Full scale humanoid runs over 7 meters per second and corners aggressively
Push the Limit: Learning High-Speed Sprinting and Aggressive Cornering on a Full-Scale Humanoid
Robotics
Summary
Controlling a full-size humanoid robot to sprint very fast and make sharp turns is hard because fast running causes strong oscillations that confuse simple control methods. The authors developed a new way for the robot to understand these rapid movements by using a special attention model and redesigned rewards that focus on overall direction rather than noisy turning data. They also trained the robot to switch between different skills smoothly by using multiple style checks depending on the desired motion. This lets the robot run fast and turn sharply with natural-looking movements, tested successfully in both simulations and on a real robot.
What this means in practice
- •For robotics engineers: Develop control software that enables humanoid robots to sprint and turn rapidly with stable, natural gaits for advanced mobility tasks.
- •For robotics hardware developers: Design full-scale humanoid robots with sensors and actuators optimized to support high-speed dynamic motions guided by learned policies.
Authors
Huaxing Huang, Ruidong Li, Jinchao Han, Ziyang Dong, Wenhao Cui
Abstract
We present a learning-based framework that enables a full-scale bipedal humanoid to sprint at over 7 m/s with motion-prior-guided gait and perform aggressive cornering at high speed with a pronounced trunk lean of up to $25^\circ$. Achieving this combination is challenging: above approximately 4 m/s, periodic foot impacts induce severe, high-frequency yaw-rate oscillations intrinsic to high-speed gaits. This high-frequency temporal structure corrupts the history of proprioceptive feedback that flat-MLP policies rely on, while rendering conventional yaw-rate tracking rewards counterproductive, as they fight against the robot's natural oscillatory dynamics. Moreover, a single style prior shared across skills cannot tell which skill is commanded and inherently compromises the low-speed skill fidelity. To overcome these limitations, we adopt a causal multi-head attention (MHA) module to explicitly model the high-frequency temporal feedback. In place of the oscillatory yaw rate, we design a heading-direction reward that supervises only the low-frequency velocity direction, allowing agile turning at high speed without suppressing natural gait oscillations. We further introduce skill-conditioned multi-adversarial-motion-prior (MAMP) discriminators that condition the style reward on the commanded skill. Ablation experiments confirm that each mechanism is essential: replacing structured temporal encoding with a flat-history MLP caps the top speed; reverting to yaw-rate tracking prevents high-speed turning; and a single shared discriminator makes in-place turning collapse into a quasi-static stance. Our framework bridges the gap between raw sprinting speed and reference-motion fidelity, demonstrating stable high-speed sprinting and aggressive cornering in both simulation and real-robot deployment.