Computational Design of Closed Linkages For Robotic Limbs

Abstract

Legged robots require low-inertia limbs capable of carrying a high payload. The design of such limbs poses challenges in integrating optimal kinematic structures with practical design considerations. In the search for optimal design parameters, advantages of optimization algorithms can be applied. This paper introduces an open-source framework for optimizing topology and parameters of closed linkage mechanisms, addressing the need for task-specific robotic limbs. Closed-loop structures are motivated by two main purposes: (1) to decrease robotic limb inertia by relocation of actuators close to the robot’s body and (2) to redistribute efforts among actuators. The framework leverages joint-based spatial graph representations, kinetostatic criteria, and multi-objective genetic algorithms to optimize mechanism topology and parameters. Focusing on kinetostatic criteria such as Jacobian metrics and inertia properties, the framework swiftly explores the design space to balance trade-offs in robot linkages. We demonstrate the framework pipeline for the task of optimizing planar robotic legs with 2 degrees of freedom. Project github page: https://licaibeerlab.github.io/jmoves.github.io/