Adding Intelligence To Robot arm Path Planning Using A Graph-match Analogical Reasoning System .
Diane J. Cook · 2005
Analogical plaiining provides a. nieaiis of solving problems in some situations where other ma.chine lea,riiing methods fail, because it, does not, require iiuinerous previous examples or a rich doiiia.in theory. Given a problem in an unfamiliar doiimiu (tlie target ca.se) ~ an analogical planning system locates a. succ fill 11la.n in a similar domain (the base case), a.nd uses the siinihrities t.0 generate the target plan. One planning doinain which can ma.ke use of analogical planning to improve performance is robot arm path planning. In the Intelligent, Mechanisms Group at NASA Ames Research Center, algorithms are designed to guide a Puma. robot arm with 7 degrees of freedom through a 3dimensional workspace. The search spa.ce is immense, a.nd current inotion planning techniques do not t,alie advant,age of the fact that many planning situatioiis are similar. This paper describes a method of using analogies to improve the performance of robot arm path planning. C' veil an enviroiiiiient description, tjhe system looks for a siinilar previous case with a. successful plan. Thatm plan is ina.pped to the current problem environment^ tmo suggest a. possible path for the robot. An example of how the robot a.rm at, NASA Anies can iimke use of amlogies to generate plans is detailed.