Tracking and guidance with intermittent obscuration and association uncertainty

A track-based multiple hypothesis filter is developed for tracking and guidance. The required target may be occasionally obscured for substantial periods resulting in a sequence of distinct “tracklets”. The association between these elements is uncertain due to spurious tracklets from other random objects. Each tracklet is processed independently and then combined according to a hypothesis structure using a novel partitioning approach. The filter produces a Gaussian mixture whose components correspond to feasible sequences of tracklets. A (near optimal) guidance demand is generated by applying a bounded cost function to the distribution of “zero effort miss” derived from the filter. The optimisation is computed efficiently via a convolution. Illustrative simulation results are presented.