A Unified Methodology for Motion Planning with Uncertainty for 2D and 3D Two-Link Robot Arm Manipulators

Abstract

Most of the work on robot path planning revolves around two basic models: the model with complete information (often called the "Piano Mover's Problem") and the model with in complete information (called here the "South Pole Search Problem'). The approach of dynamic path planning intro duced in Lumelsky and Stepanov (1987) and Lumelsky (1987) is based on the latter model and produces nonheuristic (provable) algorithms for simple robot arm manipulators operating in an environment with unknown obstacles of arbi trary shapes. The algorithms assume local on-line input information coming from the arm sensors. Unfortunately, the existing techniques require that each kinematic configuration be considered separately, which results in different algo rithms for different kinematics. In this paper, a unified meth odology is presented for designing dynamic path planning algorithms for two- and three-dimensional two-link arm manipulators. The approach is independent of the specifics of the kinematic configuration and imposes no constraints on the shape of the arm links or obstacles in the environment. The methodology exploits some important topological char acteristics of the configuration space. However, no explicit computation of obstacles in the configuration space ever takes place, which results in fast real-time algorithms.