Optimal design of self-aligning robot gripper jaws.

摘要

Although grippers play a key role in automated manufacturing, they are commonly designed by human trial-and-error. Design is complicated when the initial resting pose of the part differs from the orientation desired for assembly. This dissertation presents a minimalist algorithmic approach to designing grippers that align parts. The idea is to arrange contacts on two parallel jaws so that the part is aligned in the vertical (gravitational) plane as the jaws are closed.; We consider jaws, based on trapezoidal modules, that maximize contact between the gripper and the part at its desired final orientation. Given the n-sided 2D convex projection of an extruded polygonal part, we present an O(k2n 3 log n) numerical algorithm to design such jaws, where k is the number of sample points on each edge of the part. Our algorithm is based on several new functions that represent the mechanical and geometric properties of toppling, jamming, non-liftoff, accessibility, and form-closure.; After computing optimal jaw designs, an O(n log n) algorithm finds a tolerance class for jaws (specified as maximum and minimum material conditions). An Internet-based implementation is available for any Java-enabled web browser at www.ieor.berkeley.edu/∼goldberg/sa-gripper/. We apply this implementation to study circumstances under which jaw designs exist and give formal geometric conditions for a class of triangular parts. We also conduct physical experiments to verify our jaw designs and report sensitivity data.