Differential GPS Control of Starcar 2

Abstract

This paper assesses the ability of GPS to guide a robotic vehicle with precision. To secure accurate positional estimates, we use differential GPS positions and Doppler velocities, combined in a filter that exploits the high accuracy of velocity data. A model of vehicle dynamics overcomes a time lag of GPS fixes, which amounts to 1.5 s even for an advanced 6-channel sensor. The system attains positional accuracies around 0.5 m along each horizontal axis with no changes of satellites and 2.4 m over long periods. The system is able to steer a car around a 400 m track so accurately that the laps merge into common pixels on a video display.