Detection and Removal of Erroneous GPS Signals Using Angle of Arrival
Type of Degreethesis
MetadataShow full item record
In this thesis, an algorithm for detecting multipath and spoofed GPS signals based on signal angle of arrival (AOA) is developed. As the first step in determining the AOA of the signals, a multi-antenna array of GPS antennas is used to determine the attitude of a test vehicle by calculating precise relative positioning vectors (RPVs) between the antennas. The RPVs are calculated with the real-time kinematic (RTK) positioning algorithm, which allows for an RPV error on the order of a centimeter. The precision of the RPVs allows for sub-degree accuracy of the attitude angles. Abstract A GPS/INS extended Kalman filter is used to propagate the attitude estimate between GPS measurements. The propagated attitude estimate allows for estimation of the RPVs between the antennas without using GPS measurements. To check an incoming set of GPS measurements, an expected AOA with respect to the antennas is computed using the estimated RPVs and the known unit vectors to the GPS satellites. The actual AOAs may be estimated using the incoming GPS measurements and then compared with the expected AOAs. If the difference between the expected and estimated AOA for a signal is not within a specified threshold, the signal may be rejected as a faulty signal. Single-differenced pseudorange measurements and single-differenced carrier phase residuals are explored as alternate metrics for determining faulty signals. Abstract Finally, AOA and the alternate metrics are experimentally tested for their abilities to detect multipath signals and are compared with each other. The single-differenced carrier phase residual proved to be the most reliable multipath detection metric when used with calculated RPVs, having a near perfect success rate with the collected data. Using the carrier phase residual metric with attitude-generated RPVs proved to be less effective with a success rate of approximately 50%. The AOA approach had a similar success rate of about 50%, with worse results in instances of repeated multipath. The single-differenced pseudorange metric proved unable to reliably detect multipath due to the short antenna baselines and relatively large error on pseudorange measurements. The experimental results are validated by comparing positioning solutions before and after rejecting the signals as well as observing the pseudorange measurements of signals detected as multipath.