Monte Carlo Realisation of A Distributed Multi-Object Fusion Algorithm

We consider the problem of distributed target tracking in amulti-object, multi-sensor scenario in which the structure of the jointdistribution of the estimate between different nodes is unknown. Inthis paper we present a preliminary implementation of GeneralisedCovariance Intersection (GCI) fusion rule for multi-object posteriorsthrough a Monte Carlo realisation. We discuss the subtleties in the caseof multi-object distributions and derive a scheme for sampling fromExponential Mixture Densities which are at the heart of the GCI. Wedemonstrate the improvement in localisation of multiple targets in asimulation scenario.