Ashish Kumar

Berkeley College, University of California, Berkeley

Papers

Total Citations

476

H-Index

About

Ashish Kumar is a robotics researcher whose work sits at the intersection of reinforcement learning, locomotion, and real-world robot deployment. His most celebrated contribution, **Rapid Motor Adaptation (RMA)** (2021), addresses one of the central challenges in legged robotics: enabling quadruped robots to adapt in real-time to unpredictable real-world conditions such as shifting terrains, variable payloads, and mechanical wear. The algorithm has garnered over 447 citations, establishing Kumar as a prominent voice in adaptive locomotion research and influencing a generation of subsequent work in agile robot control. Beyond locomotion, Kumar has explored strategic robot behavior through vision-based pursuit-evasion policies, tackling the complex problem of planning under physical and intentional uncertainty. His earlier work on learning navigation subroutines from egocentric video data demonstrated a keen interest in hierarchical reinforcement learning and sample-efficient planning. He has also contributed to democratizing robotics research through the OffWorld Gym, an open-access physical robot environment designed to push the community toward real-world reinforcement learning benchmarks. Across these efforts, Kumar's research reflects a consistent commitment to bridging the gap between simulation and physical deployment, making robots genuinely capable of operating in the messy, unpredictable real world.