基于Unscented粒子滤波的传感器管理算法

Sensor Management Algorithm Based on Unscented Particle Filtering

传感器管理是信息融合技术的重要研究方向,以往的传感器管理算法主要是针对线性融合系统,现实中非线性系统更为普遍,而针对非线性融合系统的传感器管理算法研究较少。粒子滤波是目前非线性领域中应用最广的滤波算法,该算法的主要思想是使用一个带有权值的粒子集合来表示系统的后验概率密度。U nscen ted粒子滤波采用U nscen ted卡尔曼滤波计算提议概率密度分布,粒子的产生充分考虑当前时刻的量测,使得粒子的分布更加接近状态的后验概率分布。针对非线性系统,提出了一种基于U nscen ted粒子滤波的传感器管理算法。首先利用U nscen ted粒子滤波对目标进行状态估计,求出目标的协方差;然后利用信息熵计算目标的信息增量;最后利用信息增量最大对传感器资源进行分配,并对该算法进行了仿真。

Sensor management is an important part of information fusion. In the past most of sensor management algorithm mostly fuscous on linear system, actually nonlinear system is more commonly. Particle filtering is the most popular filtering in the fields of nonlinear system. The key idea of this technique is to represent the posterior density by sets of weighed samples. Unscented particle filter generates a proposal distribution by UKF, the production of particles is considered of the current measurement so the distribution of particles is close to the state posterior probability distribution. Aiming at the nonlinear system, a sensor management algorithm based on unscented particle filtering is proposed. It first uses unscented particle filtering to generate the state estimate and the corresponding covariance of the target, and then uses information entropy to gain information increment of the target, at last the sensor resources are allocated by maximizing the information increment. The algorithm is simulated by the experiment.