Max Robotics — Global Service Robot Directory + US Certification

About

Michel Tokic is a researcher whose work sits at the intersection of reinforcement learning, robotics, and intelligent autonomous systems. He is perhaps best known for developing innovative physical platforms to make reinforcement learning tangible and accessible, most notably through the "Crawler" — a compact two-degree-of-freedom crawling robot capable of learning a forward-walking policy from scratch in under 20 seconds. This work, along with related publications on teaching reinforcement learning through physical robots, has garnered significant attention in educational and research communities, accumulating citations that reflect its practical impact on how the field is taught and demonstrated. Beyond pedagogy, Tokic has made meaningful contributions to the theoretical and applied challenges of reinforcement learning, particularly around the exploration-exploitation trade-off. His research on meta-learning of these parameters using eligibility traces and his investigations into robust exploration strategies for safety-critical applications highlight a sustained commitment to making autonomous systems both adaptive and reliably safe. His work on learning safety knowledge from human demonstrations further underscores an interest in bridging human expertise and machine autonomy. Through the open-source Teaching-Box framework, Tokic has also worked to lower the barrier for engineers building learning robots, leaving a practical legacy across education, robotics, and safe autonomous systems research.