Development of a Neural Network Method to Estimate the Workspace of a Parallel Robot

In order to perform complex industrial operations, robots have to respect certain criterions as specific dimensions of the elements or of the workspace, kinematic or dynamic performance, etc. This is achieved through a proper design of the robots, which is a cumbersome task that usually requires a large amount of knowledge and practice of the human designer. This is why, several methods have been developed that assist the human designer in the process of the design of robots. The scientific problem addressed in this paper is the development of a machine learning method, a feedforward neural network regression, to estimate the volume of the workspace of a robot based on the parameters that describe the architecture of the robot. The method is implemented on a 6 degrees of freedom parallel robot with rotary actuators.