Reactive Humanoid Multi-Contact Using Learned Stability Models

2026-09-30

AI Summary

This study addresses the vulnerability of humanoid robots to falls in low-stability scenarios where reliance on foot contact alone is insufficient, proposing a reactive hand-support stabilization method. The core innovation lies in replacing conventional optimization solvers with a learning-based center-of-pressure region model, which is integrated with centroidal dynamics rollout simulation and multi-stage contact planning to enable rapid evaluation of candidate contact points and generation of optimal control policies. Both simulation and hardware experiments demonstrate that the proposed approach improves resistance to impulsive disturbances by 89%, reduces standing stabilization time by 43%, and increases walking recovery efficiency by 18%. These results indicate that the method significantly enhances the dynamic disturbance rejection capabilities of humanoid robots.

Abstract

We present a planning and control approach to reactively use hand contacts to stabilize a humanoid in low stability scenarios, where only using feet contacts may result in a fall. Candidate contacts are sampled within the robot's reachable workspace, and a preview is computed by rolling out the centroidal dynamics through pre-impact, impact and post-impact phases. Sampled points are scored based on the Center of Pressure (CoP) control authority at the post-impact phase. Central to our approach is a learned model of the robot's CoP region during post-impact, which enables rapid evaluation of candidate contact points compared to traditional optimization-based methods. The presented planner has two stages: the first selects an optimal bracing region and the second computes an optimal bracing point within the region. Our simulation results demonstrate an average increase in impulse resilience of 89% over recovery without hand contacts and 17% over a naive planning strategy (closest reachable region). We validate our framework on hardware, performing push tests while standing and walking. The standing trials show an average 43% reduction in stabilization time compared to naive hand placement and the walking trials demonstrate a 18% reduction compared to baseline recovery (without hand contacts).

Problem

• Humanoid robot
• Multi-contact stabilization
• Reactive balance
• Low stability scenarios

Innovation

• Humanoid Multi-Contact
• Learned Stability Model
• Centroidal Dynamics
• Reactive Planning
• Center of Pressure