One-degree-of-freedom planar link leg mechanisms design without predetermined trajectory using a genetic algorithm

Abstract

Legged robots have an advantage when moving on uneven ground and have become a significant focus of extensive studies. The essence of developing a legged robot lies in the selection of effective leg mechanisms. Among the various approaches, leg mechanisms using one-degree-of-freedom link mechanisms stand out for their simpler control. Despite this advantage, the design of link mechanisms presents challenges because of their nonlinear nature. Although various methods have been proposed to surmount this problem, they are unsuitable for link leg mechanism design in terms of their design purpose. In this study, we propose a novel design method using a genetic algorithm to design link leg mechanisms without predefined shapes, topologies and trajectories. Link mechanisms design without predetermined trajectories was implemented by evaluating the movement of mechanisms in a 2D rigid body simulation. In order to accomplish the design of link mechanisms without predetermining the number of links or topology, a decoding technique was devised to create a link mechanism from chromosomes. Our proposed method successfully designed link mechanisms with various shapes, topologies and link numbers. A fabricated link mechanism based on a design demonstrated adequate performance in a real environment. On the other hand, we found a bias in the link number of the design results, which indicates the need for further refinement of our method.