An Intelligent Method for Target Intention Recognition and Trajectory Prediction in Complex Aerial Combat Scenarios

With the development of new weapons such as decoy and self-defense interceptor, the air combat environment has become more complex. In this paper, an autoencoder intention recognition method is proposed to distinguish the types of targets. Additionally, a prediction method based on long short-term memory (LSTM) and convolutional neural network (CNN) is prosed for predicting target trajectories. Firstly, a three-degree-of-freedom relative motion model for missile-target interactions is established. Maneuver trajectories for target warplanes, decoys, and interceptors are designed using prior knowledge to establish the maneuver database. Secondly, an autoencoder is designed to compress and extract features from the trajectory data. Then, Softmax discriminant network is built to discriminate the intention of the trajectory data. Finally, the trajectory prediction network is designed based on LSTM and CNN to extract the features of the target trajectory from both temporal and spatial dimensions. The simulation results show that the intention recognition method proposed in this paper can accurately distinguish target warplanes, decoys and interceptors. And the trajectory prediction method can effectively improve the prediction accuracy compared with the traditional methods, which is suitable for the target prediction task in the complex air combat environment.