Proceedings of the 3 rd International Conference on Manufacturing Engineering (ICMEN), 1-3 October 2008, Chalkidiki, Greece Edited by Prof. K.-D. Bouzakis, Director of the Laboratory for Machine Tools and Manufacturing Engineering (ΕΕΔΜ), Aristoteles University of Thessaloniki and of the Fraunhofer Project Center Coatings in Manufacturing (PCCM), a joint initiative by Fraunhofer-Gesellschaft and Centre for Research and Technology Hellas, Published by: ΕΕΔΜ and PCCM 475 ROBOT PATH PLANNING OPTIMIZATION, FREE OF COLLISIONS, USING A HYBRID ALGORITHM S. Mitsi 1 , K.-D. Bouzakis 1 , D. Sagris 1 , G. Mansour 1 1. Laboratory for Machine Tools and Manufacturing Engineering, Mechanical Engineering Department, Aristoteles University of Thessaloniki, Hellas ABSTRACT In this paper an algorithm for the optimum collision-free path planning of a spatial robot, using multiple optimization criteria is developed, in the frame of a total inquire of optimum serial robots manipulation. The interface platform is an offline control system that exploits graphical data of the robot parts and workspace obstacles in order to import them in the optimization algorithm. The main optimization criteria are the total travel time from an initial known to a final known configuration through a user-defined amount of intermediate calculated poses, the avoidance of singular configurations and the path smoothness, taking into account the collision avoidance of all robot parts and obstacles, as well as the joint angles limits. The optimization problem is solved through a hybrid method that combines a genetic algorithm, a quasi-Newton algorithm and a constraints handling method, using a multi-objective function and various constraints. The feasibility of the proposed algorithm is verified using an offline control system of a 6-DOF manipulator. KEYWORDS: Robot manipulator, Path planning, Genetic algorithm, Optimization method 1. INTRODUCTION The present paper constitutes a part of a doctoral thesis regarding problems optimization in fields of robotics, based on a hybrid method. The proposed algorithm combines a genetic algo- rithm (GA) with a hill climbing method (quasi-Newton algorithm - QNA) and furthermore a con- straints handling method (CHM) is involved. The developed algorithm uses the above- mentioned methods in order to avoid the disadvantages and exploit the advantages of each in- dividual method. GA applied alone has the advantage of searching the whole space of solutions as well as not being entrapped in a local minimum. Furthermore GA is efficient only for limited number of variables and weakness signs are detected when the number of variables or the total search space rises. QNA, being a gradient-based search method, has the advantage of detect- ing local minimums for higher number of variables, but is strongly depending on the initial vari- ables vector. The synthesis of these methods combines their advantages and detects local minimums in the whole search area. Finally CHM is applied in order to reduce the variables lim- its, reduce the space to be searched and accelerate the whole procedure. The proposed algo- rithm is very efficient in finding the optimal solution in a reduced computation time. In order to test the efficiency of the developed method, it is applied on several fields of robotics. The determination of optimum robot base position and joint angles, considering discrete end- effector positions of spatial 6-DOF /1/ and 5-DOF /2/ manipulators with revolute joints, is ap- proached using this method. Furthermore it is applied to robot design in order to determine the optimum robot geometry, robot base position and joint angles of a 2-DOF spatial RR manipula- tor /3/. The fitness function that quantifies the problem in each case consists of the sum of the deviations squares between the prescribed poses and the real poses of the end-effector, the