ForPC: Force-Aware Whole-Body Control for Precise Humanoid Cart Pushing
ForPC: Force-Aware Whole-Body Control for Precise Humanoid Cart Pushing
This paper introduces ForPC, a force-aware whole-body control framework for humanoid robots designed to push wheeled carts in manufacturing and logistics settings. The method uses two-stage reinforcement learning to regulate cart velocity while maintaining balance under varying loads and postures. Trained entirely in simulation, the policy transfers to a physical Unitree G1 robot without additional training or fine-tuning, successfully pushing carts carrying up to 30 kg at 0.5 m/s. Experiments show significant improvements over baseline methods in both cart velocity tracking and cross-track error on multi-bend routes.
The study presents a simulation-trained control framework that transfers to a physical Unitree G1 robot for pushing carts up to 30 kg at 0.5 m/s without payload-specific adjustment or real-world fine-tuning, reducing cart-velocity RMSE by up to 68.2% relative to a locomotion-only baseline.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
Reported numbers
cart payload capacity
≤30
View original evidence
motion-only baseline (Loco-only) and cross-track RMSE by up to 74.2% on the multi-bend route. Trained solely in simulation, the policy also transfers to a physical Unitree G1 without real-world training or fine-tuning and, with no payload-specific adjustment, pushes a cart carrying up to 30 kg at 0.5 m/s. These results suggest that humanoid robots could eventually replace humans for efficient material handling in dai
Open source S5cart pushing speed
0.5
View original evidence
motion-only baseline (Loco-only) and cross-track RMSE by up to 74.2% on the multi-bend route. Trained solely in simulation, the policy also transfers to a physical Unitree G1 without real-world training or fine-tuning and, with no payload-specific adjustment, pushes a cart carrying up to 30 kg at 0.5 m/s. These results suggest that humanoid robots could eventually replace humans for efficient material handling in dai
Open source S5
- paper: not_reported
- code: not_reported
- weights: not_reported
- dataset: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
rnal force sensors, and fed to a residual policy that coordinates bounded arm, waist, and leg adjustments with base-velocity commands. Multi-objective rewards penalize bimanual internal opposition, grasp slip, and actuator saturation, encouraging the policy to apply contact force on demand without human demonstrations. Sim-to-sim experiments show that ForPC reduces cart-velocity RMSE by up to 68.2% relative to a loco
Open source S4
motion-only baseline (Loco-only) and cross-track RMSE by up to 74.2% on the multi-bend route. Trained solely in simulation, the policy also transfers to a physical Unitree G1 without real-world training or fine-tuning and, with no payload-specific adjustment, pushes a cart carrying up to 30 kg at 0.5 m/s. These results suggest that humanoid robots could eventually replace humans for efficient material handling in dai
Open source S5
Source：arXiv Robotics — research abstracts · arxiv.org