In the method of artificial potentials, the robot behaves like a charged particle that is attracted by the goal position and repelled by the obstacles. An overview of the method is presented which describes the common variations in a unified framework, compares the performances of the different algorithms, and corrects some common misunderstandings. It is shown that force control algorithms with generalized potentials (which depend on velocity as well as position) offer advantages over other alternatives. Also discussed are the problems of stability and convergence.>

In this presentation we shall discuss some of the ways in which CAD-style object descriptions are currently being used to facilitate the planning, programming, and verification of robotic applications, and also discuss some open issues that are the subject of ongoing research. Computer-based representations of parts, tools, fixtures and so forth are used today in the context of robot programming primarily as a source of data for solving, either automatically or with human assistance, some of the geometric and kinematic problems that arise in robotics. Some examples that will be discussed include the automatic planning of collision free motions, robot work cell layout, graphical simulation of robot motions, and the specification of initial values for certain parameters in a robot program. In each case, the problem is modeled and analyzed in terms of geometric objects of perfectly known shape and size with precise constraints on the relative positions and orientations of objects. A number of difficult problems remain, however, in CAD-based robot programming. These include: the representation and analysis of tolerancing specifications and other forms of uncertainty; modeling, analysis, and verification of robot programs incorporating the use of sensors; the use of geometric models to represent task constraints in compliant motion control; automatic updating of robot programs upon modification of part design data; and many more.

Many practically important problems in computational geometry may be regarded as a generalization of "clipping," and may be formulated precisely in terms of a function called "membership classification." This function operates on a pair of point sets called the reference and candidate sets; it segments the candidate into three subsets which are "inside," "outside," and "on the boundary of" the reference. Examples of classification problems include clipping, polygon intersection, point inclusion, and solid interference.

PADL (Part and Assembly Description Language) is a language for defining solid objects via constructive solid geometry, i.e. as (regularized) set-theoretical compositions of primitive solid “building blocks”. The current processor for Version 1.0 of the language accepts PADL statements either in batch or in keyboard-interactive mode and produces line drawings of objects on CRT's and plotting devices. A variety of drawing styles is available: dimensioned orthographics, perspectives, section views, interference drawings, etc.