Max Robotics — Global Service Robot Directory + US Certification

About

Mateo Guaman Castro is a robotics researcher pushing the boundaries of autonomous navigation and manipulation in complex, unstructured environments. His work centers on self-supervised learning for off-road driving and large-scale robot manipulation, with a focus on enabling robots to reason about terrain interaction dynamics and risk without costly human labels. Guaman Castro’s major contributions include the development of TartanDrive 2.0, a large-scale off-road driving dataset with over seven hours of high-speed data up to 15 m/s, which has become a key resource for self-supervised learning research. His paper “How Does It Feel?” introduces a novel method for learning terrain traversability by estimating interaction dynamics, while “Learning Risk-Aware Costmaps” applies inverse reinforcement learning to generate safer navigation policies. His work on the DROID dataset (108 citations) provides a massive in-the-wild robot manipulation corpus, accelerating progress toward robust manipulation policies. More recently, he has explored agile continuous jumping for quadrupeds on discontinuous terrains like stairs. With over 230 total citations and a growing portfolio of influential datasets and methods, Guaman Castro is a rising leader in field robotics, bridging perception, learning, and control for real-world autonomy.