FALCON: Learning Force-Adaptive Humanoid Loco-Manipulation

TL;DR

FALCON tackles forceful loco-manipulation in humanoids by decoupling control into lower-body locomotion and upper-body end-effector tracking through two specialized policies trained jointly with shared proprioception. It introduces a torque-limit-aware 3D force curriculum that bounds end-effector forces within joint-torque limits while progressively increasing difficulty and randomizing contact points to improve robustness. Across simulations and real robots (Unitree G1 and Booster T1), FALCON achieves superior upper-body tracking and stable locomotion under substantial external forces, with faster training and strong cross-platform transfer, including teleoperation and autonomous tote logistics. Limitations include focusing solely on end-effector forces and not addressing multi-contact scenarios or torque disturbances beyond end-effectors, pointing to future work on multi-contact reasoning and torque-adaptive policies.

Abstract

Humanoid loco-manipulation holds transformative potential for daily service and industrial tasks, yet achieving precise, robust whole-body control with 3D end-effector force interaction remains a major challenge. Prior approaches are often limited to lightweight tasks or quadrupedal/wheeled platforms. To overcome these limitations, we propose FALCON, a dual-agent reinforcement-learning-based framework for robust force-adaptive humanoid loco-manipulation. FALCON decomposes whole-body control into two specialized agents: (1) a lower-body agent ensuring stable locomotion under external force disturbances, and (2) an upper-body agent precisely tracking end-effector positions with implicit adaptive force compensation. These two agents are jointly trained in simulation with a force curriculum that progressively escalates the magnitude of external force exerted on the end effector while respecting torque limits. Experiments demonstrate that, compared to the baselines, FALCON achieves 2x more accurate upper-body joint tracking, while maintaining robust locomotion under force disturbances and achieving faster training convergence. Moreover, FALCON enables policy training without embodiment-specific reward or curriculum tuning. Using the same training setup, we obtain policies that are deployed across multiple humanoids, enabling forceful loco-manipulation tasks such as transporting payloads (0-20N force), cart-pulling (0-100N), and door-opening (0-40N) in the real world.

Figures (12)

• Figure 1: FALCON enables versatile forceful loco-manipulation tasks for humanoids: (a) Transporting Payloads: walk, squat, twist torso with payloads; (b) Cart-Pulling with significant longitudinal forces; (c) Door-Opening using both arms with multi-directional forces. Videos: https://lecar-lab.github.io/falcon-humanoid
• Figure 2: Overview of FALCON. (a) Two agents with different sub-tasks are jointly trained with shared whole-body proprioception. During training, we apply 3D external forces bounded by upper-body joint torque limits on the end-effectors; (b) FALCON is deployed with either teleoperation or an autonomy pipeline including FoundationPose wen2024foundationpose for pose estimation and motion planning.
• Figure 3: (a) Progression of the Apply Force Scale $\alpha_g$; (b) Upper-body Joint Tracking Errors During Training
• Figure 4: Comparison of FALCON and M-WB-RL: (a) action noise std; (b) tracking errors and penalties.
• Figure 4: Real-world Tracking Errors.