The generalized labeled multi-Bernoulli (GLMB) filter of Vo and Vo is an exact closed-form solution of the multitarget Bayes' filter and is, therefore, provably Bayes-optimal. Its recent implementations are extremely fast, with computational order O(n2m) where n,m are the current numbers of tracks resp. measurements. This paper generalizes the GLMB filter to fully integrated multitarget tracking and sensor management, in which dynamically moving sensors can appear and disappear and in which the states of these sensors are estimated via measurements collected by internal actuator sensors.
|