Grasp quality is a score that rates how good a grip is — not just whether it holds, but how much disturbance it can survive — letting a robot rank candidate grasps and pick the most robust one.
Grasp quality is a number rating how strong a grip is. Two grips might both hold an object, but one can take a much bigger bump before slipping — grasp quality captures that difference so the robot picks the sturdier grip.
A robot often finds many grips that would technically hold an object. Which should it use? The one least likely to fail. Grasp quality puts a number on that, so the robot can choose.
From "holds" to "how well"
Force closure answers a yes/no question: can this grasp resist any disturbance at all? But two force-closure grasps aren't equal — one might barely hold while another could survive a hard shove. A grasp-quality metric measures how strongly a grasp holds, turning ranking into math.
Score every candidate, pick the best
Quality reduces each grasp to a comparable number, so the planner isn't guessing — it selects the most robust grip it can reach.
What the score captures
Common grasp-quality ideas:
Worst-case wrench resistance (the ε-metric). The classic measure: the largest disturbance the grasp can resist in its weakest direction. A bigger number means the grip is far from slipping no matter how the object is pushed.
Contact geometry. How well-spread and opposed the contacts are — grips that surround the object's center of mass and press along good directions score higher (an antipodal grasp is a simple high-quality pattern).
Robustness to uncertainty. How much the grasp still works if the object's pose or the contacts are slightly off — crucial because perception is never perfect.
In modern systems
Analytic planners compute these metrics explicitly. Learned systems (visual grasp detection) instead predict a grasp success probability directly from images — effectively a learned quality score — trained on millions of real or simulated grasp attempts (as in Dex-Net and similar). Either way, the robot ends up ranking grasps and choosing the top one.
Why it matters
Grasp quality is what lets a robot be reliable, not just occasionally lucky — choosing grips that tolerate real-world bumps and sensing error. It's the deciding metric in grasp planning and a core idea in making manipulation dependable.
