基于极大似然目标状态估计的传感器管理

Sensor Management Under Maximum Likelihood Target State Estimation

摘要:

为估计目标运动的状态估计量,设计相应的传感器管理方案,文章依据贝叶斯概率知识,采用极大似然的最优化方法来估计目标运动的状态估计值；依据目标状态估计的误差协方差矩阵与其Fisher信息矩阵间的联系,推导似然函数关于目标状态各个分量的偏导数；采用随机离散时间系统知识,推导出对角式Fisher信息矩阵所有对角线上的各个元素；以Fisher信息矩阵的迹范数作为传感器管理中的最优代价函数,采用0-1整数规划算法来求解传感器的分配矩阵；最后将极大似然的目标状态估计及其相应的传感器管理方案应用于无人直升机的飞行控制系统中,以验证本文方法的有效性。

Abstract:

To estimate the state estimation about the considered target, and devise the corresponding sensor management, this paper uses Bayesian probability knowledge and maximum likelihood optimum method to estimate the target state estimation. This maximum likelihood method can not only solve the explicit solution of the target state with optimum strategy, but also deal with many constraints about the measurement and state variables. With the closed connection between the error covariance matrix and its Fisher information matrix, we derive the differentiation of the likelihood function with respect to each variable. Using some knowledge from stochastic discrete time system, we give every element which lies in the diagonal line of the diagonalized Fisher information matrix. The trace operation of the Fisher information matrix is applied to be the optimal cost function in the sensor management and then one 0-1 mixed integer numerical programming is used to obtain the sensor distribution matrix. Finally, we apply the sensor management under maximum likelihood target state estimation strategy into the flight control system of UAV in order to confirm the efficiency of the proposed strategy.