Problem
Industrial inspection is dominated by two failure modes: scheduled inspections that miss developing problems (because monthly walks can’t catch a slow leak or a creeping vibration shift), and unscheduled inspections that don’t happen (because climbing onto a 40-foot rooftop HVAC platform takes two people, a permit, and three hours). For HVAC systems, structural integrity, security patrol, and asset tracking, the gap between what we ought to be inspecting and what we actually inspect is enormous.
Fixed cameras solve part of this but only see what they’re pointed at. Stationary sensor networks (temperature, vibration, gas) catch known anomalies but miss anything not pre-instrumented. Humans cover everything but expensively, infrequently, and with safety risk for hard-to-reach areas — roof platforms, confined spaces, electrical substations.
Agitation
- Compliance pressure is rising. Regulators are pushing inspection cadence up (e.g., post-Surfside structural reviews in commercial buildings, OSHA enforcement on confined-space entries). Existing manual processes don’t scale to meet it without doubling inspector headcount.
- GPS-denied environments dominate. Industrial spaces — server rooms, warehouses, substations, ship interiors, manufacturing floors — almost always block or distort GPS. Outdoor drones that rely on GPS+VIO degrade fast indoors.
- Inspector safety. Roof access, confined-space entry, electrical substations all have non-trivial injury rates. Removing humans from the routine inspection loop is a measurable safety win.
- Cost asymmetry. A serious failure (HVAC compressor seizure, structural crack, undetected security breach) costs orders of magnitude more than the inspection that would have caught it. The economics favor frequent, cheap inspection over occasional, expensive ones.
Solution
A small indoor/semi-outdoor drone that flies an autonomous inspection route — taking photos, thermal readings, or sensor data at preset waypoints — and returns to charge. Key properties:
- GPS-independent localization. Layered ToF + optical flow gets sub-meter accuracy indoors without external beacons. Same hardware that makes delivery work makes inspection routes feasible.
- Reusable routes. Once a route is taught (manually walked the first time, or programmed via waypoints), the drone can run it on a schedule — daily, hourly, on-demand — without human supervision.
- Tight cost target. Sub-$1,500 BOM means a facility can deploy multiple units rather than rationing a single expensive industrial platform. Multiple units also enable parallel routes and redundancy.
- Sensor extensibility. The same airframe accepts thermal cameras, RGB cameras, gas sensors, or RFID readers depending on the inspection target. The MAVROS layer abstracts payload integration.
Concrete inspection targets
- HVAC: rooftop platform thermal scans, ductwork visual inspection, filter status checks
- Structural: weld inspection on industrial racks, crack monitoring on concrete columns, leak detection in basement plumbing
- Security: after-hours patrols of warehouses and yards, perimeter checks of fenced facilities
- Asset tracking: inventory scans in large warehouses (in conjunction with RFID payload), pallet-count verification in receiving bays
What we’re not claiming
We’re not replacing certified inspectors for regulated processes (boilers, pressure vessels) — the drone collects data; certification still requires a human signoff. We’re also not solving outdoor BVLOS (beyond visual line of sight) inspection — that’s a regulatory problem, not a technical one. For routine intra-facility inspection where the bottleneck is human cost and access risk, the economics work.
Where to go next
- Sensor stack — what we measure and how
- Last-mile delivery — adjacent application, shared hardware
- 2026 H2 roadmap — when payload integration features land