Merging probabilistic data of multiple targets detected by multiple sensors

The aim of this work is to present an extension to current data fusion techniques and associated results for the implementation of a collaborative multi-platform, multi-target detection system. A multi-sighting data fusion algorithm has been simulated. The groundbased platforms have been assumed autonomous, with fully operational guidance systems. The attached sensors have associated errors that are controlled through the simulation. The target sightings and errors are translated into estimates with associated covariances in both the major and minor axes represented by 3-dimensional Gaussian distributions. Data merging is performed in two stages using the Jointly Gaussian Probability Density Function (JGPDF) with global alignment and minimal acceptable distance calculations. Empirically data highlights that, when using this approach a better estimate of the target’s location can be obtained when more observations are made, along with distinguishing between multiple targets. This paper aims to explore the issues surrounding localising detected targets within a know region from data gathered by multiple platforms. The problem is addressed by using a platform, having a known map to perform self-localisation, to detect and localise a target with respect to itself. The technique of interest for localisation is Simultaneous Localisation and Map building (SLAM) while research will be conducted into data fusion techniques for merging of the resulting target acquisition data. The main objectives of this work are to:

• gain a theoretical understanding of SLAM and the surrounding issues,
• compare and contrast the estimation techniques employed within SLAM,
• perform a short study into appropriate sensor suites and fusion techniques and
• develop a practically feasible solution to the described SLAM problem.