基于运动参数预测的群组移动节点定位算法

Localization algorithm of group-mobility nodes based on mobility parameters prediction

针对群组移动节点定位算法普遍基于不切实际的假设,存在普适性欠佳和精度不高的问题,提出一种基于运动参数预测的群组移动节点定位算法。该算法根据群组移动节点具有相似运动的特点,运用Hermite插值多项式预测、过滤节点运动参数。为确保定位精度,应对节点移动性带来的采样区域变化,运用预测节点运动参数构建粒子有效采样区域;为节省时间开销,基于采样粒子真实分布与其极大似然估计值之间的最大K-L(Kullback-Leibler)距离确定能够满足不同采样区域的最少粒子数目;为改善算法收敛性,运用预测运动参数创建滤波公式,并选取优质粒子参与节点位置估计。在与经典算法MCL(Monte Carlo localization)法和加权最小二乘法的MATLAB对比实验中,分析了节点移动速度、自由度、K-L距离阈值、采样方格边长对定位精度的影响。结果表明,较上述算法,本算法的定位误差和时间开销较小,无须锚节点辅助,普适性较好。

This paper proposed a localization algorithm of the group-mobility nodes based on mobility parameters prediction.This algorithm aimed to solve the bad universality of other localization algorithms of group-mobility nodes,which were based on some unpractical hypothesizes.First,according to the similar move pattern of the group-mobility nodes,it used Hermite interpolating polynomials to predict and filtrate the mobility parameters of the group-mobility nodes.Due to the variational network topology caused by the mobility of nodes,the valid sampling areas were changing all the time.To solve the proplem,the valid sampling areas could be confirmed based on the predicted mobility parameters.This measure aimed to keep a high precision of the proposed algorithm.Besides,it calculated the least numbers of sampled particles of different sampling areas according to the Kullback-Leibler distence(KLD)between the real distribution of the sampled particles and its maximum likelihood estimation.This measure aimed to reduce the time expenditure of the proposed algorithm.Then it selected high quality particles for the following position estimation,according to the filter formulation designed by the predicted mobility parameters.This measure could improve the convergence of the algorithm.Finally,comparative experiments were made in MATLAB.It analyzed the effects of the main factors on the localization precision.The main factors contained the velocities of the group-mobility nodes,the freedom degree of the group-mobility nodes,the threshold value of the KLD and the size length of the sampling gird.Experiment results show that the proposed algorithm without any assistant localization measures is a good localization method of group-mobility nodes,which has a smaller time expenditure and higher precision than classical Monte Carlo localization(MCL)method and weighted least square(MLS)method.