 citations, demonstrates how tactile feedback can be integrated with variational autoencoders for real-world assembly. He has also advanced imitation learning through his domain-adversarial and conditional state space models, which learn domain-agnostic state representations from partially observable data, achieving 11 citations. More recently, Okumura has explored safe variable stiffness learning via multi-objective Bayesian optimization, prioritizing compliance alongside task performance. His multi-view dreaming framework extends reinforcement learning to multi-camera setups, enabling more robust scene understanding. With a growing body of work that consistently addresses real industrial challenges, Okumura is establishing himself as a key figure in practical robot learning.