Hypersonic Maneuvering Target State Estimation Algorithm Based on Recurrent Neural Network
ZANG Hongyan1， XIE Xiaolong2， XU Yazhou2， TAO Ye1， GAO Changsheng1
1. School of Astronautics， Harbin Institute of Technology， Harbin 150001， China；
2. Shanghai Electro-Mechanical Engineering Institute， Shanghai 201109， China
Abstract：To achieve real-time and highly accurate estimation of the motion state of a hypersonic maneuvering target, a hypersonic maneuvering target state estimation algorithm based on recurrent neural network was proposed in the study. Firstly, the aerodynamic parameters were introduced into the hypersonic vehicle motion model in the half-speed system as a slow variable, allowing the state estimation model of the hypersonic maneuvering target to be established and the ground radar measurement model to be determined. Then, considering that the model uncertainty caused by the strategic maneuvering of the hypersonic target would exceed the processing ability of the conventional filtering method, the target maneuvering behavior data set was created based on the hypersonic vehicle motion model and the motion behavior that may occur in the actual combat operations was simulated. After that, the multi-filter joint estimation algorithm was designed by combining multi-model filtering with recurrent neural networks. Finally, the proposed algorithm was simulated and verified under different simulation conditions. The simulation results show that the proposed hypersonic maneuvering target state estimation algorithm can successfully achieve rapid response to hypersonic target maneuvering changes, which are smooth, highly precise, and beneficial to subsequent prediction.
