Lock-On After Launch: Machine Vision, Cheap Terminal Guidance, and the Export-Control Fight

For a decade, counter-drone defense rested on a bet: that a drone must remain connected to something — its operator's radio link, its satellite navigation — and that severing the connection defeats it. The bet built an industry of jammers and spoofers and shaped procurement on every continent. Machine-vision terminal guidance is breaking it. A munition that can see its target and guide itself through the final kilometers does not need the link the jammer exists to cut, and no amount of radiated power changes that.

The technology, at the class level, is mundane, which is exactly the problem. A camera, an edge-compute module running detection and tracking software, and a guidance loop closed against the airframe's controls: the operator designates a target — or the system proposes one — and from lock-on the munition flies itself. The components sit squarely inside the commercial drone economy; wartime reporting describes add-on autonomy modules in the low hundreds of dollars appearing on FPVs on both sides of the Ukraine war. The premium lineage is older: the Lancet's electro-optical terminal guidance demonstrated the concept years ago at a higher price point. What changed is that the capability slid down the cost curve into the expendable class, where it multiplies.

The operational effect concentrates in the terminal phase, and it compounds. Against jamming, an autonomous terminal leg removes the defense's cheapest layer precisely where it mattered most — the final kilometers, where the target has no time to reposition and the jammer no signal to attack. Against maneuver, a tracker updating tens of times per second outperforms a human operator working through a lagging video link, raising hit probability against moving vehicles. Against operator supply — the true bottleneck of FPV warfare — lock-on after launch lets a trained crew prosecute more targets per hour, because the cognitively hardest part of the engagement is delegated to the module. Each effect is incremental; together they describe a different weapon.

The training implications deserve a paragraph of their own. Manual FPV proficiency is expensive: it demands hundreds of simulator hours, steady hands under fire, and a talent pool that attrition consumes faster than courses replace. Terminal autonomy compresses that pipeline by moving the hard skill from the wrist to the designation decision — pick the right target, at the right moment, from the right angle — which is easier to teach, easier to test, and easier to standardize across a mass force. The operator does not disappear; the job changes from pilot to mission commander, and the ratio of effect to training hours shifts accordingly. Recruitment pools widen with it, since the designation task draws on skills far more common than precision flying. Armies that struggled to scale pilot pipelines will find this the most consequential change terminal guidance brings — more than any single engagement statistic — because it converts a talent problem into a training problem, and training problems are the kind that states can actually solve on a schedule.