基于 Renyi熵的非线性系统中传感器管理算法

摘要

针对高斯非线性系统中的多目标跟踪问题，提出了一种基于Renyi熵的传感器管理算法。该算法首先根据无迹卡尔曼滤波计算预测误差与滤波误差，度量目标跟踪精度要求；利用Renyi熵结合Parzen窗函数对概率密度函数进行近似估计，得到目标的信息增量，以此作为代价函数。同时，引入目标优先级（即威胁度），得到效能函数，形成传感器管理模型；最后利用该模型实现了传感器资源的分配。仿真结果表明，该算法利用Renyi熵可以表达非线性系统高阶特性的特点，结合Parzen窗函数，保持精度的同时减少运算量，较好地度量了跟踪过程中信息的不确定性，降低了跟踪误差，优化了系统的跟踪性能。%　　A sensor management algorithm is presented based on Renyi entropy for Gaussian nonlinear multi-target tracking system .First,the target tracking accuracy requirement was analyzed based on the prediction error and the filtering error calculated by using UKF .Then,the approximate estimation of probability density function was implemented by using Renyi entropy together with Parzen window function,and the target information gain was obtained .The target priority,i.e.,the threats degree,was introduced,and the efficiency function was obtained .Finally,the efficiency function was used to distribute the sensors .The simulation results show that:1 ) the use of Renyi entropy can represent the high order characteristics of nonlinear system;and 2 ) Combined with Parzen window function,it can not only maintain accuracy but also reduce the calculation cost,and can measure the uncertainty of the information in the tracking well,reduce the tracking error,and optimize the tracking performance of the system .