S.A.R.A.H.: The Bipedal Robot with Machine Learning Step Decision Making

Abstract

Herein, we describe a custom-made bipedal robot that uses electromagnets for performing movements as opposed to conventional DC motors. The robot uses machine learning to stabilize its self by taking steps. The results of several machine learning techniques for step decision are described. The robot does not use electric motors as actuators. As a result, it makes imprecise movements and is inherently unstable. To maintain stability, it must take steps. Classifiers are required to learn from users about when and which leg to move to maintain stability and locomotion. Classifiers such as Decision tree, Linear/Quadratic Discriminant, Support Vector Machine, K-Nearest Neighbor, and Neural Networks are trained and compared. Their performance/accuracy is noted.