We propose to approximate the PHD by a set of weighted random samples which are propagated over time using a generalised SMC method. The resulting algorithm is very attractive as it is general enough to handle non-linear non-Gaussian dynamics