A sequential Monte Carlo method for PHD approximation with conditionally linear/Gaussian models

A new sequential Monte Carlo procedure for approximating the probability hypothesis density is proposed. The algorithm, based on the replacement of numerical approximation with exact computation, is applicable to the class of conditionally linear/Gaussian models. The proposed algorithm is applied with an efficient, measurement-directed importance density to multiple target tracking using range-bearings measurements. Performance for a given sample size is significantly better than the previously proposed SMC-PHD.