Continuous Robot Motion Planning In Unknown Environment

An overview is given of a class of non-heuristic algorithms for planning collision-free motion for robotic systems (such as an autonomous vehicle or a manipulator arm) operating in an environment with obstacles, when any point of the (multi-link) robot body can be subject to collision. In this approach (called Continuous Path Planning, CPP), obstacles are considered to be unknown, and the system has a feedback providing it with information about its immediate surroundings. Such local information is shown to be sufficient to guarantee reaching a global goal, while generating reasonable (if, in general, not optimal) paths. No constraints on the shape of the robot or of the obstacles are imposed. The general idea is to reduce motion planning to the analysis of simple closed curves on the surfaces of appropriate two-dimensional manifolds. The approach is readily compatible with real-time and sensory feedback applications, and thus presents an attractive alternative to the Piano Mover’s approach [1,3] where full information about the environment is assumed. In this paper, versions of CPP algorithms are reviewed and examples are given for the cases of a point automaton and of planar and three-dimensional manipulator arms.