ResearchResearch paperRobotics & Embodied AI1 source · Oct 21, 2026

Test-Time Motion Steering for Perceptive Humanoid Fall Recovery

We present RISE, an approach to perceptive humanoid recovery that lets task objectives steer how the robot returns to standing at inference time.

Humanoid fall recovery requires selecting a motion that suits both the surrounding scene and the task.
We learn a recovery vocabulary that connects intended movements to motor actions through sensory feedback.
A visual diffusion planner learns from executions of this vocabulary to predict motion intent as future body states.
Simulation experiments demonstrate diverse recoveries, terrain adaptation, and test-time steering without retraining, complemented by real-world demonstrations on a Unitree G1.

[1]Test-Time Motion Steering for Perceptive Humanoid Fall Recovery
arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 21, 05:20 PM
We present RISE, an approach to perceptive humanoid recovery that lets task objectives steer how the robot returns to standing at inference time.
Humanoid fall recovery requires selecting a motion that suits both the surrounding scene and the task.

Extractive summary: sentences quoted from the sources.

Before this