Performance between Algorithm and micro Genetic Algorithm to solve the robot locomotion

Abstract

Gait robot refers to the locomotion platform coordinating the leg motions. Several studies are using Genetic Algorithms (GA) to solve the problem of gait learning. These studies coincide with the high computational cost and high energy consumption; this is mainly due to the handling of large numbers of individuals compromising the memory space in the system. To solve this problem, and with the intention of being implemented in hardware, it is proposed the use of micro Genetic Algorithms (μGA) that use populations of reduced size in this research twelve individuals. This article presents a comparison of the performance of both algorithms, μAG versus the standard GA, solving the problem of the generation of gait patterns for a quadruped robot. It demonstrated that the μAG has a rapid convergence to the solution and generate the robot locomotion. The implementation of the algorithms have performed in an embedded system with four cores to validate the results.