Locomotion Classification of Bipedal Humanoid Robot using Fast Fourier Transform

Abstract

A bipedal strolling robot is a kind of humanoid robot. These robots interact with the environment and may encounter external disturbances such as collisions. In this paper, a simple and robust methodology to detect disturbances during unidirectional walking of a humanoid robot is proposed. The procedures incorporate complex deep learning ideas which may require extra equipment, or strategies where various sensors are required bringing about complex multi-sensor information combination. The paper provides two techniques that can be effectively used to classify the state of a robot using existing gyroscope and accelerometer sensors. The first classification approach uses Fast Fourier Transform (FFT). The adopted methodologies allow detection of instability during walking and the experimental results obtained that suggests suitability to effectively classify the motion of robot during walking.