This project is concerned with accurately and reliably determining the state of a vehicle relative to a specified trajectory (e.g., a lane centerline). We are utilizing inertial navigation methods based on inexpensive solid state inertial sensors aided by external sensors such as carrier phase differential GPS, magnetometers, and roadway height. Due to this integration of sensors, reliability is increased relative to a single sensor approach and the changes required to the roadway infrastructure may be significantly decreased. This projects objectives included preparation for and participation in DEMO2002, research into INS aided GPS tracking of satellites, and research into methods to use auxiliary sensors (roadway height or magnetometers) to aid the integer resolution process. This project has been very interesting and successful. Although Demo2002 was ultimately cancelled our preparations for it were fruitful and we did participate in a smaller project demonstration with PATH at Crows Landing. In preparing for participation in Demo2002, we constructed much more robust prototypes of our GPS aided INS hardware. This include significant software re-writing that improved the reliability of the software. Overall, the hardware and software are now much more reliable and easier to use and install on test vehicles. In the area of INS aiding of a GPS receiver, research was performed and a new algorithm was design, but not implemented. We expect that this approach would lead to better satellite tracking during brief interruptions of the satellite signal. Implementation would require extensive interactions with the receiver design team, which is no longer possible due to changes in the GPS industry. In the area of aided integer ambiguity resolution, we have developed new algorithms to use information from non-GPS sensors to facilitate and speed-up the process of integer ambiguity resolution. This algorithm was evaluated in two experiments at UCR with significant improvements in the ability to correctly identify the integer ambiguities in a single epoch with fewer than 6 satellites. This algorithm was also used during the Crows Landing testing. Finally, UCR and PATH worked at PATH and Crows Landing and generated an impressive set of data as shown in Figures 12-25 of Section 4.2. This set of experiments was the culmination of the project. The experiments mounted the GPS/INS hardware on a bus that already was instrumented with the magnetometer hardware. We demonstrated that (1) the INS control state and magnetometer control state matched very well both when only the DCPGPS is aiding the INS and when both the magnetometer and DCPGPS are aiding the INS; (2) seamless transitions between magnetometer control, INS control, and manual control; (3) advanced maneuvers such as lane changes.

This dissertation considers reliable real-time navigation with Real-Time Kinematic (RTK) GPS aided Inertial Navigation System (INS). To improve the accuracy and reliability of the aided INS, a Contemplative Real-Time framework (CRT) is proposed by combining the conventional Kalman Filtering and the Bayesian smoothing which considers all the navigation information over a time window. To facilitate the formulation and the solution of the CRT problem, a probabilistic graphical model called Factor Graph is utilized.

To enhance the robustness of the navigation system to faulty measurements, a novel robust graph optimization method referred as Hypothesis Test based Least Soft-thresholding Square (HT-LSS) is proposed. Due to the integer ambiguity inherent in the RTK GPS carrier phase measurements, in this work the classical Factor Graph modeling is extended for RTK GPS/INS applications by incorporating integer unknowns. Nonlinear Mixed Integer Least Square (NMILS) is required to solve the CRT RTK GPS/INS problem. The major contribution of this thesis is the proposition of a novel Common-Position-Shift method to reduce the computational cost of the NMILS in this problem. In addition, a robust real-time differential correction computation approach was developed for reliable DGPS/RTK applications with internet transported differential information in RTCM and Ntrip standard. The proposed algorithms are evaluate using data acquired with the sensor platform mounted on an automotive vehicle to illustrate the performance.

Stockpile reclaiming using a Bucket Wheel Reclaimer (BWR) is an important part of stockyard management. The growth in demand for material handling over the years has drawn attention to improve the automation of the process. However, studies have shown that stockpiled products are being reclaimed at approximately 50% of their potential. This study focuses on the challenges in the automation of stockyard management system using a BWR.

For high accuracy point cloud computation and surface reconstruction of the stockpiled materials, accurate calibration is of crucial importance. This dissertation presents a calibration technique for multi-LIDAR systems to estimate the GNSS-LIDAR extrinsic parameters of BWR's. The approach presented works with one or more 2D LIDARs and does not require special markers (e.g., reflective tape) or surveyed locations (other than a DGNSS base station antenna). The method and its accuracy have been demonstrated using experimental data from a stockyard environment.

Regarding real-time management and control, the dissertation presents a technique for real-time point cloud management, visualization, and feature extraction for large scale stockyards environments. The software solution described continuously manages the point cloud in real-time as the sensors stream data. It also displays the current stockpile on an interactive interface that allows the user to see the surface from different view-points, interrogate the coordinates of any surface location, and computes the BWR entry and exit point for automated operation. The software is tested using experimental data from a port located in Yantai, China.

To operate, guide and control vehicles in low visibility conditions, it is critical that the states of the vehicle are accurately estimated, which includes the three dimensional position, velocity, and attitude. This can be accomplished by GPS (Global Positioning System) aided encoder or GPS aided inertial approaches. The overall positioning accuracy of either approach will be determined by the GPS performance. Real-time centimeter accuracy GPS positioning can be achieved using carrier phase measurements. This requires

fast and reliable on-the-fly integer ambiguity resolution.

In this dissertation, we focus on resolving GPS ambiguity problem, including both integer ambiguity estimation and integer ambiguity validation. For integer ambiguity estimation, a brief overview of pervious work on integer ambiguity resolution is first presented.

Then, an improved integer ambiguity resolution method is proposed. Subsequently, simulations and real-world data are presented to demonstrate the effectiveness of the method. We also present integer ambiguity algorithms with auxiliary measurements and algorithms with multiple epoch measurements, both of which are useful in GPS challenging areas. For integer ambiguity validation, a brief overview is first presented, and then analytic discussion and test results on several popular validations methods are studied. Finally we discuss GPS modernization and its effect on integer validation.

This report describes the results of an effort to implement and analyze the performance a vehicle control system using control state information obtained from a carrier phase Differential Global Positioning System (DGPS) aided Inertial Navigation System (INS). Keywords: Vehicle Positioning Systems, Global Positioning System, Inertial Navigation, Differential Carrier Phase, Advanced Vehicle Control

Many applications require reliable, high precision state estimation while mitigating measurement outliers. This dissertation presents a novel state estimation approach to the challenge of preventing outlier measurements from affecting the accuracy and reliability of state estimation. Since outliers can degrade the performance of state estimation, outlier accommodation is critical. The most common method for outlier accommodation utilizes a Neyman-Pearson (NP) type threshold test in a (extended) Kalman filter (KF) to detect and remove residuals greater than a designer specified threshold. Such threshold based methods may use residuals arbitrarily close to the threshold, even when they are not needed to achieve an application's performance specification. Outlier measurements that pass the residual test (i.e., missed detections) results in incorrect information being incorporated into the state and error covariance estimates. Once the state and covariance are incorrect, subsequent outlier decisions may be incorrect, possibly causing divergence.

The major contribution of this dissertation is changing the focus from outlier detection, to looking for a subset of measurements which have minimum risk while achieving a lower bounded information for state estimation. Risk-averse performance-specified (RAPS) state estimation works within an optimization setting to choose a set of measurements that achieves a performance specification with minimum risk of outlier inclusion. This dissertation derives and formulates the RAPS solution for outlier accommodation which applies to both linear and nonlinear applications. The approach is also extended to moving horizon state estimation problem. Global Navigation Satellite Systems (GNSS) and inertial measurements for moving vehicle state estimation are used as an example to show the performance of the proposed approach.