Abstract: The optimization of robot trajectories with genetic algorithms represents a multi-criterion problem because each trajectory consists of many points to be reached with different positions of each joint. The performance of the optimization of a multi-objective problem depends on the coding of the problem and the used algorithm. In this paper first the used coding of a robot arm is presented. Then the algorithm and three possibilities of weighting of the various genes with three different transition functions for each possibility are introduced. In contrast a special technique of selection is presented which performs a manipulation in the case of the selection leading to a speeding up of the whole algorithm. Each of the presented algorithms converge to the Pareto-optimal set by using a special criterion for the end of the optimization.
Key words: Multi-criterion optimization, robot trajectories, Pareto-optimal solution.