萤火虫算法智能优化粒子滤波  

Firefly Algorithm Intelligence Optimized Particle Filter

作　　者：田梦楚[1] 薄煜明[1] 陈志敏[1,2] 吴盘龙[1] 赵高鹏[1] 

机构地区：[1]南京理工大学自动化学院,南京210094 [2]中国卫星海上测控部,江阴214431

出　　处：《自动化学报》2016年第1期89-97,共9页Acta Automatica Sinica

基　　金：国家自然科学基金(61501521;U1330133;61473153;;61403421;61203266);国防重点预研资助项目(40405070102)资助~~

摘　　要：针对粒子滤波(Particle filter,PF)重采样导致的粒子贫化以及需要大量粒子才能进行状态估计的问题,本文结合粒子滤波的运行机制,对萤火虫算法的寻优方式进行修正,设计了新的萤火虫位置更新公式和荧光亮度计算公式,并在此基础上提出了萤火虫算法智能优化粒子滤波.该方法引入了萤火虫群体的优胜劣汰机制以及萤火虫个体的吸引和移动的行为,使粒子群智能地向高似然区域移动,提高了粒子群的整体质量.实验表明该方法提高了粒子滤波的预测精度,同时大大降低了状态值预测所需的粒子数量.Given the particle impoverishment due to particle filter（PF） resampling and given the need of a large number of particles for state estimate, the optimization mode of firefly algorithm is revised in combination with the operating mechanism of particle filter, and a new update formula of firefly position is designed as well. On this basis, an intelligent optimized particle filter of firefly algorithm is proposed. By means of firefly group s mechanism of survival of the fittest and individual firefly s attraction and movement behaviors, this algorithm enables the particle swarm to move toward the high likelihood region with the purpose of improving the total mass of particle swarm. The experiment has shown that the algorithm has upgraded the prediction accuracy of particle swarm and substantially reduced the quantity of the particles required by the prediction of state value.

关 键 词：粒子滤波 萤火虫算法 粒子贫化 状态估计 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]