在不确定噪声类型的非线性无线传感器网络环境下,采用分布式代价参考粒子滤波算法对移动目标进行跟踪。针对出现的概率密度函数方差不收敛的问题,结合最优控制思想,提出算法的改进。分析自适应函数中方差更新的特点,将修正系数γ替换步长k,以最接近真实轨迹为依据,选取最佳粒子;建立非高斯噪声的目标运动模型,利用仿真工具比较改进前后算法得出的多步数状态估计值方差,验证可行性,并给出修正系数γ的最佳取值范围。%Under an environment of nonlinear wireless sensor network with uncertain noise, distributed cost reference particle filter algorithm is used in moving targets tracking. To address the problem of diver-gency in probability density function of variance, an improved algorithm is proposed based on optimized control thought. Firstly the characteristics of variance updating in adaptive function are analyzed, and the correction factor γ is used to replace the step k to choose the optimal particle according to which is the most close to the real trajectory; the variance of more steps state estimation value is then compared be-tween before and after the improved algorithm by simulation tool, and the feasibility is verified, and at last, the optimal value range of correction factor γ is given.
展开▼