Maneuver Recognition Based on Hierarchical CNN-MDN with Partially Trusted Data
Yu Hong Zhao, Yonglin Li, Jiyang Xiao, Zhenxing Zhang, Bo Pang, Xunan Huang · 2024
As the appearance of new aircraft and the change in combat mode have led to urgent needs for intelligent air combat, maneuver recognition and decision-making are the core of such combat. On the other hand, deep learning technology has achieved remarkable results in the civil field, which provides new ideas for research in this aspect. When the given flight data is missing and partially credible, the credibility parameter is specially added to the model of maneuver recognition. Thus, a convolutional neural network-hybrid density network algorithm is proposed, and the action with the maximum output probability brings the recognition result. In the simulation experiment, the convergence of the network is analyzed and compared with other algorithms, which show that the proposed network produces the best effect in maneuvering recognition.