STIRNet: A Spatio-Temporal Network for Air Formation Targets Intention Recognition

Air formation is a common style of air combat, which demonstrates a high degree of flexibility and strategic value in complex battlefield environments. The activity state of air formation is the result of the intertwining of time domain and air domain, which requires accurate execution of tactical processes in the time axis and skillful deployment of forces in three-dimensional space. Therefore, air formation target combat intention recognition is a complex and challenging task that requires an in-depth understanding of the dynamically changing behavioral patterns of the formation. To address this problem, this paper proposes the STIRNet (Spatio-Temporal Network for Intention Recognition) model, which abstracts the air formation as a spatial graph structure composed of vehicle nodes and combines its temporal data evolving over time. The model autonomously adjusts its attention to different moments and spatial locations through the spatio-temporal attention mechanism, focusing on the important spatio-temporal features that are crucial for recognizing the combat intention of the air formation; and simultaneously captures and integrates the feature information of the air formation in both the temporal and spatial dimensions through the spatio-temporal convolutional operation, which effectively solves the deficiencies of the traditional methods in dealing with the complex spatio-temporal dependency relationships. The experimental results show that the model proposed in this paper effectively improves the accuracy of the combat intention recognition of air formation targets, which is of great value for command decision-making and air battlefield situation assessment.