粒子滤波算法的关键技术应用

针对基于贝叶斯原理的序贯蒙特卡罗粒子滤波器出现退化现象的原因,以无敏粒子滤波(UPF)、辅助粒子滤波(ASIR)及采样重要再采样(SIR)等改进的粒子滤波算法为例,对消除该缺陷的关键技术(优化重要密度函数及再采样)进行了分析研究.说明通过提高重要密度函数的似然度、引进当前测量值、预增和复制大权值粒子等方式,可以有效改善算法性能.最后通过对一无源探测定位问题进行仿真,验证了运用该关键技术后,算法的收敛精度和鲁棒性得到进一步增强.

Application of Key Techniques based on Particle Filters

We analyze the degeneracy phenomenon of sequential Monte Carlo particle filters based on bayesian theorem,put focus on the key techniques(good choice of importance density and use of resampling) to reduce its effects.Several improving schemes such as the Unscented Particle Filters(UPF),the Auxiliary Sampling Importance Resampling(ASIR) and the Sampling Importance Resampling(SIR) algorithms are introduced to illustrate through increasing the likelihood of the importance density or incorporating new measurement,or replicating particles with large weights within the generic frame of particle filters,the convergence accuracy and robustness behaviors of the algorithm can be effectively improved.A typical passive detection and location problem is simulated to prove above conclusions.