Study Of Particle Filter Algorithm

The classic Kalman filter and extended Kalman filter are the most commonly used in target tracking fields.For the former linear systems, which is used for nonlinear systems. Target tracking is a typical dynamic system state estimation problem. When the system is linear Gaussian conditions, the Kalman filter can give the best estimation. However, in reality, most of the system state is nonlinear, non-Gaussian, so the traditional filtering algorithms can not apply. On the 20th century the particle filter algorithm is rising, with the non-linear, non-Gaussian conditions, and strong processing power, many researchers move to this attention. Particle filter is a theoretical framework based on Bayesian Monte Carlo sampling method, using a series of weighted point set to approximate the particle posterior probability density distribution. Firstly, the traditional filtering algorithm is studied, and points out its shortcomings and scope of application. Then, in-depth study of the particle filter and the defects of particle filter was introduced and improved methods was introduced. In the fourth chapter, the particle swarm optimization Rao-Blackwellized particle filter algorithm is introduced. For the state of the nonlinear part, by particle swarm optimization, driven by all the particles move to the high likelihood region, so that fewer particles can be used to achieve better estimation performance. For the linear part of the state, still using Kalman filter for processing. Chapter presents a differential evolution of the particle swarm co-evolution particle filter algorithm. Biological groups based on the ecosystem within among species allelopathy, and common development of ideas arising from the co-evolution optimization algorithm to optimize the particle filter, simulation results show perfect performance.