On the sensor fusion in the walking robots design

Abstract

The aim of this paper is to provide a sensor fusion for a real laboratory model of the underactuated two degrees of freedom walking robot. More precisely, the laboratory model consists from two pairs of legs with the actuated knees connected in the actuated hip. The angle between the stance leg and the walking surface is both unactuated and insusceptible to direct measurement by precise encoders. In such a way, the convenient fusion of different sensors plays cornerstone role in underactuated walking robots control design. The each link of the leg is equipped with sensors for DC motor input current and its angular position and velocity measurement. The data from sensors are fused together via Extended Kalman Filter in order to eliminate measurement noise and increase accuracy of the measurement. The fused measured data can be used as input data for the tracking or estimating algorithms for the robot walking control or for leg link movement control via a DC motor.