|Abstract:||This paper will describe a GNSS/Inertial vehicular navigation mechanization that has been developed to enable robust navigation capabilities in harsh GNSS environments such as urban canyons, tunnels and parking garages. Results of extensive experimental validations will be presented to demonstrate precise and reliable navigation performance in dense urban environments. GIVE is implemented as a completely self-contained solution without any external sensor inputs (such as odometer aiding), thus enabling ease of installation and operational use. A tightly coupled navigation engine fuses GNSS measurement data (carrier phase and pseudoranges) with IMU measurements and vehicular motion model (non-holonomic constrains and zero-velocity updates). The implementation utilizes cell-phone quality inertial sensors (at a $2-5 price range), which significantly reduces the overall system cost. A key aspect of reliable navigation is ensuring consistency of navigation outputs. In other words, solution assurance has to be maintained with navigation error metrics being consistent with actual errors. Assured navigation is especially important for safety-critical applications such as connected cars and self-driving cars. Dense urban scenarios that are characterized by rich multipath environments represent a significant challenge for assured navigation. To address this challenge GIVE implements the following technological enablers: 1) Use of GNSS carrier phase for sensor fusion. Specifically, GIVE utilizes temporal carrier phase differences over time. Temporal differencing eliminates integer ambiguities while fully capturing the underlying motion dynamic for the INS error-state estimation. The use of temporal differences is particularly beneficial for the identification and removal of close-range non-line-of-sight (NLOS) multipath errors. It may be challenging to identify multipath signals reflected by buildings within a close range in the range domain. Yet, significant Doppler differences are generally present (due to the difference in LOS between direct and multipath signals), which makes multipath errors readily distinguishable in the temporal phase domain. 2) Residual thresholding applies INS-based statistical gating to eliminate those GNSS measurements that do not agree with their values predicted based on inertial outputs. 3) Probabilistic density association filtering (PDAF) further mitigates the influence of outliers that pass through the residual check. PDAF modifies the Kalman filer estimation step by directly incorporating a probability of outlier missed-detection into the measurement/prediction weighting scheme. The use of PDAF proves to be especially beneficial for “transitioning” cases (GNSS-denied to GNSS-challenged) such as exiting from a tunnel into an urban canyon. For such cases, residual sigma bounds are increased (due to unmitigated inertial drift) and multipath errors can leak through leading to detrimental effects on the navigation performance. De-weighting of potentially corrupted measurements based on their respective missed-detection probabilities significantly improves the robustness of the multipath mitigation performance. The paper will present the overall system mechanization of GIVE. Extensive test results in difficult urban environments will be used to demonstrate the system performance. Example scenarios will include downtown San Francisco and downtown Chicago tests with an example case of exiting from a tunnel into an urban canyon.|
Proceedings of the 32nd International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2019)
September 16 - 20, 2019
Hyatt Regency Miami
|Pages:||1335 - 1355|
|Cite this article:||
Soloviev, Andrey, Vadlamani, Ananth, Sharon, JD, Yang, Chun, "GNSS/Inertial Vehicular Engine (GIVE) for Automotive Navigation," Proceedings of the 32nd International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2019), Miami, Florida, September 2019, pp. 1335-1355.
ION Members/Non-Members: 1 Download Credit