A Distributed Joint Integrated Probabilistic Data Association (JIPDA) Filter with Soft Object Association

Abstract

We propose a distributed multisensor joint integrated probabilistic data association (JIPDA) filter for multiobject tracking in decentralized sensor networks. Conventional Chernoff fusion of the posterior multiobject distributions of neighboring sensors presupposes a correct “hard” association of the objects tracked at the sensors. To avoid detrimental effects of an incorrect hard association, we develop a fusion method based on probabilistic (“soft”) object association. Our numerical results demonstrate significant performance gains relative to the hard association approach.