Haptic Probing for Subsurface-Aware Force Distribution on Layered Soft Tissue

TL;DR Before pressing on soft tissue, a robot hand gives it a brief low-amplitude vibration to tell muscle from bone-backed tissue, then spreads the force over its tactile fingertips under a pressure limit set by that class. Kneading the shoulders of three people with a dual-arm robot, it cut median force-tracking error to 2.22 N from 4.04 N for a fixed-compliance baseline.
What they do
Robots that massage, scan or retract soft tissue have to hold a set force while the tissue underneath changes from muscle to bone-backed. Usual force controllers infer the substrate from the loading signal, so they adapt only after loading begins. This method senses the tissue type first with a small haptic probe, then uses it to decide how much pressure the fingertips may apply and how much contact area the hand must recruit.
How it works
- Haptic probe. Under a light preload, a brief low-amplitude excitation measures the local dynamic admittance (peak frequency, damping, response magnitude). The tissue is labeled bone or muscle by whichever of two references, recorded on the same subject at anatomical landmarks, it is nearer to.
- Regime-conditioned force distribution. The label sets the impedance and a maximum pressure. An allocation step equalizes pressure across the engaged fingertips using contact areas from their tactile arrays, and recruits another finger when the area is too small to carry the force under that pressure limit.
- Re-probing and safety. At a detected boundary the hand unloads, probes again and commits new settings. Probes, switches and finger recruitment are paid from a partitioned, non-refilling passivity energy budget.
Results
Pressing and kneading the shoulders of three subjects with a dual-arm robot and three-fingered hands, compared with a fixed-compliance baseline and a reactive (agnostic) controller:
- Median force-tracking error was 2.22 N, versus 3.37 N (agnostic) and 4.04 N (baseline), a 45% reduction on the baseline.
- Where the bone pressure bound binds, pressure uniformity was 1.161 versus 1.252 and 1.380, and 95th-percentile peak pressure was 62 kPa versus 68 and 228 kPa.
- Above 5 N the hand held 182 mm² of contact versus 163 mm² for the baseline, about 12% more.
Limitations
Boundary detection is still reactive, so each new region is briefly loaded with the previous settings. The finite budget caps recruitments per traverse, tissue is split into only two classes, and peak taxel pressure is not regulated. Tests cover three subjects, the tactile array on one hand clipped, and the paper does not report the probe’s classification accuracy on its own.
Figures



Abstract
Force-regulated contact with layered soft tissue fails most often where the subsurface composition changes: controllers infer the substrate from the loading signal alone, so the estimate guiding them lags the change---more acutely, we argue, over compliant than over bone-backed tissue. We remove that lag with a separate haptic channel. Under a light preload, a brief low-amplitude excitation lets the local dynamic admittance discriminate muscle from bone-backed tissue relative to a per-subject baseline, before task force is applied---active haptic sensing used to commit a control decision rather than to characterise tissue. A force- and tactile-driven distribution law then consumes that regime throughout execution: an instantaneous allocation equalises pressure across the engaged fingertips, while the regime sets the admissible pressure and hence the area the hand must recruit to deliver the target force---making patch area a controlled variable rather than a by-product. Probing, switches and recruitment are priced in a partitioned passivity budget, so safety monitoring cannot suppress the probe and a refused recruitment forfeits force tracking rather than the pressure bound. We report force tracking and pressure uniformity over curved, migrating contact on three human subjects, with regime labels from anatomical landmarks.
Summaries are written automatically from each paper's text. Figures and videos come from the paper and its project page; all credit goes to the authors.
