Trust Region Based Sequential Quasi-Monte Carlo Filter

A trust region based sequential quasi-Monte Carlo filter is proposed for system state estimation and object tracking which are the non-linear and non-Gaussian random procedures with multi-source uncertain information.In the proposed algorithm,the quasi-Monte Carlo(QMC) technique is used to optimize the distribution of the sampling particles in the state space,which can obtain a small error of the integration in the filtering process and a better accuracy of the state estimation.Furthermore,a trust region(TR) procedure is used to move particles to regions of high likelihood,which results in a fewer particle selection and lower computational cost.Experimental results show that the proposed algorithm overcomes the particle impoverishment,reduces the computational complexity of the QMC filter,and gets a more accuracy estimation than existing algorithms such as particle filter and QMC filter in system state estimation and object tracking.