Shortest non-synchronized motions parallel versions for shared memory crew models

Abstract

This papers treats parallel versions of the path finding problem for robots whose joints cannot be controlled in such a way that the endeffector follows a prespecified trajectory. This means, if two or more joints are moving at the same time, the relative positions at each time moment for the joints, i.e. the exact positions of the endeffector, are not known. This may be due to the low level control of the robot (for example, with heavy load robots), or due to a complicated kinematics structure. For such mechanisms a motion is specified by certain intermediate positions (values for all joints) along a desired path. These intermediate positions (“synchronization points”) and the requirement that the motions in the single joints are monotonous between consecutive synchronization points guarantee a certain structure of a path. There are several possibilities for parallelizing such algorithms: one may apply known parallel graph search techniques or develop a new concept. We discuss known approaches and argue why they are not really appropriate for our algorithm. New concepts for parallelizing the algorithm are discussed.

Supported by the Austrian Ministery for Science and Research.