A LABELED MULTI-BERNOULLI FILTER FOR SPACE OBJECT TRACKING

To maintain custody of the increasing number of detectable objects in Earth orbit, tracking systems require robust methods of multi-target state estimation and prediction. One alternative to the classic multiple hypothesis and probabilistic data association methods uses a random finite set for modeling the multi target state. The common forms of such filters sacrifice knowledge of specific targets for the sake of tractability. This paper presents a labeled multi-Bernoulli filter for tracking space objects, which allows for the identification of individual targets. This version of the filter includes a new-target birth model based on the admissible region and non-Gaussian propagation of the single-target state probability density function. The benefits of the filter are then demonstrated for the tracking of both previously known and newly detected objects near geosynchronous orbit.