To solve the particle weights degradation and sample impoverishment problems in particle filter which was used in non-Gaussian nonlinear systems, a new kind of particle filter was proposed. The sampling particles could move to high likelihood area by using particle swarm optimization idea, and the weights of degradation of the particles were reduced. Then through the variation operation of the artificial immune, the range of the optimal value was expanded when searching for the optimization, and the diversity of the particles was increased, Simulation results showed that the state estimation accuracy of new algorithm is improved nearly 40 times than standard particle filter. The filtering efficiency is 37.523%, 37 times of the standard particle filter. This algorithm has better real-time and state estimation precision, and could effectively relieve the particle of the weights of the exhaustion of sample degradation.
