Max Robotics — Global Service Robot Directory + US Certification

About

Michael T. Rosenstein is a leading researcher in robot motor learning and human-robot interaction, whose work bridges reinforcement learning, control theory, and biologically inspired robotics. His most influential contribution is the development of supervised actor-critic reinforcement learning, a framework that integrates supervisory signals with traditional RL to accelerate skill acquisition—a method detailed in his highly cited 2012 paper (69 citations). Rosenstein’s early work on robot weightlifting via direct policy search (80 citations) demonstrated how simple search algorithms combined with biological constraints could achieve effective motor skill learning. He also made foundational contributions to dynamic manipulability analysis, introducing velocity-dependent measures (42 citations) that remain essential for robot design and control. His research on mixed-initiative teleoperation (36 citations) addressed user fatigue by developing methods for extracting user intent, while his work on autonomous humanoid tool use (26 citations) advanced the DARPA Mobile Autonomous Robot Software program using NASA’s Robonaut platform. Rosenstein’s continuous category learning algorithms (31 citations) enabled robots to autonomously organize sensory experiences. With over 350 total citations across his most-cited works, his research has shaped modern approaches to robot learning, particularly in making reinforcement learning practical for complex, real-world tasks through structured policies and supervisory integration.