Rao-Blackwellized Unscented Particle Filter for a Handheld Unexploded Ordnance Geolocation System using IMU/GPS

Abstract

The existence of Unexploded Ordnance (UXO) is a serious environmental hazard, especially in areas being converted from military to civilian use. The detection and discrimination performance of UXO detectors depends on the sensor technology as well as on the processing methodology that inverts the data to infer UXO. The detection systems, typically electro-magnetic induction (EMI) devices, require very accurate positioning (or geolocation) in order to discriminate candidate UXO from non-hazardous items. For this paper, a hand-held geolocation system based on a tactical-grade IMU, such as the HG1900, was tested in the laboratory over a small, metre-square area in sweep and swing modes. A camera position system was used to emulate GPS or alternative ground-based external ranging systems that control positioning errors. The proposed integration algorithm is a combination of linear filtering (Extended Kalman Filter) and nonlinear, also non-Gaussian filtering (Unscented Particle Filter) in the form of the Rao-Blackwellized Particle Filter (RBPF). The test results show that the position accuracy was improved by applying nonlinear filter-based smoothing techniques in both the straight and curved sections of the sweep and swing trajectories.