A Field Guide to Cognitive Electronic Warfare
Part of the Field Guide to Cognitive Electronic Warfare. All information is derived from unclassified, publicly releasable sources.
For most of its history, electronic warfare operated by lookup: match the detected emitter against a library of known threats and apply the response the library prescribes. Software-defined emitters broke that method, because they change waveform from one transmission to the next and appear in forms held in no library. Cognitive electronic warfare replaces the lookup with a system that senses the spectrum, decides, acts, and learns from the result within a single engagement and with limited human involvement. The field is straightforward to define and difficult to field, and the difficulty has a definite structure. Four problems account for most of it.
The four problems
A cognitive EW system cannot directly observe the effect of its own action on a hostile system, so it optimizes and learns against feedback it can only infer. It has to act on novel threats for which no training data exists, often from a single observation under a deadline. Its adversary adapts in return, so the objective is a payoff in a contest rather than a fixed target. And because it changes its own behaviour, it cannot be certified like fixed software, so it has to earn the trust required to be allowed to act. These four problems organize this guide.
One constraint runs through all four: a cognitive EW system can act only on what it can measure. It optimizes a score built from what it observes, learns from feedback it obtains, and is trusted only insofar as its behaviour can be checked. The quantity hardest to measure, the effect of an action on an adversary that does not report its state, is the one that matters most, and that single difficulty is why the four problems are hard at the same time. Electronic Support and Battle Damage Assessment develops the constraint in full.
Scope
This is a field orientation, written for someone who needs a working command of the subject rather than an implementation guide. It covers what cognitive EW is, why it is hard, and where the difficulty and the current capability sit. It assumes a working knowledge of the electromagnetic spectrum and nothing more. It contains no classified or controlled material, and describes no organization's methods or work.
How to read this guide
Five pages, in reading order. Read them through for the whole subject, or start with the problem that matches your question.
- What Is Cognitive Electronic Warfare? — why library-based EW failed, what separates adaptive, automated, and cognitive systems, and the four hard problems.
- Electronic Support and Battle Damage Assessment — measuring the environment and the effect: the constraint that bounds the whole field.
- Machine Learning for Electronic Warfare — learning from almost no data, during the mission, and the RF foundation models that make one observation enough.
- Electronic Attack, Protection, and Cognitive Radar — the contest: synthesizing countermeasures, reading the adversary's objective, and attacking the machine learning itself.
- Testing, Trust, and Fielding Cognitive EW — why demonstrations stall, how a learning system is tested and bounded, and what fielding actually requires.