Papers
11
Total Citations
252
H-Index
7
About
Mathieu Geisert is a leading researcher in legged robotics, specializing in the intersection of model-based control and data-driven learning for quadrupedal locomotion. His work focuses on enabling dynamic, terrain-aware movement on complex, uneven terrain—a critical challenge for robots like the ANYmal platform. Geisert’s major contributions include developing the RLOC framework, which unifies reinforcement learning and optimal control to map sensory feedback into real-time footstep plans, and pioneering methods for reliable long-horizon trajectory planning using analytical costs and learned initializations. His research on real-time trajectory adaptation via deep reinforcement learning has allowed robots to circumvent computationally expensive online optimization, achieving robust, agile locomotion. With over 250 total citations, his most influential work (122 citations) demonstrates the practical impact of his hybrid approach. Geisert has also advanced contact planning with acyclic reachability-based methods and introduced disentangled gait representations through VAE-Loco, enabling versatile, continuous gait modulation. His contributions to the Loco3D project highlight his role in multi-contact locomotion for complex environments, cementing his reputation as a key innovator in making quadrupeds reliably traverse the real world.
Research Focus
Key Achievements
Top Papers
- 1
- 2
- 3
- 4
- 5
- 6
- 7
- 8
- 9Rapid Stability Margin Estimation for Contact-Rich Locomotion3 citations · 2021
- 10Regularized Hierarchical Differential Dynamic Programming2 citations · 2017