| Author: | Shao, Jianbo |
| Title: | Robust state estimation and integrity monitoring for inertial-based multiple sensors navigation system in urban environments |
| Advisors: | Chen, Wu (LSGI) |
| Degree: | Ph.D. |
| Year: | 2025 |
| Department: | Department of Land Surveying and Geo-Informatics |
| Pages: | xiii, 140 pages : color illustrations |
| Language: | English |
| Abstract: | The prosperous development of intelligent transportation has heightened the demand for precise positioning of autonomous vehicles in dense urban environments like building canyons and viaducts, where satellite signals are unavailable. Intelligent vehicles typically utilize navigation and positioning functions by employing an inertial-based multiple-sensor integrated system (INMS) for navigation, which combines inertial navigation with satellite systems, odometers, and optical sensors. However, the integration of more sensors increases the likelihood of data errors and outliers due to challenging urban conditions, impacting the performance of state estimation methods and noise statistics identification that rely on Gaussian noise assumptions. In addition to the accuracy, autonomous driving must evaluate the confidence of position solution to ensure quick system responses and safe mode switches when position solution is unreliable, thereby reducing accident risks and enhancing safety. Integrity is crucial for assessing position confidence, yet most current INMS integrity monitoring methods are derived from aviation satellite navigation techniques. Due to heavier-tailed noise distributions and higher outlier rates, the unique challenges of multi-failure, non-Gaussian integrity monitoring in urban environments have been insufficiently explored. Consequently, the following key issues have been addressed to improve navigation accuracy and reliability in an urban environment. To address the problem of INMS state estimation performance degradation due to mismatched noise assumptions in urban environments, a robust resampling-free filtering algorithm based on the adaptive kernel-sizes maximum correntropy criterion is proposed. The cost function of the resampling-free update framework is constructed based on the maximum correntropy criterion, which effectively exploits the non-Gaussian moments of the state distribution caused by the non-closed mapping, ensuring the resampling-free estimation optimality and preventing the loss of the higher-order moment information from Gaussian reconstruction. Subsequently, an adaptive method for kernel size of correntropy is developed to realize the online optimal adjustment of the kernel size and ensure robustness against outliers. The simulation experiment demonstrates that the proposed algorithm can optimize the correntropy kernel size and improve the INMS state's estimation performance under non-Gaussian noise in urban environments compared with existing robust filters. To mitigate outlier interference with the measurement noise covariance matrix (MNCM) estimation, a robust noise adaptation algorithm is proposed based on a posterior smoothing variational approximation. The inverse Wishart distribution is used as the conjugate prior model of the MNCM, and a joint variational approximation analytical solution of the MNCM and smoothing state is derived. Then, the inverse Wishart distribution's inverse scale matrix is reconstructed based on the correntropy matrix to suppress the interference of measurement outliers on the MNCM estimation. The simulation experiment demonstrates that the proposed algorithm can effectively suppress the interference of measurement outliers on MNCM estimation and accurately identify the measurement noise statistics. To quantitatively assess the reliability of INMS state estimates in urban environments, an autonomous integrity monitoring (IM) algorithm based on multiple fault-missing detection assumptions is proposed. A consistency factor in the state domain is calculated using the sequential probability ratio over sliding windows. Under the multi-fault missing detection assumption, the horizontal protection level is calculated based on the maximum eigenvalue combined with the consistency factor to quantitatively assess the confidence of the position solution. The simulation experiment demonstrates that the proposed algorithm can effectively quantitatively evaluate the confidence of the position solution and monitor the navigation integrity in the case of measurement outlier disturbance. To validate the effectiveness of proposed algorithms in practical engineering, an invehicle experiment is conducted. The experimental results demonstrate that: 1) The proposed robust state estimation algorithm reduces the root mean square error (RMSE) of the horizontal position estimation by more than 5.0% compared with the existing robust estimation methods and has a higher robust state estimation accuracy in adverse urban areas; 2) The proposed noise adaptation algorithm provides a smoother and reliable MNCM estimation, which reduces the corresponding position RMSE by more than 13.6% compared to the existing methods and effectively suppresses the interference of measurement outliers on the MNCM adaptation; 3) Compared to the existing IM methods, the proposed IM algorithm has the higher reliability of protection level (99.85%) and does not produce any hazardous misleading events, which can effectively assess the position confidence and monitor the navigation integrity. Therefore, the experiment verifies the effectiveness of the proposed algorithm in practical engineering applications. |
| Rights: | All rights reserved |
| Access: | open access |
Copyright Undertaking
As a bona fide Library user, I declare that:
- I will abide by the rules and legal ordinances governing copyright regarding the use of the Database.
- I will use the Database for the purpose of my research or private study only and not for circulation or further reproduction or any other purpose.
- I agree to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.
By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms.
Please use this identifier to cite or link to this item:
https://theses.lib.polyu.edu.hk/handle/200/13926