Fault Detection and Exclusion in Deeply Integrated GPS/INS Navigation
Type of Degreedissertation
MetadataShow full item record
This dissertation presents a novel fault detection and exclusion method in a centralized deeply integrated GPS/INS navigation system. The method presented is also demonstrated in a centralized vector tracking GPS receiver. Also, a new multipath error model and a range variance parameter are developed to better deal with the additional challenges faced by vector-based receivers. These methods and analysis extend the field of robust navigation, particularly with regards to advanced tracking architectures. GPS was originally designed to operate in clear line of sight view to the supporting satellite constellation. However, recent advances in receiver hardware and computing power have pushed positioning into more difficult scenarios. New ways of using radionavigation sensors, such as vector tracking and deeply integrated GPS/INS, have been developed to handle these situations. However, integrity has been difficult to maintain in these configurations. Fault detection and exclusion can provide this need by keeping the navigation solution free from erroneous measurements that occur in difficult environments. This dissertation presents these contributions in four stages. In the first, a multipath model for vector tracking is developed. This model is based on a delayed signal's interaction with the direct signal correlation peak rather than the scalar early and late correlator outputs. To demonstrate both the model and improved performance of vector tracking, simulated results show the vector receiver tracking the signal with 0.015 m less range error. Experimental results show vector tracking performing better in multipath environments by several meters. Second, the range variance parameter is derived as a means to monitor a vector receiver's tracking situation. Since traditional lock detection does not apply directly to a vector receiver, another approach is needed. The range variance gives an indication of how the receiver's position uncertainty would translate to range uncertainty. This is done in a geometry-free way so the receiver can determine the maximum impact its error would have. The variance parameter is demonstrated in an environment with significant blockage to show its response to the tracking situation. Third, fault detection and exclusion are applied to a centralized vector tracking architecture. This integrity method is based on the normalized innovation test parameter. When new measurements are provided to the navigation filter, they are normalized by their expected variances. Faulty signals are shown to increase this test parameter and pass the detection threshold. Live sky demonstrations of fault detection and exclusion yield position improvements on the order of one meter. Lastly, a deeply integrated GPS/INS algorithm is presented. The vector fault detection and exclusion method also applies to this fused navigation system. After dealing with IMU synchronization, results are presented in which an automotive grade IMU is integrated with a vector software receiver. Using the fault detection and exclusion method, positioning performance is improved by several meters. In total, this dissertation demonstrates two navigation methods: the GPS vector receiver and the Deeply Integrated GPS/INS system. Both of these methods are made more robust by performing fault detection and exclusion. This method is shown to remove velocity drifts and position jumps due to signal errors in difficult scenarios. The result is a highly robust navigation system for continuous positioning in GPS degraded environments.