Target Situation Awareness via Electromagnetic Spectrum Mining Based on TSACGCN

In long-range electronic confrontation scenarios, the electromagnetic spectrum serves as a crucial resource for assessing adversarial situations. For large-scale swarms in maritime and aerial domains, however, obtaining complete electromagnetic spectrum data is challenging due to the expanded motion space and prolonged communication cycles, which greatly limits the reliability of situation assessment. In this article, swarm is mapped as a graph utilizing communication links, and a two-layer situation awareness model is proposed based on a temporal graph convolutional network with sampling, aggregation, and concatenation. In the inner layer, graph convolutional network with sampling, aggregation, and concatenation are used to mine the spatial features of electromagnetic spectrum information for target type identification. In the outer layer, bidirectional gated recurrent unit is used to capture the temporal dependencies of spatial features, thereby inferring swarm intentions. In response to dynamic adversarial environments, an aggregation mechanism is used to standardize data dimensions across layers. Experimental results show that the proposed model effectively identifies target types and action tasks even in cases of incomplete or insufficient data, so as to provide reliable information support for strategy formulation and resource allocation.