适于惯导系统初始对准的神经网络实时算法研究

通常卡尔曼滤波器被用于解决惯导系统的初始对准问题。由于卡尔曼滤波的运算时间与系统阶次的立方成正比 ,所以当系统阶次较高时 ,滤波器会失去实时性。而神经网络具有函数逼近性能 ,实时性又好。为此 ,本文研究了一种基于扩展卡尔曼滤波原理的权值更新多层神经网络学习算法 ,对此算法进行了详细的推证 ,并将该算法运用到惯导系统的初始对准过程。仿真结果表明了这种神经网络结构用于惯导系统初始对准问题的有效性 ,既真正获得了与扩展卡尔曼滤波器相同的对准精度 ,又大大提高了系统的实时性

A Real- time Neural Network Learning Algorithm Adapt to the Initial Alignment of Inertial Navigation System

As a rule, the Kalman filter has been used to solve the initial alignment of inertial navigation. Whereas the computer time of the Kalman filter depends on the dimension n of the inertial navigation system model state vector. The number of computations per iteration is on the order of n\+3. Any more number of states would take leave of real time in computation time. We all know that the neural network has the ability of self-learning and good performance of real time. A learning algorithm for multiplayer neural network based on the Extended Kalman filter theory is studied. The theoretical procedure of the algorithm are described in details, and using the algorithm to the initial alignment of the inertial navigation system. Simulation results proved the availability of the neural network algorithm for initial alignment of nonlinear inertial navigation system. Not only can surely obtain alignment accuracy which is similar to that of the Extended Kalman filter, but also the alignment time is reduced considerably. Consequently, an available algorithm of the neural network for the initial alignment of nonlinear inertial navigation system is discovered.