基于分层空间的Rao-Blackwellised化粒子滤波算法

RAO-BLACKWELLISED PARTICLE FILTER ALGORITHM BASED ON LAYERED SPACE

针对于粒子滤波在目标跟踪里面粒子使用率低且单一等问题,提出一种基于分层空间的Rao-Blackwellised化粒子滤波算法。该算法通过Rao-Blackwell定理将线性变量边缘化从而减少状态空间维数,提高估计精度。在满足同等精度要求时,可大大减少所需粒子数目,因而大大降低计算负荷;分层理论可以把粒子空间分成多重空间,利用权重实现合理分配粒子,可以提高相关估计的准确性。实验结果表明,所提算法的均方根误差相比传统算法降低了50%,精确度在每个仿真时间内都有提高。

Concerning the problem of low and single particles utilisation rate of the particles filter in target tracking,we propose aRao-Blackwellised particle filter algorithm which is based on layered space.The proposed algorithm marginalises the linear variable throughRao-Blackwell theorem so as to reduce the state space dimensionality and to improve the estimation precision.It can greatly reduce thenumbersof particles required while meeting the same accuracy,thus the computational load is reduced greatly as well.The layered theory candivide the particle spaces into multidimensionality.It allocates the particles reasonably by using the weight and can improve the accuracy ofthe related estimation.Experimental results show that the root mean square error of the proposed algorithm reduces by 50% than thetraditionalalgorithm,and the accuracy was improved in each simulation time.