Martin Stolle

Google (Switzerland), Carnegie Mellon University, Forbes Hospital

Papers

Total Citations

257

H-Index

About

Martin Stolle is a leading researcher in robot locomotion and machine learning, whose work has fundamentally advanced how robots move across challenging terrain. His key research areas include legged locomotion, trajectory optimization, and policy transfer in robotics. Stolle’s most influential contribution is his 2011 paper on “Optimization and learning for rough terrain legged locomotion,” which has garnered 122 citations. In this seminal work, he introduced a hierarchy of fast, anytime algorithms that plan footholds and dynamic body motions, enabling robots to navigate uneven surfaces with unprecedented efficiency. This approach revolutionized the field by embedding optimization directly into the control loop, making real-time adaptation possible. Stolle also pioneered the use of trajectory libraries for policy creation, as seen in his 2006 paper (77 citations), which offered a powerful alternative to traditional value-function methods. His 2007 work on policy transfer (38 citations) further extended this concept, allowing robots to reuse learned behaviors across different tasks—a critical step toward generalizable robotic intelligence. By bridging optimization, learning, and control, Stolle has shaped modern locomotion research, inspiring new generations of robots that can walk, climb, and adapt in the real world.