Track-to-track fusion using split covariance intersection filter-information matrix filter (SCIF-IMF) for vehicle surrounding environment perception

Vehicle surrounding environment perception is an important process for many applications. Nowadays, a tendency is to incorporate redundant and complementary sensors into an intelligent vehicle, in order to enhance its perception ability; then an essential issue arises naturally, i.e. what fusion architecture can be used to combine the data from multiple sensors? In this paper, we propose a new track-to-track fusion architecture using the split covariance intersection filter-information matrix filter (SCIF-IMF). The basic idea is to use the IMF (adapted for estimates in split form) to handle the track temporal correlation of each sensor system and to use the SCIF to handle track spatial correlation. The proposed architecture enjoys complete sensor modularity and thus enables flexible self-adjustment. A simulation based comparative study is presented, which shows that the track-to-track fusion architecture using the SCIF-IMF can achieve centralized architecture comparable performance.