Tracking in clutter using probabilistic data association

The probabilistic data association (PDA) formulae have been rederived with the specific goal of obtaining formulae for probability of track existence as a by-product. Originally, Bar-Shalom and Tse (1975) derived the PDA algorithm from expressions conditioned on track existence. This effectively removed the information related to the probability that the track exists. The authors algorithm is derived from expressions in which the track existence is an event with a certain probability. This results in an integrated probabilistic data association algorithm (IPDA) which provides both data association and probability of track existence formulae. They describe IPDA in more detail and show by simulation that it provides more reliable and accurate tracking than IMMPDA. Finally, they consider implementation of PDA-algorithms in situations where there are a large number of real and false tracks. Their approach is to exploit the single-instruction multiple-data structure of the massively parallel MasPar MP-1 computer to achieve PDA track processing rates of about 20000 track per second.