Wiping, sanding, and inspection robots press a sliding tool on a surface with controlled force. Controlling that force needs the surface normal, which is unknown on real-world, unmapped surfaces. Surface-normal estimators can help estimate the surface normal, but they either need a friction model or assume that the instantaneous tool velocity lies along the surface; however, walking violates this assumption. We propose SCRUB, a contact-based surface-frame estimator, which lets an armless quadruped slide a body-mounted tool along near-vertical surfaces in the absence of known, axis-aligned, or vision-predicted normals. SCRUB's predictive controller presses and slides the tool using a Kalman-filter estimate of the normal from two cues: 1) A force cue measures the estimate's tilt across the motion. If friction directly opposes the motion, the cue is zero exactly when that tilt is zero, for any friction coefficient. However, if the cue's sliding direction is wrong, friction appears as a false tilt. 2) A velocity cue measures tilt along the motion. Walking and pressing move the tool in and out. SCRUB therefore subtracts the commanded press rate from the tool velocity, averages the result over each gait cycle, and treats this mean as lying along the surface. On Unitree Go1 hardware at nominal speed and load, SCRUB completed every wall traverse and column orbit. No orbit completed with either cue removed, without gait averaging, or with a fixed normal.
Figures & tables
Fig. 1: Hardware and surface frame. Go1 with the trunk-mounted tool (rigid spherical tip in a silicone bumper), inline FT17 force/torque sensor, and surface frame (t^,b^,n^s) : tangent along the slide, binormal across it in the surface (near vertical here), and the adopted outward normal used by the predictive and whole-body controllers. Background removed.
Fig. 2: SCRUB overview. (a) Estimation builds the chart S from force and tool motion. (b) MPC, depth reference, and WBC turn S into joint torques for (c) the robot, which returns contact force and state. Photo in (c): background removed.
Symbol
Meaning
Surface and frames
n
true outward surface normal (toward the robot)
uslip
true sliding direction of the tip on the surface
t^ , b^
slide tangent; binormal b^=n^s×t^
κt
chart curvature along t^ (controller only)
Estimator
TABLE I: Notations used in this work. A hat marks an estimate.
Setting
Symbol
Value
Estimator
Force-row noise
σf,0 , βf
0.3 N, 0.05
Friction bound
μmax
0.8
Stiction bound (start)
μs
0.6
Velocity noise
σv,0
2 mm/s
Load–depth slope, lower bound
kmin
200 N/m
TABLE II: SCRUB implementation settings.
Fig. 3: Estimator cues (tilts exaggerated). (a) Force row, tool sliding out of the page. The true (orange) and estimated (blue) sliding planes are edge-on. Load Fn acts in the page and drag Ft ( ⊗ ) into it, so fe stays in the orange plane. Its component perpendicular to the blue plane, the force-row residual rf=−m^⊤fe , thus measures cross-slide tilt θm . (b) Velocity row, top view. From t0 to t1 , the tool moves by Twvˉc (dashed). Without press motion, this displacement is ideally perpendicular to n at mid-window. The prior n^− deviates from this right angle by along-slide tilt θu .
Fig. 4: Test surfaces. OSB wall (left) and drywall column (right). Arrows: slide direction. Backgrounds removed.
Fig. 5: Example trials. Wall (a)–(e) and column (f)–(j) trials to completion. (a), (f) Tool path, with x into the surface at the start and a dashed display-only fit. (b), (g) Force, with the ±2N band shaded. (c), (h) Speed. (d), (i) Normal error, whose RMSE is for this trial. (e), (j) Curvature. Black lines mark targets, dashed for the true curvature. Force, speed, and normal error use a 0.1s display window.
Metric
OSB wall
Drywall column
Completion and coverage
Completed trials (of 5)
5
5
Falls (of 5)
0
0
Traverse [m]
1.000
1.904
Orbit swept [deg]
–
360.1
Surface and force tracking
TABLE III: SCRUB results (medians of five trials).
Fig. 6: Speed and load sweep. Median vertical-tilt RMSE of five trials per cell. Dashed box: the cell with one fall.
Condition
Done
Falls
Orbit
Normal
Force
In band
Low load
Speed
of 5 ↑
of 5 ↓
deg ↑
deg ↓
N ↓
% ↑
s ↓
mm/s ↓
SCRUB
5
0
360.1
2.48
2.06
81.2
0.00
2.66
Estimator baselines and ablations
Fixed normal
0
0
16.1
30.88
4.94
43.1
0.29
16.9
Friction comp.
0
5
33.0
25.67
4.13
57.5
0.52
18.7
Inst. velocity
0
2
214.0
4.56
2.93
75.3
0.35
10.1
TABLE IV: Column baselines and ablations (medians of five trials).
Robotic surface swabbing requires sustained interaction between a compliant tool and heterogeneous environments, where accurate estimation of tip-level contact force is critical for consistent sampling performance. However, deformable tool dynamics introduce nonlinear viscoelastic hysteresis that decouples wrist-mounted force measurements from true contact forces, while tool-integrated sensors are impractical for deployment due to sterility and disposability constraints. This paper presents a data-driven framework for contact force estimation in Deformable Tool Manipulation (DTM) that leverages proprioceptive sensing without requiring explicit physical models or permanent embedded sensing hardware at the tool tip. A recurrent architecture is first identified through a comparative evaluation of temporal models, where a compact LSTM achieves the lowest estimation error and sub-millisecond inference latency. To address generalization across unseen surfaces and tool compliance conditions, we introduce a parameter-isolated few-shot adaptation strategy that augments a frozen recurrent backbone with low-dimensional context embeddings using feature-wise linear modulation (FiLM). Experiments on a UR5e platform across nine tool-surface interaction regimes demonstrate that the proposed approach significantly improves robustness under domain shift, reducing zero-shot estimation error by up to 63% while preserving baseline performance without catastrophic forgetting. These results show that separating shared deformation-history dynamics from domain-specific conditioning enables reliable force estimation for DTM in non-stationary environments.
Siavash Mahmoudi, Chaitainya Kuppar Reddy, Yang Tian +1
Department of Biological and Agricultural Engineering, University of Arkansas, Fayetteville, AR 72701, USA · Department of Food Science, University of Arkansas, Fayetteville, AR 72701, USA
Payload forces must be accommodated during locomotion, while leash forces can specify desired motion. We investigate whether a shared three-dimensional force estimate in newtons, inferred from proprioceptive history under sustained loading, can support both tasks. An estimator and locomotion policy are jointly trained with supervised force and velocity outputs and learned latent context. The estimated force conditions locomotion and additionally generates planar-velocity and yaw-rate commands for leash guidance through an analytical map. In sustained-force simulation sweeps, temporal means of componentwise force root mean square error range from 1.44 to 2.83,N. Compared with a domain-randomized baseline, the framework reduces velocity-tracking and base-orientation error scores by 21.6% and 46.5%, respectively, and increases mean survival from 68.29% to 94.60% in separate sustained-force tests. Unitree Go1 experiments demonstrate stationary vertical and horizontal force estimation, locomotion with an 8.5,kg payload whose weight exceeds the 70,N training force limit, and leash guidance using the same force-estimation interface.
Run Wang, Xu Yang, Alapati Tuerxun +1
Department of Automation, Tsinghua University, Beijing, China
This paper presents the Kinetics Observer, a novel proprioceptive state estimator for legged robots designed to provide the feedback required for versatile locomotion and physical interaction with the environment, with enhanced robustness to contact slippage. Its core contribution is a tight coupling between whole-body kinematics and external wrenches through a contact dynamics model, enabling the consistent fusion of leg kinematics, IMU, and wrench sensor measurements for real-time joint estimation of centroidal kinematics, contact rest poses, contact wrenches, and disturbance wrenches. By exploiting contact wrench measurements as correction terms, the proposed observer increases sensing redundancy, which provides observability of contact slippage relative to the centroid frame and enhanced robustness to modeling and sensor errors. The approach is experimentally evaluated on two humanoid robots across three scenarios totaling thirteen walking sequences, including long-distance walking with repeated contact changes, locomotion over slippery obstacles, and non-coplanar multicontact motion.
Arnaud Demont, Mehdi Benallegue, Abdelaziz Benallegue +5