PMHT with timing uncertainty

Abstract

This paper considers the problem of tracking a target using measurements where the temporal information is noisy or unreliable. The measurements are modelled as being received with a known stochastic time delay and also with time error parameters that are to be estimated. By treating the true measurement time as missing data and using expectation maximisation, a modified version of the probabilistic multihypothesis tracker (PMHT) is derived for the uncertain-timing problem. This modified PMHT thus associates each measurement to the targets and to time instants. The method is validated using simulations, and the improvement over the standard PMHT is quantified. Further simulations compare the algorithm with existing alternatives dealing with timing uncertainty.