| A GNSS Rx baseband has two functions: acquisition and tracking, which isbasically a detection and estimation problem of statistical signal respectively. Inrecent years, vector tracking attracts far more attention than other issues in GNSSreceiver due to its promising advantages in weak signal processing, interferencemitigation and potential for migration into a INS/GNSS deep integration system. Theword vector, as it implies, means there are more than one state to estimate in thesignal estimation (tracking) problem and hence when it comes to implementation, thestate is a vector (e.g. a vector of positions, velocities and clock parameters of thevehicle) rather than a single variable (e.g. code phase or carrier phase in a traditionalscalar receiver) in the filtering mechanization. Two questions remain totallyunresolved in this field: first is lack of unified theorem explaining the superiorities ofvector loop over scalar loop and second is the real-time implementation of a vectorsoftware receiver. This dissertation is dedicated to solving the first problem and tryingto shed some light on the second one.The thesis focus on the theory and implememtation of vector tracking loop in aGNSS receiver. In detail, it includes the following three aspects:(1) Performance metric. Obviously we need a performance metric to prove theeffectiveness of our filter, but existing ones are inappropriate to use. As a result, thisthesis proposes a unique method–tracking loss in meters to evaluate front-endparameters’ effects on the variance of the output of a code-tracking loop. Suchparameters are front-end bandwidth, number of bits used in quantization and samplingfrequency. The metric gives a single indicator instead of three separate values, toindicate how different combinations of these three parameters will degrade theprecision of code tracking, as opposed to the ideal situation where positive infinitequantization bit, sampling frequency and front-end bandiwth are assumed. This metricis used to determine the frequency-plan of the real-time SW receiver, described inChapter7in this thesis.(2) Theory and mechanization of a vector tracking loop. Take vector delay-lockloop (VDLL) as an example. VDLL is a further generalization of the extended kalman filter (EKF). VDLL closes the loop all the way back to the signal correlators insteadof having two separate sets of shorter loops (delay lock loops and the EKF loop).Transition from a scalar receiver to its vector version nessecitates examination oftheory and mechanization of a VTL. Kalman filter parameters such as process andmeasurement covariance matrices have to be taken with extreme care. The secondsubtask is to explain the superiority of a vector receiver over a scalar receiver usingcovariance analysis.Vector tracking loop (VTL) including VDLL (Vector DLL), VPLL (Vector PLL)and VFLL (Vector FLL) is one of the most promising architectures fornext-generation Global Navigation Satellite System (GNSS) receiver due to itsrobustness during unplanned GNSS satellite outages and in adverse environments (e.g.Radio Frequency Interference (RFI) or user high dynamics), improved accuracy ofnavigation solutions compared with that of traditional receivers using scalar loopsunder regular conditions, and most important of all, easy integration with inertialNavigation System (INS) measurements to form an ultra-tight INS/GNSS integratedsystem. In the current State High Technology Project (or863), a real-timehigh-sensitivity receiver with Interference Detection and Mitigation (IDM) capabilityis required to form the core of the GNSS Vulnerability Assessment and ValidationPlatform (VAVE). As a result, an ultra-tightly coupled GNSS/INS structure wasdecided on and as a natural first step, a software vector GNSS receiver has to berealized for research and fast-prototyping.There has long been interest in theoretical analysis and implementation of VTL.Nonetheless, previous efforts to implement the vector tracking structure often focuson either post-processing data collected from a front-end or offering limited view ondetails of real-time implementation. In addition, system parameters of the VTLimplementation are often chosen in an empirical rather than an analytical way. Thispaper tries to address these two issues by presenting a panoramic view on design andreal-time implementation of an open-source Global Navigation Satellite System(GNSS) software receiver restructured to use vector delay-lock loop for signaltracking. Drive test results on navigation solutions under stringent conditions areprovided to prove the performance margin over traditional scalar loop.This particular part is structured as follows. The second sub-section defines theproblem being analyzed, introduces the notation and assumptions that are employed,and develop the measurement and system model for the extended Kalman filter (EKF)structure of VTL. However, for real-time employment of the equations, this configuration has to be slightly restructured. Therefore a cascaded (or federated) EKFis developed and described.The third subsection describes the implementation of a post-processing VTLreceiver, which has been developed as a research and teaching tool in university lab toprovide useful insights in parameter tuning, especially the determination of the Q andR matrix (i.e. the covariance matrix of system noise due to user dynamics andnon-modeling error, and the covariance matrix of measurement noise represented bydiscriminator output noise and due to the nonlinearities of receiver such as the jointeffects of precorrelation filtering, sampling and quantization) of the Kalman filter.(3) ImplementationFirst we give the details of the real-time implementation of VTL in a softwarereceiver. A widely-used USB FE, GN3sV2, designed by CCAR of University ofColorado, Boulder is used to provide digital Intermediate Frequency (IF) samples tothe receiver. This FE has sufficient bandwidth for GPS L1C/A and Galileo E1B/C(only BOC(1,1)). Since originally, there is an upper limit to the data grabbed by thefront-end, it has to be flashed to lift off this limit for continuous operation of thesoftware receiver. The VTL software receiver is a modified version of an open-sourcereceiver, GPS-SDR, which is widely used in researches on GNSS signal-processing.The architecture of the VTL receiver is exposed by using a state flow chart, where theinitialization of the VDLL and the definition of measurement epoch and integrationepoch are emphasized. The system parameters are also provided and threshold for lossof lock in the context of VDLL is established. Later this GN3sV2FE will be replacedby a USRP2to explore SCA.Next we present the navigation performance by comparing the vector trackingreceiver with its scalar tracking counterpart in drive test. The results confirm that thatthe former is much superior to the latter in terms of robustness in poor satellitevisibility and noise in the navigation solutions.The last subsection summarizes the paper and discusses the framework for futurework, especially its promising prospects in UAS (Unmanned Aerial System)All in all, a VTL is the very starting point for INS/GPS deep integration. This isthe ultimate goal of the work described in this thesis–to provide theorectical andexperimental results for future efforts to implement a lab deep integration prototype. |
PDF Full Text Request