Analysis of improvement to two-wheel robot navigation using low-cost GPS/INS aids

Abstract

This paper shows the use of sensor fusion (GPS/INS/odometry) to estimate and eliminate errors which are unobservable for a two-wheeled robot using odometry information alone, including the accuracy benefit that combinations of these sensors provide. The combinations examined are GPS/odometry and INS/odometry, both of which are low-cost options. It is shown that the effects of longitudinal wheel slip and tire radius error can be lumped together as an ‘effective radius’ and estimated effectively using GPS velocity and course measurements. This estimated radius can account for these error modes at the speed at which it was estimated but the estimation effectiveness degrades as speed changes. This paper shows the improvement to the navigation of a two-wheel robot platform using this estimation method.