关键词: 全域火力场, 态势认知, 超视距空战, 威胁预测, 动态逃逸, 深度学习

Abstract:

To address the threat prediction issue in beyond-visual-range air combat situational awareness， a global fire field-based threat prediction and dynamic evasion methodology is proposed. Firstly， a vector field theory-driven all-domain fire field modeling framework is established， formally defining the concepts and mathematical models of single-aircraft all-domain fire field， joint all-domain fire field， and their corresponding evasion fields. Subsequently， a real-time fire field computation method based on Multi-modal Residual Fusion Network （MRFNet） is developed， which resolves the computational bottlenecks and discrete field distortion issues inherent in traditional Monte Carlo approaches via an “offline training-online inference” deep learning paradigm. Concurrently， a short-term trajectory prediction algorithm employing a Vector Autoregressive Model （VAR）is introduced to enable real-time forecasting of multivariate flight states for both adversarial and friendly aircraft. The proposed methodology allows real-time computation of global and local threat assessments in complex distributed combat scenarios， providing targeted threat warnings and evasion recommendations. Experimental results demonstrate that the MRFNet-based approach reduces single fire field computation time from minute-level to millisecond-level while maintaining fitting errors below 5×10^-4， exhibiting excellent data smoothing and extrapolation generalization capabilities. The VAR-based trajectory prediction achieves longitude/latitude errors below 8.73×10^-4， outperforming various state-of-the-art deep learning-based time series prediction methods， with relative error losses remaining under 22% under strong positional deviation interference in threat detection. Comprehensive simulation analyses confirm that the proposed methodology aligns with pilots’ cognitive logic in real combat scenarios， demonstrating high fault tolerance， strong robustness， and low latency characteristics. This work shows significant practical value and operational relevance for adapting to complex distributed combat environments.

Key words: all-domain fire field, situational awareness, beyond-visual-range air combat, threat prediction, dynamic evasion, deep learning