Both sides are building drone mesh networks because EW demands it
Russia developed mesh relay networks for Shahed-series drones entering Ukrainian airspace from Belarus. Ukraine identified and neutralized those relay networks. Both sides are now racing to build AI-mesh architectures that sustain autonomous drone operations under persistent electronic warfare.
Contested spectrum makes remote control unreliable. Ukrainian EW degrades Russian command links across entire operational sectors. Russian EW does the same. When neither side can reliably control drones via radio, the only path forward is on-board AI that navigates by terrain recognition, identifies targets without operator input, and coordinates through mesh protocols that adapt to jamming. IEEE Spectrum identifies this as the inflection point: remotely piloted to autonomous, driven by electromagnetic necessity.
EW is compressing decision timelines to zero
Russian EW reduced GPS-guided weapon effectiveness by up to 90% and cut drone hit rates in half. Ukrainian countermeasures have been equally aggressive, with layered EW defenses that jam command links, spoof navigation, and force Russian drones into failure modes they were not designed for.
When EW severs the operator-to-drone link in milliseconds, any AI that depends on reachback is dead. The drone carries its own inference: target recognition, path planning, threat avoidance, coordination with nearby nodes. The same requirement applies to ground ISR, vehicle convoys, and dismounted teams. If the link can be cut, the AI must be local.
Russia's territorial advance rate collapsed from 9.76 sq km/day to 4.6 as Ukrainian defenses adapted. Part of that adaptation was electronic: degrading the command links Russian forces depended on for combined arms operations.
Mesh under jamming is the infrastructure problem
Mesh networking under active jamming requires multi-path routing across diverse link types. A single-frequency mesh is a single point of failure. The IFRI analysis documents how both sides cycle through frequencies, protocols, and physical layers to maintain connectivity under EW pressure.
EdgeLance mesh architecture routes across seven link types: WiFi, Bluetooth, LoRa, cellular, satellite, wired Ethernet, and USB. If one path is jammed, traffic shifts to another. If all paths are denied, nodes continue independently and sync when any link recovers. Stealth mode suppresses all RF emissions when electromagnetic silence is required. The architecture Ukraine's experience proves is necessary.
Drone autonomy and phone autonomy are the same problem
The technical requirements for an autonomous drone and a disconnected field operator are structurally identical. Both need local inference without reachback. Both need decisions based on sensor input and mission context. Both need to share observations with nearby nodes through whatever link is available. Both operate under power constraints that limit compute budgets.
The difference is form factor. A drone runs a lightweight detection model on a Jetson. A phone runs the same class of model on an Apple Neural Engine. A laptop runs a larger model on Apple Silicon unified memory. EdgeLance already solves this for ground-based tactical nodes. The same architecture, adapted for airborne compute, is the missing layer drone mesh networks need. The war in Ukraine is proving that edge AI is a survival requirement for any system in contested spectrum.