HALO RL framework teaches humanoid robots to co-carry loads, hitting 86% success
CyberRobooo · x · 2026-10-30
A human and a Unitree G1 humanoid jointly carry a large tire, with the robot adjusting speed and body motion in real time as the person shifts posture or direction.
HALO is a new reinforcement learning framework for human-robot collaboration that uses Lyapunov stability constraints to reduce conflicts between independently learned human and robot policies.
- Achieves an 86% task success rate in simulation vs. 82.1% for the PCGrad baseline
- Real-world G1 tests show better responses when a human suddenly stops or changes carrying height
As humanoids enter factories, warehouses and homes, this kind of adaptive teamwork will be essential.
Related event: HALO Framework Teaches Humanoid Robots to Collaborate with Humans(4 posts)→
More from Embodied
- ETH Zurich and NVIDIA teach a quadruped robot to solve a Rubik's cube with its feet — lukas_m_ziegler · 2026-10-30
- OpenFlyV2: a 1B onboard VLA model for real-time aerial drone navigation — IPEC-COMMUNITY · 2026-10-30
- Robots from Asimov and Menlo Research survive brutal hardware testing — alysha_lobo · 2026-10-30
- XPENG's IRON humanoid robot walks off its new production line, 80%+ core processes automated — ingliguori · 2026-10-30
- Lenovo ThinkSmart View boots mainline Linux 7.x, turning the $30 device into a Claude/OpenAI speaker — DavidBennett__ · 2026-10-30
- Amazon's DreamFAR world model nearly doubles π₀ real-world performance for robot training — CyberRobooo · 2026-10-30