Robot path planning using genetic algorithms

Abstract

This paper describes the application of genetic algorithms (GAs) for finding a collision-free path for a 3-DOF revolute robot manipulator between specified start and goal configurations among known stationary obstacles. In addition to collision avoidance and staying within specified joint limits, the path can be optimized for minimum distance, time, joint torques, or some combination of these. Attention is focussed on the coding scheme to represent the robot trajectory in a form suitable for the GA, and in the fitness evaluation that is used to drive the GA towards convergence on the optimal path.