Until 2024 the standard answer to small drones was electronic warfare: jam the control link or the satellite navigation and the drone falls out of the sky. Then came fibre-optic FPVs, which trail a hair-thin optical cable instead of a radio link. They are immune to jamming, and they now make up a large share of FPV sorties. A defence built only from jammers has a hole in it that an adversary can drive a whole raid through, which is exactly what the EW-only comparison in this twin shows.
RF jammer / EW is cheap, wide and never runs out, but cannot touch fibre-optic or inertially-guided threats. High-power microwave defeats a whole swarm with one pulse and was first demonstrated against a fibre-optic drone in December 2025, but its reach is short. Laser DEW costs pennies a shot yet needs line of sight and dwell time, so saturation beats it. Guns are cheap and very short-ranged. Kinetic interceptors reach furthest and defeat anything, with the smallest magazine and the highest cost per shot. No single one is an answer; the mix is.
We Monte Carlo the raids — hundreds of attacks, each a set of drones of sampled type on sampled approach corridors, some diving into valleys to break line of sight against real elevation data. Siting then maximises expected drones defeated across the whole ensemble, before they cross the standoff ring:
One binary per candidate site and effector type, sparse geographic couplers between options whose footprints overlap on a threat, and a budget constraint. That is the shape that embeds well on today's quantum annealers, unlike dense routing problems — the same formulation North Wales Police piloted with D-Wave for police vehicle placement, made stochastic and heterogeneous.
Coverage drives the search because set operations are fast. The headline numbers come from a timed engagement simulation instead: magazines deplete, effectors cool down between shots, microwave pulses catch everything in the beam at once, and anything still flying at the standoff ring is a leaker. The gap between the two is magazine depth, and it is invisible to a pure coverage model.
The inner siting problem is QPU-native today. The outer Monte Carlo over raid scenarios is what quantum amplitude estimation accelerates: estimating a rare leakage probability to error ε needs about 1/ε² sampled raids classically but about 1/ε quantum queries. Rare, high-consequence raids are precisely what matters here, which makes this a natural early fault-tolerant research track.
Terrain from the national DEM tile pack used by our driving-simulation programme. Threat and effector parameters are representative open-source figures for illustration, not a specification of any fielded system. Rhumb's role is planning and simulation — siting studies on a twin — not effector control.