改进的SRCDKF-PF算法及在BOT系统中的应用

Application of Improved SRCDKF-PF for BOT System

针对纯方位目标跟踪(Bearing-Only Tracking,BOT)系统强非线性特点,提出一种新的解决方案:采用平方根中心差分卡尔曼滤波器(Square-RootCDKF,SRCDKF)产生粒子滤波提议分布,融入最新的观测数据影响;增加改进措施以提高滤波性能,如采用系统重抽样算法减少方差、应用马尔可夫链模特卡罗(Markovchain Monte Carlo,MCMC)方法消除粒子贫乏等。仿真表明该算法是有效的,针对当前BOT系统,比传统EKF、PF算法可靠性更好,跟踪精度更高。

A new solution was proposed to the bearings-only tracking （BOT） system, due to its strong nonlinear nature of the system. In the solution, the Square-Root CDKF （SRCDKF） proposal distribution was presented to update the particles, and it allowed the particle filter to incorporate the latest observations into a prior updating routine. Many improved methods were introduced to increase the filter performance, e.g. the systematic resample algorithm was used to decrease the covariance, and the Markov chain Monte Carlo （MCMC） method was used to eliminate the impoverishment of the samples, etc. Simulations show the improved SRCDKF-PF algorithm indicates higher reliability and better accuracy than the traditional PF and EKF algorithms for the current BOT system.