In this paper we describe a sequential K-track L-list Viterbi algorithm for Bayeasian measurement /track association and estimation of the number of target tracks K. This algorithm can handle missed detections, false alarms and track initialization. Compared to direct computation of the solution of the Bayesian problem, this algorithm manages computational cost by sequentially eliminating from further consideration candidate K-track sets which have too great a Bayesian cost. The trellis diagram framework onto which the algorithm is built is generally applicable in that: it can accommodate other Bayesian an ML tracking problem solutions, as will as practical approximations to these; and it can be used for situations other than the one addressed here (e.g. for sequential track initiation and elimination).
We plan to further develop this generalized Viterbi tracking algorithm approach, by extending the algorithm to implement sequential track initiation and elimination, by considering further computation reduction via additional candidate K-track set pruning, and by incorporating practical Bayesian/ML motivated techniques such as joint probabilistic data association filtering (JPDAF) [11], probabilistic multi-hypothesis tracking (PMHT) [12] and interactive multiple model (IMM) [13].