未知状态模型下基于高阶容积卡尔曼滤波和神经网络的状态估计算法

STATE ESTIMATION ALGORITHM BASED ON HIGH ORDER CUBATURE KALMAN FILTER AND NEURAL NETWORK WITH UNKNOWN STATE MODEL

针对非线性系统状态模型未知的情形,提出一种基于高阶容积卡尔曼滤波和神经网络的状态估计算法,解决了未知非线性系统模型的状态估计问题。在算法的实现过程中,首先利用神经网络对非线性系统建立状态空间模型,然后把神经网络的权重和系统的状态变量组合在一起作为新的状态变量,并采用高阶容积卡尔曼滤波对新的状态进行实时更新,从而达到神经网络对非线性系统模型的真实逼近以及对状态值的精确估计。最后的目标跟踪仿真表明,该算法具有更高的估计精度。

In view of the nonlinear state model of the system is unknown, this paper presents a state estimation algorithm based on high order cubature Kalman filter and neural network to solve the problem of state estimation of unknown nonlinear system model. The neural network is used to establish the state space model for the nonlinear system. Then, the weight of the neural network and the state of the system variables together are combines as the new state variables. And the new state is updated in real time by high order cubature Kalman filter, so as to achieve the neural network on the nonlinear system model of the real approximation and accurate estimation of the state value. The final target tracking simulation shows that the algorithm has higher estimation accuracy.