Sensorless damage-safe grasping

Yusei Shuto; Danilo Vasconcellos Vargas

cs.ROAug 25, 2026

Sensorless damage-safe grasping

Authors: Yusei Shuto, Danilo Vasconcellos Vargas

Organizations: Kyushu University, Fukuoka, Japan

Abstract

Robotic fruit harvesting must hold produce securely without bruising it, yet compression stiffness varies several-fold with ripeness within a single species, so no fixed grip force spans the range. Rather than tune force, we bound deformation: a controller closes the gripper until the object's estimated compression strain reaches a user-specified limit $\varepsilon$ , using only the encoder position and motor-effort signal on every servo gripper---no tactile or force-torque sensor. Dividing an effort-based contact force by a lower bound on object stiffness makes the stop provably conservative---true compression stays at or below $\varepsilon$ ---for any $\varepsilon$ above a contact-detection strain floor we identify and quantify: robust detection itself spends compression, linearly in closing speed, making speed an explicit throughput--gentleness knob. Unlike a hand-tuned force threshold, $\varepsilon$ is a certified, size-scaling, operator-interpretable damage limit, and a ready safe-action parameter for learned grasping policies. In MuJoCo simulation over a realistic fruit-stiffness range, under a sensor-noise model calibrated to the real servo, the controller holds $\ge 98\,\%$ grasp at $0\,\%$ damage across all medium-to-firm stiffnesses for the entire certified $\varepsilon$ range, which neither fixed-force baseline attains; on stiffness-graded 3D-printed TPU cubes it matches baseline grasp success at roughly half the grip force and cuts soft-object damage from $100\,\%$ to $40\,\%$ .

Figures & tables

Panel of Figure 1 — Figure 1: Positioning of this work. Prior damage-aware grasping either couples dedicated tactile/vision sensing with heuristic (learned or hand-tuned) damage handling, or applies a fixed hand-tuned force with no damage model; sensorless force estimation supplies the signal but no damage bound. This work occupies the previously empty region: a certified deformation bound from proprioceptive signals alone.

Figure 2: Method pipeline. Encoder position $x$ and motor effort $I$ are read at $50\,\mathrm{Hz}$ ; the estimation block detects contact by amplitude threshold and computes the compression-strain estimate $\hat{r}=\hat{F}/(k_{\min}D)$ ; the control block applies a proportional law that drives $\hat{r}\to\varepsilon$ , producing the velocity command (close or hold). Fruit diameter $D$ and the user-set bound $\varepsilon$ enter as external inputs; the stiffness lower bound $k_{\min}$ sets the strain scale inside the estimate.

Figure 3: Conservative compression-strain bound (§ 3.5 ). The controller divides the current-based force by the stiffness lower bound $k_{\min}$ , so its estimate $\hat{r}$ over-bounds the true strain $r_{\mathrm{true}}$ (which scales with the actual $k_{\mathrm{true}}\geq k_{\min}$ ); stopping when $\hat{r}=\varepsilon$ leaves the true compression at $\varepsilon\,k_{\min}/k_{\mathrm{true}}\leq\varepsilon$ , the gap being the certified conservatism.

Figure 4: Simulation environment: the SO-ARM101 sandwiches a cube (fruit proxy, red) between two large jaw pads in MuJoCo. Cube side $D$ and whole-fruit stiffness $k$ are sampled per episode.

Method	$F$ at term.	Grasp	Damage
Proposed ( $\varepsilon=0.7\,\%$ )	$1.0\,\mathrm{N}$	100 %	0 %
Proposed ( $\varepsilon=1.3\,\%$ )	$1.0\,\mathrm{N}$	98 %	0 %
Proposed ( $\varepsilon=1.4\,\%$ )	$1.0\,\mathrm{N}$	100 %	0 %
Proposed ( $\varepsilon=1.5\,\%$ )	$1.2\,\mathrm{N}$	98 %	0 %
Proposed ( $\varepsilon=2.0\,\%$ )	$3.1\,\mathrm{N}$	100 %	0 %
Baseline aggressive ( $0.10\,\mathrm{A}$ )	$11.4\,\mathrm{N}$	100 %	100 %

Table 1: Headline comparison at

k=4000\,\mathrm{N/m}

(medium-soft tomato, 50 seeds per cell). Per-controller grasp success, damage rate (

r_{\text{hook}}>1.5\,\%

sustained for

\geq 5

ticks), and mean grip force

F

at closing exit.

Method	$k\,(\mathrm{N/m})$
Method	2000	4000	6000	8000	10 000
Proposed ( $\varepsilon=0.3\,\%$ )	00 6 / 00 0	0 64 / 00 0	0 62 / 00 0	0 52 / 00 0	0 38 / 00 0
Proposed ( $\varepsilon=0.7\,\%$ )	00 0 / 00 0	100 / 00 0	100 / 00 0	0 98 / 00 0	100 / 00 0
Proposed ( $\varepsilon=1.3\,\%$ )	0 34 / 00 0	0 98 / 00 0	100 / 00 0	100 / 00 0	100 / 00 0
Proposed ( $\varepsilon=1.4\,\%$ )	0 48 / 00 0	100 / 00 0	100 / 00 0	100 / 00 0	100 / 00 0
Proposed ( $\varepsilon=1.5\,\%$ )	0 40 / 00 6	0 98 / 00 0	100 / 00 0	100 / 00 0	100 / 00 0

Table 2: Sim grid: grasp / damage rate (%) across

(\varepsilon,k)

for the proposed method and the two fixed-force baselines (2000 trials; 50 seeds per cell;

D\in\{35,40,45\}\,\mathrm{mm}

pooled). Bold

=

clean corner (100 / 0); 17 of 30 proposed cells hit it.

Method	Cube	Grasp	Damage	Timeout	Mean $\hat{F}$
Proposed ( $\varepsilon=0.7\,\%$ )	soft	100 %	40 %	0 %	53
Proposed ( $\varepsilon=0.7\,\%$ )	medium	100 %	0 %	0 %	52
Proposed ( $\varepsilon=0.7\,\%$ )	stiff	80 %	0 %	0 %	49
Proposed ( $\varepsilon=1.3\,\%$ )	soft	100 %	80 %	0 %	55
Proposed ( $\varepsilon=1.3\,\%$ )	medium	100 %	0 %	0 %	56
Proposed ( $\varepsilon=1.3\,\%$ )	stiff	100 %	0 %	0 %	60

Table 3: Real-world validation on the SO-ARM101 with TPU cubes.

Figure 5: Representative real-world trial sequences on a soft TPU cube. Top: the proposed method at $\varepsilon=0.7\,\%$ — grasp succeeds without visible damage. Bottom: the fixed-high baseline ( $F_{\mathrm{th}}=25$ load units) — the controller times out and the cube is visibly damaged. Columns: pre-contact, grasp commit, lifted, final.

Figure 6: Measured mean true compression strain across stiffness (solid) against the equality-case prediction $\varepsilon\,k_{\min}/k_{\mathrm{true}}$ (dashed), with detection latency suppressed (reduced speed during the persistence window; simulation). The strain stays below the user bound $\varepsilon$ throughout and follows the prediction’s $1/k$ shape down to the residual ${\approx}0.15\,\%$ threshold term — visible as the $\varepsilon=0.7\,\%$ curve flattening. At the operating speed of the main sweep the latency term dominates instead (Fig. 7 ).

Figure 7: Settled strain at stop versus closing speed ( $k=10^{4}\,\mathrm{N/m}$ , $\varepsilon=0.7\,\%$ , 10 seeds per point, simulation). The detection floor is linear in $v$ with an intercept set by the noise-robust threshold (fit: $0.10\,\%+0.093\,\%\cdot v\,[\mathrm{mm/s}]$ ), matching $\varepsilon_{\mathrm{floor}}$ of § 6.3 : closing speed is the operator’s throughput–gentleness knob. All points remain below the user bound $\varepsilon$ .

Robotic Grasping

Robotic grasping studies how robots perceive, plan, and execute stable, task-appropriate contacts with objects across varied shapes and conditions.

See all 302 papers

Sep 11, 2026

Control Architecture for Safe Grasping of Fragile Objects Using a Coarse Position-Controlled Gripper

Robots are increasingly used in unstructured environments. The need for them to safely grasp unknown objects without damaging them becomes crucial. Humans achieve this by sensing and quickly responding by adjusting their grasping force. Similarly, effective grasp acquisition in robots requires compliant interaction strategies that can adapt to uncertain object properties and adjust to any instabilities during manipulation. We present a geometry-aware force/torque-based contact estimation method for a coarse position-controlled gripper, combined with an adaptive admittance controller for safe grasp acquisition. The desired contact forces are estimated online to keep stable contact with objects of unknown properties. This enables compliant and stable grasps while avoiding excessive forces. Experiments with objects of different sizes, shapes, stiffnesses, and weights show that the proposed algorithm not only prevents slippage but also applies minimal force to safely grasp an object without causing excessive deformation.

Marko Pavlic, Moritz Geier, Timo Markert +1

Machine Vision and Perception Group, Chair of Robotics, Artificial Intelligence and Real-time Systems, Technical University of Munich, Munich, Germany · Resense GmbH, Klingenberg, Germany · Semantic Information Systems Group, Osnabrück University, Osnabrück, Germany

Robotic Grasping Contact-Rich Robotic Manipulation Admittance Control Force-Controlled Robotic Manipulation

Ask PDF

Oct 8, 2026

Tactile Sensing-based Secure and Gentle Grasping of Unknown Objects

Reliable robotic manipulation requires maintaining a firm grasp while strictly avoiding both contact slip and structural damage to the object. Satisfying these inherent secure-and-gentle grasping constraints is challenging for unknown objects. We present a novel online data-driven predictive control framework for safe dynamic robotic grasping of unknown objects. The proposed system features a pair of compact, single-channel tactile-sensing modules that characterize the gripper-object dynamic interaction in real time, entirely bypassing the need for prior knowledge or offline training. A finite-state-machine-based grasping force probing mechanism is developed to establish the initial grasping force and explicitly identify grasping force constraints. Through online dynamics learning, a Gaussian Process (GP) model captures gripper-object interaction dynamics. The model uncertainties are further used to form robust grasping constraints, which are embedded into a model predictive control scheme. We demonstrate error convergence in a probabilistic sense. The proposed control framework runs at 100~Hz in real time to dynamically manipulate and handle objects rather than static holding. Extensive experimental evaluations across diverse contact surface topologies and various types of objects validate superior performance and efficiency compared to baseline strategies. The proposed method is able to manipulate objects ranging from delicate items, such as raw tofu and marshmallows, to rough objects, such as rocks.

Feng Han

Department of Mechanical Engineering, New York Institute of Technology, Old Westbury, NY 11568 USA

Robotic Grasping Tactile Sensing for Robotic Manipulation Model Predictive Control

Ask PDF

Sep 18, 2026

ZeroTouch: Tactile-Supervised Visual Contact Estimation for Contact-Rich Manipulation

Reliable robotic grasping benefits from estimating the evolving physical interaction and selecting a grasp-dependent compression target. Tactile sensors provide direct interaction measurements but require dedicated hardware at deployment. We introduce ZeroTouch, a tactile-supervised framework that predicts dense contact deformation, the instantaneous six-axis wrench, and a grasp-dependent desired compression target from wrist RGB observations, gripper state, and local gravity direction. Tactile measurements are used only as privileged supervision during training and are not required at deployment. On the full validation set, the complete architecture reduces normal-force MAE from 2.017 N for a state-only baseline to 0.531 N. In physical evaluation with 20 trials per condition, ZeroTouch achieves 95% success on an unseen object, 80% in a seen-object/unseen-grasp condition, and 90% under a content/load shift. Under the same evaluation protocol, OpenVLA achieves 25%, 40%, and 55%, while SmolVLA achieves 10%, 25%, and 35%, respectively.

Dmitriy Kosenkov, Daniia Zinniatullina, Miguel Altamirano Cabrera +3

Intelligent Space Robotics Laboratory, Skolkovo Institute of Science and Technology, Moscow, Russian Federation.

Robotic Grasping Contact-Rich Robotic Manipulation Contact Force Estimation Visual-to-Tactile Prediction Robotic Perception Force-Controlled Robotic Manipulation

Ask PDF

Abstract

Robotic Grasping

Explore similar work

Control Architecture for Safe Grasping of Fragile Objects Using a Coarse Position-Controlled Gripper

Tactile Sensing-based Secure and Gentle Grasping of Unknown Objects

ZeroTouch: Tactile-Supervised Visual Contact Estimation for Contact-Rich Manipulation