Motion Planning for Robotic Manipulators

Abstract

In this thesis motion planning for robotic manipulators is considered. A mathematical model describing the kinematics and dynamics of a manipulator based on a bounding box model of ABB's IRB140 is developed. Variants of rapidly exploring random trees (RRTs) are implemented and compared to show that it is in general better to maintain trees at both the initial and final vertex as opposed to only at the initial vertex. Although, in most cases, connecting longer paths results in less overall running time than connecting shorter path segments, when the density of obstacles increase all variants struggle and perform similarly. A method for optimizing paths based on RRTs is presented with a sketch of proof that under some conditions this method guarantees to find the path of smallest cost according to some criteria. The resultant path is used as a foundation for optimal trajectory specification which is subsequently used as a reference for a control system which in addition to tracking the desired trajectory remains robust in the presence of unmodeled dynamics which is shown by simulating the system.