Abstract
We study supported reaching to extend a humanoid robot's workspace by bracing against the environment for stabilization using exclusively local sensing. While current solutions for whole-body humanoid control aim at avoiding contact, contact is needed to reach objects that are not directly placed at the edge of a table. We employ a sampling-based model-predictive control (MPC) framework for whole-body leaning, bracing, reaching, and recovery on a Unitree H1-2 humanoid. The controller runs in real time at 34 Hz, transfers sim-to-real, extends reach by ≈50 cm beyond the unbraced standing workspace, and sustains the loaded posture for over four minutes, before pose estimation drifts. The controller reaches the commanded target in 89% of trials and completes the full reach-and-recovery cycle in 81%, across 54 trials spanning six targets and three table heights. We further characterize stability, contact loading, and end-effector precision, and analyze failure modes and sampling-based planner sensitivity.
- Reach beyond the unbraced standing workspace+50 cm
- Loaded posture sustained4+ min
- Commanded reaches completed (48/54, six targets × three table heights)89 %
- Full reach-and-recovery cycles (44/54)81 %
- Target accuracy, wrist-camera measured (±0.9, n=48 holds)2.7 cm
- Maximum-extent reaches completed on hardware (11/15)73 %
- Planning rate, 20 rollouts over a 1 s horizon, laptop CPU34 Hz
Results