Reduction of GPS Noise for Precision Control of Robot Navigation in Confined Areas
Most modern navigation systems solve the localization problem by extensively using global positioning system (GPS) data. Unfortunately the GPS data quality depends on several factors, which can lead to large positioning errors. Known GPS errors fall in two groups: either atmospheric errors, or multipath errors. Because of these errors, differential GPS systems have been developed using both ground based and satellite based reference systems. The cost of a differential GPS unit such as a Novatel range from a little over $2000 to over $9000, which can be prohibitive for use on certain home service robotic vehicles such as autonomous snow plows or autonomous lawn mowers. This paper discusses ways of mitigating GPS errors on low cost single frequency GPS units such as Copernicus II, Skytraq and U-Blox, which cost far less than $100 each, and hence are attractive for use in many robotic applications such as those mentioned above. The paper will present a model of GPS noise and use that model to process GPS data for use in navigation of an autonomous snow plow designed for use in residential driveways and side-walks; it will be supported by experimental results only.