Autonomous site monitoring

A Massachusetts-based construction company managing active sites had an inspection problem every large builder recognizes: safety and progress inspections depend on someone physically walking the site, and a photo taken from a slightly different angle each visit is worthless for tracking how a structure actually develops.

Site walks are slow, episodic, and expose inspectors to the very hazards they are documenting. The company needed continuous, comparable, repeatable site intelligence, without adding headcount or putting more people in harm's way.

An autonomous quadruped monitoring system running DogOS, Notum's spatial-intelligence platform. The robot surveys and maps the site on its own. An inspector opens a tablet, selects an area of interest (stairwell 2B, a scaffold line, an excavation edge) and sets the schedule. From then on the robot navigates there unattended, frames the identical shot from the identical position, and delivers a photographic time series in which deviation is impossible to miss.

Between waypoints, DogOS's onboard reasoning layer watches the scene itself: flagging blocked egress routes, missing edge protection, and personnel in exclusion zones, and pushing findings to the inspector instead of waiting to be asked. A charging station closes the loop: missions launch, run, and recover with no operator on site.

DogOS is Notum's hardware- and software-agnostic autonomy platform: a spatial-intelligence and locomotion layer that turns commodity robotic platforms into autonomous workers. It fuses onboard sensing into a live spatial model of the site, plans and adapts its own missions, and pairs classical autonomy with near-real-time LLM-driven reasoning and recognition. The robot doesn't just record the site, it reads it.

Because DogOS sits above the hardware, the same stack deploys on an indoor Go2 Edu or a weatherproof industrial B2 carrying 40 kg of thermal, lidar, or through-wall sensing: whatever the site demands.