Improved Evolutionary Particle Filter Algorithm Applied in Radar Tracking

In particle filter algorithm, resampling is always used to release sample impoverishment phenomenon, but it weakens the diversity of samples set and cause the algorithm unrobust. Based on imitating biology evolvement regulation, paper (Mo Yi-wei, et al., 2005) brought forward the evolutionary particle filter (EPF) algorithm. On the cost of much calculation, this method ameliorates the diversity of samples set to relieve the effect caused by samples impoverishment, but in paper (Mo Yi-wei, et al., 2005) the way to select the variation strength is not related. In radar tracking, this paper brings forward an improved evolutionary particle filter (IEPF) algorithm, in which variation strength is based on state noise and measurement noise. What´s more, unlike EPF in previous paper, in which the evolution proceeds at each step, the algorithm sets threshold of effective particles to determine if variation is necessary at current step, and much calculation is saved. Simulations demonstrate the feasibility of proposed algorithm