Context-Enhanced Information Fusion for Tracking Applications

Abstract

Over the last decade, context has become a key source of information for tracking problems. Context inference allows refining sensor modeling and target dynamics as well as the creation of motion constraints according to the physical and operational conditions of the scenario. This work presents two example applications: indoor and inland waterway navigation where the context information is employed to reduce the uncertainty of the tracking and enhance the navigation solution. For indoor positioning a cascaded Extended Kalman Filter (EKF) and Particle Filter (PF) architecture is proposed. The system uses stance phase detection and the available floor map to construct the measurement models for the KF and motion constraints for the PF respectively. For navigating in inland waterways it is shown how to benefit from context information fusion by inferring the operating condition of the Global Navigation Satellite System (GNSS). In this scenario, context based criteria are derived for the selection of the best position estimation amongst several positioning solvers running in parallel. This work presents the basics for context fusion in tracking applications, illustrating the theory with two application examples. The preliminary results already demonstrate a performance improvement compared to state-of-the-art approaches.