Organizations: Shanghai AI Laboratory · Tsinghua University · Shanghai Jiao Tong University · Zhejiang University · The Chinese University of Hong Kong
Humanoid loco-manipulation requires coordinated locomotion, balance regulation, and manipulation under frequent contacts, making teleoperated demonstration collection difficult: a practical controller must simultaneously provide precision, stability, and contact awareness. We propose FORGE, a force-aware multi-granularity control system with a coarse-to-fine command space spanning velocity navigation, planar foot placement, wrist-pose tracking, and optional force-conditioning inputs. Policy training requires no full-body human motion trajectories or joint-angle references; deployment consumes only sparse bilateral wrist poses relative to the pelvis, coarse motion commands, and optional desired wrist-load inputs. For sustained contact, FORGE combines a momentum observer with a learned correction trained using simulator contact-force supervision. At deployment, it estimates low-frequency, wrist-equivalent interaction forces from proprioception and reconstructed low-level arm PD torques, without wrist force/torque sensors. A quasi-static virtual-target loop uses this estimate and an optional desired applied-force input to adapt the wrist target. A frozen locomotion prior and asymmetric residual actions stabilize manipulation-induced whole-body corrections. Experiments evaluate unified-control stability, task-space tracking, quasi-static wrist-equivalent force estimation, squeezing forces during sustained contact with a rigid box, and downstream diffusion-policy execution of a squat-and-lift task across payload conditions. Website: https://bb0928.github.io/forge/.
Figures & tables
Fig. 2 : Overall pipeline of FORGE. The framework has three components: multi-granularity control over navigation, stance, pose, and optional force-conditioning inputs; a physics-informed hybrid force estimator and virtual-target contact-adaptation loop; and asymmetric residual actions anchored on a frozen locomotion prior. Policy training uses sparse task-space commands rather than full-body human motion references.
Fig. 3 : Actor architecture. Solid deployment paths use only the masked observable history. “Wrist Force” denotes the observer-derived residual history, not a force/torque-sensor reading or simulator force. The adaptation decoder’s 6-D correction refines the observer at deployment; the dashed privileged quantities supply training targets only.
Fig. 5 : Sequential bottle placement. Four ordered frames show bottles placed into a basket, with vertical wrist-equivalent force traces at marked releases.
Method
Latency (ms) ↓
Foot MPE (cm) ↓
Wrist MPE (cm) ↓
Contact force (N) ↓
HOMIE
27
N/A
N/A
15.48
TWIST2
197
7.87
7.76
17.16
SONIC
252
7.35
6.42
15.32
FORGE
22
5.42
4.15
8.52
TABLE I : Comparison with Humanoid Controllers
Reward Term
Weight
Formulation / Error Term
Command Tracking
Vel X / Y Track
2.0 / 1.5
exp(−∥Δvx,y∥2/σ)
Ang Vel / Int Track
2.0 / 2.0
exp(−∥Δωz∥2/σ) , exp(−Ierr/σint)
Base Height Track
2.0
exp(−∥Δh∥⋅10)
End-Effector Tasks
Wrist Pos Track
5.0
exp(−∥Δpw∥/σw)
TABLE II : Principal Reward Terms and Formulations
Metric
w/o mask
Ours
Valid rollout (s) ↑
15.69±1.32
19.46±0.52
Lin. velocity error (m/s) ↓
0.194±0.003
0.162±0.003
Planar feet MPE (m) ↓
0.182±0.002
0.167±0.002
Wrist MPE (m) ↓
0.115±0.003
0.079±0.004
TABLE III : Masking Ablation Results
Figure 8
Fig. 8 : Representative single-trial real-world x/y/z wrist-equivalent interaction-force estimates at staged 0,5,10,15,20 N loads. Each load is repeated 10 times with digitally synchronized force-gauge references.
Fig. 9 : Contact-load adaptation and wrist-equivalent force estimation. (a) Foam-box lifting: feedback enabled (left) reduces visible deformation relative to feedback disabled (right). (b) Force-test setup. (c) Planar-load evaluation under external forces in [−30,+30] N: the hybrid estimator records lower error than the learned-only estimator. Shading denotes one standard deviation over ten repeats, using synchronized force-gauge references.
Fig. 10 : Walking, squat/lift, and foot-position control: wrist errors and pelvis acceleration briefly deviate and recover under wrist pushes and base drags.
Estimator
x axis
y axis
z axis
Hybrid
0.81/0.50
1.97/1.57
0.87/0.67
Observer-only
2.83/2.29
5.15/4.43
3.02/2.58
Learned-only
2.43/2.19
2.23/1.92
2.88/2.47
TABLE IV : Force-Estimation Errors under Staged Loads
Payload
Empty
0.5 kg
1 kg
2 kg
Success rate
100%
100%
90%
90%
TABLE V : Cross-Payload Box-Transport Success Rates
Many everyday tasks require controlling contact force as well as motion. On a humanoid, this force must often be coordinated across both hands and the whole body, and cannot be specified by kinematic trajectories alone. Therefore, we present FLORE, a framework for learning force-reactive humanoid loco-manipulation. We collect robot-free demonstrations using two force-sensing UMI-style handheld grippers for hand motion and contact wrenches, and wearable trackers for body motion. A force-reactive diffusion policy takes vision, proprioception, and measured wrenches as input, then predicts actions and per-hand force commands. A hybrid compliance whole-body controller with force tracking executes these commands while maintaining balance: an integral of the force error measured by real force/torque (F/T) sensors lets each hand track force accurately along the commanded direction and remain compliant along the others. In simulation, the controller tracks per-hand force commands of varied magnitude and direction, with little loss in whole-body motion accuracy. Across three contact-rich tasks on the robot, the force-reactive policy succeeds in 80% of rollouts, compared with 26.7% for a position-only policy trained on the same demonstrations and executed by the same controller. For more details, please visit our website: https://louis-leee.github.io/FLoRe/.
Sihang Li, Xiaofeng Guo, Jinkun Cao +5
Amazon FAR (Frontier AI & Robotics) · New York University · Carnegie Mellon University +1
Humanoid motion tracking has scaled to general policies that reproduce diverse human motion, but the resulting behavior is in free-space: the tracker knows where the body should be, not how it should interact with the environment. Other work makes the object part of the objective, yet abstracts that interaction to contact or geometry, which leaves the demanding cases under-explored. We argue that force is a natural command channel between motion imitation and object interaction, one that specifies not just where to make contact but how hard. We present Ctrl-F, a hierarchical motion-force framework that augments motion tracking with an explicit force command, a 3-D force at each designated contact body such as a hand or a foot. The hierarchical framework comprises: (1) A unified motion-force controller tracks the reference motion while applying the commanded force to the world. (2) A high-level force planner that issues a force command online based on object state and proprioception observations. In simulation and on hardware, Ctrl-F completes loco-manipulation that demands large, sustained interaction force and adjusts its command online as the load changes, both out of reach of a force-blind tracker. On hardware, Ctrl-F demonstrates hand-based dragging, pushing, and lifting, as well as foot-based kicking, successfully manipulating boxes weighing up to 20kg in pushing and kicking tasks where the force-blind baseline fails, while adapting online to additional loads introduced mid-task.
Force-controlled loco-manipulation requires a whole-body policy to coordinate locomotion and arm motion while regulating end-effector interaction forces. This is challenging under floating-base dynamics and changing support contacts, particularly when end-effector force/torque sensing is unavailable for control. This paper presents a force-aware reinforcement learning approach with hybrid sensorless force estimation for wheeled-legged loco-manipulation. The proposed method provides a structured estimate of the end-effector force as an explicit policy observation, enabling force-guided contact behavior without using an end-effector force/torque sensor for control. The force estimate is obtained by combining generalized momentum observation, contact-constrained wrench projection, and temporal residual learning: the model-based components extract the physically structured part of the whole-body disturbance, while the residual network compensates the remaining motion-dependent bias. The estimated force is integrated into a mode-conditioned whole-body policy with an axis-wise force/position selector, allowing free-space motion, pure force regulation, and hybrid force/position control within one controller. Simulation results demonstrate improved sensorless force estimation and force-control performance. Hardware experiments further validate the proposed controller through quantitative valve-rotation and hybrid wiping evaluations, together with force-guided door opening and zero-force human-guided motion on a real wheeled-legged platform.
Xuanqi Zeng, Jiaming Wang, Tianlin Zhang +5
Department of Mechanical and Automation Engineering at the Chinese University of Hong Kong, Shatin, Hong Kong