Answers are grounded in this paper and cite their sources.
Conventional modeling approaches struggle to characterize the inhomogeneous structure of complex 3-D plasma sheath and establish the dynamic correlation between the observation aspect angles and key scattering regions within the flow field. The multimodal radar cross section (RCS) prediction model integrating multidimensional flow field tensors is proposed to overcome these limitations. First, the multimodal dataset is constructed, comprising six key parameters and 3-D flow field data obtained from simulations across various flight conditions and vehicle geometries. Second, the RCS prediction model is established, integrating the angular-biased attention mechanism that dynamically links the incident angles to dominant scattering regions in the 2-D unfolded flow field, precisely capturing interference and resonance effect due to plasma inhomogeneity. Experimental results demonstrate the outstanding performance of the prediction model on the test set, achieving the coefficient of determination (R²) of 0.9311 and the root mean square error (RMSE) as low as 1.4921 dB, significantly outperforming the basic model that relies solely on vehicle parameters. This quantitatively confirms the essential contribution of integrating physical flow field information in enhancing prediction accuracy. Compared with the reference physical optics (POs) method, the proposed model reduces the per-sample inference time from minutes to milliseconds while preserving prediction accuracy. By integrating multimodal fusion and the angular-biased attention mechanism, this article achieves rapid RCS prediction for hypersonic targets. This approach provides an efficient and physically interpretable technical pathway for radar early-warning and identification systems.
No takes yet. Share an insight, caveat, or question.
Li et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: