遺伝的アルゴリズムを用いた歩行機構の設計

抄録

Link mechanisms are advantageous for implementing complex motions by a simple structure and input. However, designing the structure of the link mechanisms to accomplish the desired movements through inverse analysis poses a significant challenge. This is primarily due to the inherent nonlinearity of the link mechanism equation. Although many methods have been proposed to address this issue, most of them focus on designing the link mechanisms to perform a desired movement using the trajectory as input. However, for certain mechanisms such as walking mechanisms, it is not always clear what an appropriate trajectory might be. Thus, we are developing the design method that uses the desired "function" as input, rather than a predefined trajectory. This method employs a genetic algorithm in conjunction with a 2D rigid body simulation. By combining the evaluation of the movements on the simulation with the genetic algorithm, it enables the design of the desired link mechanism without the need for a precise target trajectory. We show the result using the proposed method for designing walking mechanisms and discuss potential improvements to our method.