Dr. Sacks is working with Leo Joskowicz of Hebrew University on automating the geometric reasoning involved in mechanism design and analysis. Geometric reasoning is a common bottleneck in computer-aided mechanism design because of its intrinsic complexity and because geometry is easier to visualize than to encode. We are tackling that bottleneck through a combination of computational geometry and knowledge-based programs. We started with the fundamental task, called kinematic analysis, of determining the possible behaviors of a mechanism subject to the constraint that parts cannot overlap or deform. Kinematic analysis is equivalent to the geometric task of partitioning high-dimensional spaces, called configuration spaces, into connected components. The complexity of the geometric reasoning has restricted automation to mechanisms whose parts are permanently connected by standard joints. But our survey of 2500 mechanisms in an engineering encyclopedia shows that most contain non-standard joints and that many involve multiple contacts. Multiple, non-standard contacts also arise in assembly and tolerancing. We have developed a computational multiple contact kinematics and have implemented a robust, efficient kinematic analysis program, called HIPAIR, that covers 90% of the 2500 mechanisms in the survey. HIPAIR supports the design tasks of simulation, functional parametric shape design, and kinematic tolerancing.
CS Annual Report - 19 APR 1996