ParkourFormer: Integrating Predictive Supervision and Sequence Modeling into Parkour Locomotion
2026-05-25 • Robotics
Robotics
AI summaryⓘ
The authors created ParkourFormer, a new method that helps humanoid robots plan better for complex movements like jumping and climbing by predicting their future body states. Unlike previous approaches that act only based on the current situation, this method looks at past movements and guesses what will happen next to make smarter decisions. Tests on different terrains showed that ParkourFormer successfully completes tricky courses much more often than older methods. This shows predicting future body positions helps robots move more smoothly and reliably.
Humanoid robotParkourReinforcement learningTransformer modelSequence modelingProprioceptionFuture state predictionCross-attentionLocomotion policySimulation
Authors
Yanheng Mai, Wenhao Xu, Zirui Huang, Yifei Fu, Shengwei Dong, Xinjue Wang, Kailun Huang, Yanzhe Xie, Renjing Xu
Abstract
Humanoid parkour requires locomotion policies to coordinate whole-body dynamics across rapidly changing terrains such as stairs, gaps, slopes, and obstacles. Existing reinforcement learning policies are largely reactive, mapping observations directly to actions without explicitly modeling future body states. Such modeling becomes critical in agile locomotion tasks where successful motion execution depends strongly on anticipating upcoming contact transitions and body dynamics.We present ParkourFormer, a Transformer-based sequence modeling framework that reformulates humanoid locomotion as a future-conditioned decision-making problem. The current robot state queries historical sensorimotor trajectories through cross-attention, while a lightweight prediction head forecasts short-horizon future proprioceptive states. The predicted future states, trained with supervised signals, are fused with temporal features to generate actions, enabling the policy to jointly reason over motion history and anticipated future dynamics. We evaluate ParkourFormer on a diverse multi-terrain humanoid parkour benchmark including stairs, gaps, slopes, rough terrain, and obstacle traversal. Experiments in simulation and on a real humanoid robot show that ParkourFormer achieves a 93.85% average traversal success rate on highly challenging terrains, with improvements of up to 42.73% over strong MLP, MoE-based MLP, and vanilla Transformer baselines, while maintaining a single unified policy across all terrain types. These results demonstrate that explicit future-state modeling significantly improves robustness and generalization for agile whole-body locomotion.