[gpm]

> because they need to have some kind of antenna to communicate externally

I mean, that's true until it isn't. Autonomous drones (otherwise known as weird looking missiles?) being used to attack a base with a fancy microwave air-defence system isn't really science fiction anymore.

I suppose cameras might be hard to shield too?

[version_five]

It would probably also need GPS. I think it should be possible to harden a camera. Light can go e.g. though an arbitrarily small honeycomb mesh, while microwave cant. But it might be limiting in terms of what a ln autonomous drone would need to see. Lidar might be harder

[gpm]

GPS is jammable, I'd assume that the base with the fancy microwave anti-air system is capable of denying it and you'd want to be capable without it anyways.

Guidance via camera has been a thing for a long time: https://en.wikipedia.org/wiki/TERCOM#DSMAC,_Digital_Scene_Ma... I don't know if it's available open source (or commercially), but I don't imagine it would take me (or any other reasonably skilled programmer) long to develop it these days. Satellite imagery is freely available from google maps if you have nothing better.

Near(ish) peer adversaries (the kind that can afford drone swarms) are definitely capable of the guidance part.

[AHOHA]

Drones can fly safely with gnss denied environments, a system we tested it was in a mine with a very narrow areas, and at a relatively high speed too, being open in the sky would be easier.