Aaron Rovinsky

Berkeley College, University of California, Berkeley

Papers

Total Citations

H-Index

About

Aaron Rovinsky is a robotics researcher whose work sits at the intersection of reinforcement learning and dexterous robotic manipulation, with a particular focus on enabling autonomous, real-world learning systems. His research addresses one of the field's most persistent challenges: teaching multi-fingered robotic hands to perform complex, contact-rich manipulation tasks without constant human oversight. Rovinsky's most influential contribution, "Dexterous Manipulation from Images: Autonomous Real-World RL via Substep Guidance" (2023, 13 citations), introduces a framework that leverages substep guidance to help robots learn intricate manipulation skills directly from visual input. His earlier work on reset-free reinforcement learning (2021, 9 citations) tackled the practical bottleneck of human intervention during training, proposing a multi-task learning paradigm that allows robots to autonomously collect experience through trial and error. More recently, his REBOOT framework (2023) advances data efficiency by reusing previously collected data to bootstrap learning for new dexterous tasks. Collectively, Rovinsky's research pushes toward genuinely autonomous robotic systems capable of operating in unstructured real-world environments—a critical step forward for practical robot deployment in everyday settings.