神经网络校正的EKF在高超声速目标跟踪中的应用

针对临近空间高超声速目标高度非线性、强耦合、高机动、时变参数、和独特气动特性等特点,综合运用军事运筹学理论与方法、系统建模技术、神经网络技术以及计算机仿真等,提出基于神经网络校正的扩展卡尔曼滤波(EKF)算法,并在高超声速目标跟踪中进行了应用研究。采取神经网络的学习能力来克服卡尔曼滤波发散问题,通过卡尔曼滤波后加一级误差处理环节使滤波收敛。仿真结果表明:该算法在目标发生较大机动时仍能保持较高的跟踪精度。

Application of Neural Network-aided Kalman Filtering Technique in Target Tracking of Hypersonic Vehicle

Aiming at the characteristics of nonlinearity, strong coupling and maneuvering, parameter varity and unique aerodynamics and so on in hypersonic vehicle, the method of neural network aided Kalman filtering for the near space hypersonic vehicle is proposed, and be applied to study of hypersonic target tracking by synthetically applying military operational research theory and method,system modeling technique, neural network technique and computer simulation technique, etc. The basic thought is making use of study capability of BP neural network to overcome the divergence problem in extended Kalman filter. After the extended Kalman filter, adding an error processing section makes filter concentrated. The simulation study shows that the algorithm is better than ones by extended Kalman filter. The Algorithm will have certain reference in target interception of hypersonic vehicle in Near Space.