Kinematic Model Pruning: A Design Optimization Technique for Simultaneous Optimization of Topology and Geometry

This paper presents a method of optimizing the design of robotic manipulators using a novel kinematic model pruning technique. The optimization departs from an predefined candidate linkage consisting of a initial topology and geometry. It allows simultaneously optimizing the degree of freedom, the link lengths and other kinematic or dynamic performance criteria, while enabling the manipulator to follow the desired end-effector position and avoid collisions with the environment or itself. Current methods for design optimization rely on dedicated and complex frameworks, and solve the design optimization only as decoupled from each other in separate optimization problems. The proposed method only requires the introduction of a simple function, called a pruning function, as an objective function of an optimization problem. The introduced pruning function transforms a discrete topology optimization problem into a continuous problem that then can be solved simultaneously with other continuous objectives, using readily available optimization schemes. Two applications are presented: the optimization of a manipulator for the inspection of radio frequency cavities and a manipulator for maintenance within the future circular collider (FCC).