Lab 13 / 15 · Interactive · real-time
Grasp ε Evaluation Bench
Move fingers, change friction, and watch grasp quality ε shift — and when it hits zero.
Grasp parameters
What to look at
Gray region in the right plot = the set of wrenches contact forces can combine. Origin inside = force closure; dashed circle = inscribed ball ε. The two-finger 120° configuration drives ε to zero when μ is small.
Web model: linearized cone (4 rays) + per-finger force cap of 1, ε = inscribed ball radius of the wrench set (160-direction enumeration). Same model as T13's live run; T13's per-direction-distance metric differs slightly numerically (two-finger opposing 0.49 vs the web's inscribed 0.52) — conclusions identical.
What problem this lab solves
«A stable grasp» can be computed as one number: the radius ε of the largest ball inscribed in the wrench set the contact forces can produce (T13). It is a min, not an avg — the worst direction decides everything. This bench runs T13's actual LP in your browser: each direction's capability is attained at a vertex, so enumeration is exact.
How to read the figure
The dots on the disc are fingertips; the small sectors are each finger's friction cone (half-angle arctan μ). The gray region is the projection of the wrench set W onto the torque-force plane — force closure = the origin inside it, and ε = the inscribed ball radius (the circle in the figure). The worst-direction arrow points at the spot on W's boundary closest to the origin.
Hands-on
Pick the «three fingers at 120°» configuration and drag finger 1's perturbation from 0° to 40°: ε falls smoothly but never reaches zero (the normal still passes through the center). Switch to «two fingers» and drag the angular separation past 120°: watch when ε crashes to 0 (criterion: arctan μ < the angle between chord and normal). Finally fix the configuration and drag μ from 0.2 to 1.0 — material (coating, texture) is a first-class citizen of grasp performance.
Done playing — where next
Related topics
Next labs (in play-learn order)
Rolling Holonomy Lab
The ball rolls back to its start, the orientation doesn't — roll one loop and see how far it drifts, and what sets the drift.
6 min
Friction Cone Slicing Bench
Drag the direction angle and wedge count: see how much tangential capability gets sliced off when the cone becomes a polytope.
5 min
Slip Precursor Detection Bench
Tune the threshold and trade detection rate against false alarms — the precursor lives in the amplitude, not the derivative.
6 min