Andrew Schwartzwald

University of Minnesota, SpaceX (United States)

Papers

Total Citations

H-Index

About

Andrew Schwartzwald is a roboticist focused on bridging the gap between simulation and reality for agile, unconventional locomotion. His primary research centers on applying reinforcement learning (RL) to control tumbling robots—simple, robust platforms capable of traversing obstacles far larger than themselves, but notoriously difficult to command due to their chaotic, non-linear dynamics. Schwartzwald’s key contribution is the development of adaptive control policies that transfer seamlessly from simulation to the physical world, overcoming the "sim-to-real" gap. His foundational 2020 paper, *"Sim-to-Real with Domain Randomization for Tumbling Robot Control"* (4 citations), introduced a domain randomization technique that allows RL policies to generalize to real-world physics without requiring exact modeling. He extended this in his 2022 work, *"Tumbling Robot Control Using Reinforcement Learning: An Adaptive Control Policy That Transfers Well to the Real World"* (3 citations), demonstrating a policy that outperforms traditional flat-terrain assumptions by enabling the robot to actively explore and exploit its full range of motion. While his citation count is modest, his work is notable for pioneering practical, transferable RL solutions for a challenging class of robots, offering a blueprint for deploying learned controllers on low-cost, rugged hardware.