OxTS have improved the Kalman filter model that the Inertial+ uses with the Novatel OEM-V GPS receiver. The new model improves the drift when GPS is not available by 5% using our standard drift test. The new model is also used in the RT family of products.
Making a model for a GPS receiver is a long task. It takes over 3 weeks to capture the necessary data and even longer to process the data and to extract critical parameters about the GPS. Previous models for the Novatel GPS relied on far less data than this.
OxTS normally quote the position error growth rate, or drift rate, of the Inertial+ as 2.5m per minute or kilometre. This is derived from a standard driving route that we use where the vehicle is driven on straight and twisty sections of road with a variety of speeds. In post-processing we simulate loss of GPS at different sections of the file. Typically 30 different sections, spread over an hour of data, are used to produce the results shown here.
The actual results from our standard test are shown below. These are RMS differences from the reference file, which always had good GPS. They are drift or error figures after 60 seconds without GPS. The Novatel GPS was in RTK mode for this test.[table “55” not found /]
Future research in to GPS receivers will be aimed at modelling each satellite measurement individually, rather than as a whole. Work in this area has already begun and some results should be available by the time the next software release is available.
This work on modelling the satellites individually promises to give much better results for lower performing GPS receivers, like the internal receiver in the Inertial+. It also leads the way to being able to reject poor satellites and preventing them from updating the inertial navigation system. These stray satellites tend to cause larger errors in urban environments.