Neural Network Aided Adaptive Filtering and Smoothing for an Integrated INS/GPS Unexploded Ordnance Geolocation System

Abstract

The precise geolocation of buried unexploded ordnance (UXO) is a significant component of the detection, characterization, and remediation process. Traditional geolocation methods associated with these procedures are inefficient in helping to distinguish buried UXO from relatively harmless geologic magnetic sources or anthropic clutter items such as exploded ordnance fragments and agricultural or industrial artefacts. The integrated INS/GPS geolocation system can satisfy both high spatial resolution and robust, uninterrupted positioning requirements for successful UXO detection and characterization. To maximize the benefits from this integration, non-linear filtering strategies (such as the unscented Kalman filter) have been developed and tested using laboratory data. In addition, adaptive filters and smoothers have been designed to address variable or inaccurate a priori knowledge of the process noise of the system during periods of GPS unavailability. In this paper, we study and compare the improvement in the geolocation accuracy when the neural network approach is applied to aid the adaptive versions of the extended Kalman filter (EKF) and the unscented Kalman filter (UKF). The test results show that the neural network based filters can improve overall position accuracy and can homogenize the performance of the integrated system over a range of relatively quiet to dynamic environments. Navigation-grade and medium-grade IMUs were compared and, with standard smoothing applied to the new filters, geolocation accuracy of 5 cm (13 cm) was achieved with the navigation-(medium-) grade unit within 8-second intervals that lack external control, which is at or close to the area-mapping accuracy requirement for UXO detection. K E Y

Integrating GRU with a Kalman Filter to Enhance Visual Inertial Odometry Performance in Complex Environments

“…However, studies like [33] predict absolute state vectors instead of vector increments using NN, which increases model complexity and requires a more extensive training process. Studies from [29,[32][33][34][35][36][37][38][39][40][41][42][43][44][45] adopted vector increments of the sensor observations and predictions during KF prediction, whilst most of the work only works on GNSS/INS navigation during GNSS outages, aiming for improving INS efficiency INS in urban settings and situations [31,38,39,41,42,46].…”

Ground vehicle gravimetry using inertial navigation system, dead reckoning, and barometric altimeter

“…Gravity information serves as the fundamental data for resource exploration, geodetic and geophysical research. In recent years, ground vehicle gravimetry has gained widespread application in groundwater surveys, urban cavity detection, and underground unexploded ordnance detection [1]- [3]. In traditional vehicle-mounted gravity measurement methods, the positioning, speed measurement, and acceleration measurement of the carrier all rely on the acquisition of GNSS information, and the quality of GNSS information directly determines the accuracy of gravity measurement.…”