Papers
7
Total Citations
307
H-Index
5
About
Martin Pecka is a leading researcher in multi-robotic exploration, reinforcement learning safety, and autonomous navigation in extreme environments. He made his most prominent mark as a core member of the CTU-CRAS-NORLAB team in the DARPA Subterranean Challenge, where his work on multi-robotic systems for GPS-denied underground exploration (128 citations) directly advanced technologies for search-and-rescue operations. Pecka has also made foundational contributions to safe reinforcement learning, authoring a key overview of safe exploration techniques (79 citations) that remains essential reading for researchers seeking to deploy RL on physical robots without catastrophic failures. His innovative work on simulating non-deformable tracked vehicles introduced the "Contact Surface Motion" effect, enabling fast yet plausible physics simulation for tracked robots. More recently, Pecka has pushed the boundaries of autonomous navigation with MonoForce, a physics-informed, self-supervised model that predicts robot-terrain interaction on deformable surfaces like tall grass and bushes. His research consistently bridges the gap between theoretical safety guarantees and real-world robotic deployment, making him a pivotal figure in field robotics and safe autonomous systems.
Research Focus
Key Achievements
Top Papers
- 1
- 2Safe Exploration Techniques for Reinforcement Learning – An Overview79 citations · 2014
- 3
- 4Fast simulation of vehicles with non-deformable tracks31 citations · 2017
- 5Autonomous flipper control with safety constraints27 citations · 2016
- 6
- 7Fast Simulation of Vehicles with Non-deformable Tracks2 citations · 2017