Population generation and validation for the task-based morphology evolution of robotic manipulators

An algorithmic platform is proposed in this paper to randomly generate and validate a set of basic morphologies of planar robotic manipulators, with combinations of serially connected links and loops. To design a robot for specified tasks of following a trajectory and possessing a prescribed stiffness, there are possibilities of working with several hybrid morphologies. Normally, one such basic morphology is selected with experience and the robot is designed for the given tasks. In the present work, basic structures of planar robotic manipulators are generated randomly, while varying the number of links and number of degrees of freedom (dof), within a stipulated number of links. This set will serve as an initial population for the evolution of an optimal mechanism for given tasks, using evolutionary algorithms. A novel concept of Mechanism Assembly Matrix has been introduced to randomly generate planar hybrid manipulators. The proposed algorithm is executed for various number of links and dof. The resultant structures are successfully matching to the possible solutions for the given number of links and the dof. The second phase involves the scrutiny of generated morphologies for validity and retain only the viable structures in the database. A brief discussion is included on the morphology selection during the process of task-based design of a robot.