Multiobjective Optimization of 6-dof UPS Parallel Manipulators

International audience; The optimal design of parallel kinematic machines goes through two fundamental stages. The first one concerns the structural synthesis. It enables, a priori, to determine the choice of families of the most adapted architectures according to the desired applications such as flight simulators, machine-tools, medical applications, etc. This can be done by applying several techniques such as: screw theory, Lie groups, graph theory, finite element method, etc. The second one concerns the dimensional synthesis and aims to determine the dimensions of the architecture that has been selected during the structural synthesis. This stage remains a major task because the criteria of performance of a given architecture are strongly dependent on its sizing. In this paper, we present a dimensioning methodology of the architectural parameters of the 6-dof UPS (U: universal joint, P: prismatic joint, and S: ball-and-socket joint) parallel manipulators (the positions of the attachment points of the actuators on the base and mobile plate as well as the radius of the base and the mobile plate). The problem will be formulated as a multiobjective optimization problem (MOOP) by taking into account simultaneously several criteria of performance such as the workspace, kinetostatic performances, stiffness, and dynamic dexterity. The SPEA-II genetic algorithm is adopted to solve this type of MOOP.