Rohde & Schwarz has outlined the growing challenge posed by mode-agile threat emitters, which operate in non-traditional modes that traditional static threat libraries cannot detect. These emitters use unexpected frequencies, modulation techniques, and hopping schemes, making them difficult to counter with predefined databases. The company emphasizes the need for adaptive systems capable of real-time threat classification and response in contested environments.
Cognitive AI/ML architectures are positioned as the solution, enabling autonomous perception, reasoning, and response to unknown threats. These systems rely on artificial neural networks, deep neural networks, fuzzy logic, and genetic algorithms to process and classify signals. The white paper also details the implementation challenges, including computational resource demands at the tactical edge, minimizing detect-to-counter latency, and managing SWaP-C constraints. Additionally, the use of hardware-in-the-loop and system-in-the-loop training systems is highlighted as critical for iterative development and validation of cognitive AI/ML algorithms.
The white paper reviews the architecture of cognitive AI/ML radar and EW systems, including key functional blocks such as RF acquisition, AI-driven analysis and inferencing, waveform synthesis, and RF generation. It also examines the challenges of training these systems — from acquiring real-world and simulated signal datasets to performing hardware-in-the-loop (HIL) and system-in-the-loop (SIL) testing — and describes how closed-loop testbeds can iteratively develop, validate, and improve the AI/ML algorithms needed to counter unknown threats.
Source: ieee
