A cheap drone forces a response that costs a thousand times more, the operator is drowning in sensor feeds with too few good data points, and today's decision support is often thousands of hardcoded rules. DomeCommand is being built to change that, with a solver and an agentic harness instead.
Read the post
Inturai Ventures has signed an exclusive letter of intent to acquire the technology behind DomeCommand. A short note on what Inturai's sensors do, and why sensors that can be carried are the part worth paying attention to.
Most sensor fusion quietly assumes the inputs are clean and the sensors sit still. Neither is true. Here is how DomeCommand models a picture that is honest about what it does not know, and then sends a drone to go find out.
Sensor fusion sounds like addition, add up your sensors and get one clean answer. It isn't. Here is a concrete, tick-by-tick walkthrough of one drone seen by four sensors, and the one distinction that makes the whole thing work.
A command system for a real fight can't assume GPS, can't assume a radio link, and can't assume the drone is even smart. Here's how DomeCommand is architected to degrade gracefully: four loops, two links, and one policy that survives all the way down to a single drone flying solo.
The data you most need to build a counter-swarm system is the data you can never collect, so we generate it. A look at the simulation engine: scenario generation across fleet and threat catalogs, why determinism matters, and how we are using engagement history to train the planner.
Two jobs, and everything else is detail: work out where you are, then work out what the motors should do about it. A ground-up explainer built on the cascaded controller I wrote for Udacity's Flying Car Nanodegree.