Improved Particle Filtering Algorithm for Maneuvering Target Tracking

Abstract

The particle filtering(PF) algorithm, which is proposed recently, is an efficient method dealing with nonlinear and non-Gaussian problems. It is widely used in the field of maneuvering target tracking which is easily disturbed by the circumstances to solve non-linear or non-Gaussian problems. However, PF is not always satisfactory as it always need to use a large number of particles to estimate the true state of the target accurately. If the number of the particles is too large, the real-time performance of the filter will become lower. But if decrease the particles, the validity and diversity of the particles will become worse. So an improved PF algorithm is proposed in this paper. The new method uses a residual, which is equal to the value of the predict measurement reducing the latest measurement, to adjust the likelihood distribution of the particle filter. Via this adjust process, the sampling particles tend to the high-likelihood region before the weights of the particles are updated. The effectiveness and diversity of the sampling particles can be maintained through the method, and the sample-dilution problem can be overcome. The simulation results show that the improved particle filtering algorithm applied in maneuvering target tracking can improves the tracking performance.