Track genomics, neuroscience and molecular biology.
Search papers, labs, and topics across Lattice.
TERRA addresses the challenge of musculoskeletal locomotion over non-flat terrain by introducing an end-to-end pipeline that reconstructs terrain geometry from kinematic trajectories and retargets motion while respecting anatomical and contact constraints. This approach enables the training of a single muscle-actuated control policy on diverse, terrain-aware locomotion data, achieving significantly improved terrain accuracy and reduced violations. The key result is a practical method for generating robust musculoskeletal locomotion across varied non-flat terrains, overcoming limitations of current flat-ground focused methods.
Muscle-actuated agents can now learn to locomote over diverse non-flat terrain by inferring terrain geometry and contact from kinematic data alone.
Recent advances in musculoskeletal modeling and reinforcement learning have enabled muscle-actuated agents to reproduce increasingly complex human motions. Yet these capabilities remain largely confined to flat ground, in part because motion datasets rarely include aligned terrain geometry and because retargeting terrain interactions to complex musculoskeletal bodies is challenging. We present TERRA, an end-to-end pipeline for terrain-aware retargeting and control of musculoskeletal locomotion. From kinematic trajectories alone, TERRA combines terrain priors, estimated contacts, and negative free-space evidence to recover task-relevant support geometry. TERRA further considers anatomical, tendon-continuity, and contact constraints during retargeting. Using the resulting motion-terrain pairs from five datasets, we successfully train a single muscle-actuated control policy on 9.4 hours of diverse locomotion. Across reconstruction, retargeting, and held-out tracking benchmarks, TERRA improves terrain accuracy, sharply reduces anatomical and interaction violations, and achieves the highest observed completion rate over supported terrain families. Overall, TERRA provides a practical route from scene-less motion data to muscle-actuated locomotion over diverse non-flat terrain. Project website: https://cnai.epfl.ch/terra/