ION GNSS 2012
Session B1: GNSS Simulation, Testing & Timing Applications 1

Title: Tracking Multipath in Received GNSS Signals through use of a Signal Decomposition and Parameterization Algorithm
Author(s): M.E. Haker, and J.F. Raquet, Air Force Institute of Technology
Date/Time: Wednesday, September 19, 2012, 11:03 a.m.
Room: 103/104 (NCC)

This paper outlines results of research into the construction of a high-fidelity GNSS signal simulator. This signal simulator makes use of outputs obtained using the previously published Signal Decomposition and Parameterization Algorithm (SDPA). The SDPA is used to obtain parameters from received GNSS signals, in order to discern what can be expected from a real world GNSS signal environment. Previously, only the algorithm framework and initial performance findings have been published.
Robust GNSS-based navigation in challenged environments (such as urban or indoor settings) is desired. In order to develop algorithms to enable robust navigation, high-fidelity modeling and simulation of the GNSS radionavigation channel is required. This is because the effectiveness of GNSS receiver signal processing algorithms must be verified. This verification requires the availability of reasonable channel models; preferably obtained using results that mirror what would be expected from real world signals.

The SDPA processes uses a signal ray model in order to process received and simulated GNSS signals. More specifically, the SDPA decomposes received data, using a global stochastic search optimization technique (simulated annealing) to simultaneously fit a GNSS signal ray model to rays suspected to be present within a single integration period of received or simulated GNSS signal data. The nominal sequence of the decomposition process includes the recording of GNSS signals in a challenged setting, the decomposition and parameterization of the recorded data in post-processing to obtain ray parameters, and the application of these ray parameters in a tracking algorithm to obtain time-varying models for ray parameters. These ray models are then used to construct a signal simulator that can output generated data sets that reflect what can be expected from real world GNSS signal environments.

The SDPA works by decomposing and parameterizing the direct path and multipath rays within an integration period to obtain four parameters for each ray: amplitude, Doppler frequency, propagation delay, and carrier phase. This is done by performing SDPA processing on an integration period-by-integration period basis. Processing involves iteratively computing an estimate search space (also known as the ambiguity function, where the data within an integration period is transformed to a space with dimensions including amplitude, Doppler frequency offset from an intermediate frequency, propagation delay relative to the suspected direct path ray propagation delay, and phase) and comparing the received and estimate search spaces. Differences between the two search spaces are resolved through the addition of ray estimates. As more ray parameters are obtained, more ray estimates contribute to the estimate search space, and the RMS error between the received and estimate search spaces is reduced. Iterations continue until the decomposition processing stopping criteria has been reached. Once this happens, the next integration period-sized data vector is then input for processing, and the algorithm continues until all data is processed. The SDPA is designed to obtain ray parameters similarly to what would be expected from a Multipath Estimating Delay Lock Loop (MEDLL) receiver, except that the SDPA itself does not make use of results from previously processed integration periods in a tracking loop, as takes place with MEDLL.

Novel research findings presented in this paper will include results of modification of the SDPA to simultaneously estimate parameters from multiple rays present within an integration period ("multiray parameterization"), enabling explicit estimation of received direct path and multipath rays in the presence of noise that present constructive or destructive interference. Multiray parameterization results from simulations and real world received GNSS data will be presented. Previously published research findings only presented results from estimates obtained using sequential estimation of ray parameters (single ray parameterization). Additionally, an algorithm to determine stopping critieria has been developed and will be presented.

New research findings in this paper will also include a framework for the tracking of ray parameters over multiple integration periods (on the order of seconds) using multiray parameterization. Tracking is permitted through use of the output of ray parameters from the SDPA. These tracking results will also be published in this paper. Tracking performance expectations will also be presented in this paper as well. Between multiray parameterization within an integration period and the subsequent tracking of ray parameters over time, GNSS ray parameter models are obtained which can then be used in the construction of a software signal data generator that simulates received GNSS signals with an expectation of high fidelity in replicating what can be expected from a real world environment.

The significance of this work is that, through use of the SDPA, a means is established by which real world data can be characterized using a set of parameter values that can then be used in synthesizing a model for what can be expected from a setting of interest. In short, if data is recorded from an environment, the data can be used in the tailoring of a simulator that outputs an infinite amount of generated data that reflects the parameters that could be expected from data that has been obtained using a recording device. This technique bridges the gap between mathematical models and record/playback systems in the generation of GNSS signal data to be used in testing GNSS receiver algorithms, leveraging the advantages of both mathematical model-based simulators and record/playback-based simulators while mitigating the drawbacks of each.

Previous Abstract Return to Session B1 Next Abstract