Robot motion planning

摘要

Manipulation and motion are the most common means that we use to act directly on the world. Vision and language serve us as inputs and we communicate with other human beings using speech, but we act on our surroundings by moving and manipulating. If AI is to deal with real life problems, autonomous intelligent systems will have to interact with the world in the way humans do. This tutorial is intended to present what robot motion planning is, what has been achieved through it so far, and what could be reasonably expected from it in the near future. The attention will be focused more on techniques for real-life applications than on theoretical formulations. Emphasis is on robot manipulators rather than on isolated moving objects, distinguising applications in 2D and 3D. The tutorial is organized on a performance basis instead of the usual methodological clasification. Three levels of performance are distinguishedr(a) Geometric theoretical algorithms, (b) approaches that work at a computer simulation level and (c) techniques that have been implemented on actual robots, or which deal with problems for real-life robots. Some of the covered topics are: collision detection, fundamentals of the problem, geometric algorithms, basic motion planning techniques, moving obstacles, multiple robot coordination, nonholonomic motion, planning with uncertainty and future directions. Throughout the tutorial many examples and case studies will serve to illustrate successful applications as well as their underlying theoretical techniques. Transactions on Information and Communications Technologies vol 1, © 1993 WIT Press, www.witpress.com, ISSN 1743-3517