Abstract:

For the problem of high computational complexity of multiple model probability hypothesis density (PHD) filters based on sequential Monte Carlo method and its improved versions, which are caused by distributing a lot of time updated particles to all over the regions where the target may appear in the next time step, a unscented transform based multiple model PHD filter for tracking maneuvering targets is proposed. The algorithm exploits the last measurements to generate the proposal distribution function and sample particle from this distribution. Consequently, most of the sampled particles will around the regions where the target likely appears in the next time step. Numerical simulations demonstrate that the proposed method can not only reduce the computational complexity but also improve the performance of multiple model PHD maneuvering target tracking.