Optimal Design of a Parallel Robot Using Neural Network and Genetic Algorithm

It is well known that parallel robots have greater rigidity, higher payload-to-weight ratio and better dynamic performance than serial robots. However, the complex forward kinematics problem and the limited workspace are the main disadvantages of this type of robots. To design a parallel robot to maximize its workspace we need the robot motion models, thus is a very difficult task. The larger the workspace, the more range of movement is available to perform different tasks. In this paper, by using neural network combined with genetic algorithm we propose an optimal design method for the parallel robot, which maximizes the volume of the workspace of parallel robots. The neural network learns the motion model of the robot, the genetic algorithm uses this model to generate the optimal parameters of the robot. As case of the study, the method developed is applied to the Stewart platform to test the effectiveness and efficiency of the algorithm.