摘要
粒子滤波是基于递推的蒙特卡罗模拟方法的总称,可用于任意非线性,非高斯随机系统的状态估计。为了减轻退化现象,引入重采样过程,但重采样过程算法复杂,计算量大,不利于硬件实现,并且会削弱粒子的多样性,从而导致滤波性能下降。提出了一种将局部重采样和优化组合算法结合的重采样算法。将粒子按权值大小分类,小权值的粒子抛弃,大权值的粒子进行复制,将复制的粒子和抛弃的粒子线性组合产生新的粒子,增加了粒子多样性并且只对大权值粒子进行运算,故降低了计算量利于实时系统的硬件实现。仿真结果证明了该算法的有效性。
Particle filtering is a sequential Monte Carlo simulation algorithm.It can be used to estimate the state of any nonlinear,non-Gaussian system.In order to reduce the degeneracy,the resampling algorithm is adopted.But the resampling process has complex algorithm architecture,which have restricted its implementation in real-time system.Resampling process also leads to the loss of diversity of particles,and the loss makes filter＇s performance worse.A new algorithm-partial resampling combined with optimizing combination resampling method is proposed.Assort the particles by their weights,the particles which have low weights are abandoned and the particles which have high weights are reproduced,and generate new particles by combining the reproduced particles and abandoned particles.This new method partly overcomes the loss of diversity and because it simply operates to the high weights particle so its calculation is simplified.And it is propitious to implement by hardware.The simulation results prove the effectiveness of the proposed method.
出处
《电子测试》
2011年第4期91-94,共4页
Electronic Test
关键词
粒子滤波
局部重采样
优化组合
particle filtering
partial resampling
optimizing combination
