Skip to Main Content
 

Global Search Box

 
 
 
 

Files

ETD Abstract Container

Abstract Header

The Estimation Methods for an Integrated INS/GPS UXO Geolocation System

Lee, Jong Ki

Abstract Details

2009, Doctor of Philosophy, Ohio State University, Geodetic Science and Surveying.
Unexploded ordnance (UXO) is the explosive weapons such as mines, bombs, bullets, shells and grenades that failed to explode when they were employed. In North America, especially in the US, the UXO is the result of weapon system testing and troop training by the DOD. The traditional UXO detection method employs metal detectors which measure distorted signals of local magnetic fields. Based on detected magnetic signals,holes are dug to remove buried UXO. However, the detection and remediation of UXO contaminated sites using the traditional methods are extremely inefficient in that it is difficult to distinguish the buried UXO from the noise of geologic magnetic sources or anthropic clutter items. The reliable discrimination performance of UXO detection system is depend on the employed sensor technology as well as on the data processing methods that invert the collected data to infer the UXO. The detection systems are required the very accurate positioning (or geolocation) of the detection units to detect and discriminate the candidate UXO from the non-hazardous clutters, greater position and orientation precision because the inversion of magnetic or EMI data relies on their precise relative locations, orientation, and depth. The requirements of position accuracy for MEC geolocation and characterization using typical state-of-the-art detection instrumentation are classified according to levels of accuracy outlined in: the screening level with position tolerance of 0.5 m (as standard deviation), Area mapping (less than 0.05 m), and characterize and discriminate level of accuracy (less than 0.02m). The primary geolocation system is considered as a dual-frequency GPS integrated with a three dimensional inertial measurement unit (IMU); i.e., INS/GPS system. Selecting the appropriate estimation method has been the key problem to obtain high precise geolocation of INS/GPS system for the UXO detection performance in dynamic environments. For this purpose, the Extended Kalman Filter (EKF) has been used as the conventional algorithm for the optimal integration of INS/GPS system. However, the newly introduced non-linear based filters can deal with the non-linear nature of the positioning dynamics as well as the non-Gaussian statistics for the instrument errors, the non-linear based estimation methods (filtering/smoothing) has been developed andproposed. Therefore, this study focused on the optimal estimation methods for the high precise geolocation of INS/GPS system using simulations and analyses of the two Laboratory tests (cart-based and handheld geolocation system). First, the non-linear based filters (UKF and UPF) have been shown to yield superior performance than the EKF in various specific simulation tests which designed similar to the UXO geolocation environment (highly dynamic and small area). The UKF yields 50% improvement in the position accuracy over the EKF particularly in the curved sections (medium-grade IMUs case). The UPF also performed significantly better than EKF and show comparable and improvement over the UKF when the IMU noise is symmetric and non-symmetric. Also, since the UXO detection survey doe not require the real-time operations, each of the developed filters was modified to accommodate the standard Rauch-Tung-Striebel (RTS) smoothing algorithms. The smoothing methods are applied to the typical UXO detection trajectory; the position error reduced significantly using a minimal number of control points. Finally, these simulation tests confirmed that tactical-grade IMUs (e.g. HG1700 or HG1900) are required to bridge gaps of high accuracy ranging solution systems longer than 1 second. Second, these result of the simulation tests were validated from the laboratory tests using navigation-grade and medium-grade accuracy IMUs. To overcome inaccurate a priori knowledge of process noise of system, the adaptive filtering methods have been applied to the EKF and UKF and it was called as the AEKS and AUKS. The neural network aided adaptive nonlinear filtering/smoothing methods (NN-EKS and NN-UKS) which is augmented with RTS smoothing method were compared with the AEKS and AUKS. Each neural networkaided, adaptive filter/smoother improved the position accuracy in the both straight and curved section. The navigation grade IMU (H764G) can achieve the area mapping level of accuracy when the gap of control points is about 8 seconds. The medium grade IMUs (HG1700 and HG1900) with NN-AUKS can maintain less than10cm under the same conditions above. Also, the neural network aiding can decrease the difference of position error between the straight and the curved section. Third, in the previous simulation test, the UPF performed best than other filters. However since the UPF needs a large number of samples to represent the a posteriori statistics in high dimensional space, the RBPF can be used as alternatives to avoid the inefficiency of particle filter. The RBPF is tailored to precise geolocation for UXO detection using IMU/GPS system and yielded improved estimation results with a small number of samples. The handheld geolocation system using HG1900with nonlinear filter based smoother can achieve the discrimination level of accuracy if the update rate of control points is less than 0.5Hz and 1Hz for the sweep and swing respectively. Also, the sweep operation is more preferred than swing because the position accuracy of the sweep test was better than that of swing test.
Christopher Jekeli (Advisor)
Burkhard Schaffrin (Other)
Alan Saalfeld (Other)

Recommended Citations

Citations

  • Lee, J. K. (2009). The Estimation Methods for an Integrated INS/GPS UXO Geolocation System [Doctoral dissertation, Ohio State University]. OhioLINK Electronic Theses and Dissertations Center. http://rave.ohiolink.edu/etdc/view?acc_num=osu1259186598

    APA Style (7th edition)

  • Lee, Jong Ki. The Estimation Methods for an Integrated INS/GPS UXO Geolocation System. 2009. Ohio State University, Doctoral dissertation. OhioLINK Electronic Theses and Dissertations Center, http://rave.ohiolink.edu/etdc/view?acc_num=osu1259186598.

    MLA Style (8th edition)

  • Lee, Jong Ki. "The Estimation Methods for an Integrated INS/GPS UXO Geolocation System." Doctoral dissertation, Ohio State University, 2009. http://rave.ohiolink.edu/etdc/view?acc_num=osu1259186598

    Chicago Manual of Style (17th edition)