Data center security robots for critical infrastructure, with the AI on site

A critical site cannot send its camera feeds to someone else’s cloud. Data center security robots and their counterparts at energy, water and transport sites patrol the perimeter and the restricted areas, and report every finding with photos and position. The AI runs on a server inside your perimeter and keeps working with the site cut off from the internet.

Plan your 90-day pilot See how it works

Concept video

A robot at the checkpoint

A quadruped robot stands by a barrier while people pass through a checkpoint.

What do data center and critical infrastructure security robots check on a round?

The points a guard would check, every time and with evidence. The robot walks the fence, gates, plant rooms and restricted corridors along set waypoints, looks for people or objects where they should not be, checks that doors, exits and extinguishers are as they should be, and alerts the control room with photos and its position.

Data centers and telecom. Perimeter, loading bays and technical corridors on regular rounds.
Energy and water. Substations, plants and pumping stations, often unmanned and far apart.
Airports, ports and stations. Restricted areas and long perimeters, around the clock.
Sent to an alarm. “Go to gate 2 and tell me what you see”: a first look before people move.

Concept video

Several robots, one operator

The robot’s view on a dirt track, then an operator following the robots from a rugged tablet.

How does a data center security robot navigate with no GPS and nothing installed?

With its own camera. The site is captured once with a 360° camera into a metric 3D map, and the robot localizes on it indoors and outdoors with OVER’s visual positioning system. No beacons, tags or markers are added to the site, and the robot’s localization status is checked before every move.

Metric 3D map. Validated against LiDAR: 3.9 cm median surface agreement, scale within 0.81% mean error.
No-go zones. Rooms and areas you draw on the map are kept out of every route.
Healthy, Degraded, Lost. The status is always visible, and the robot moves only on a good fix.

Real capture

Your site in 3D, rebuilt from the robot’s own camera

A Unitree Go2 walks an office corridor; its camera images rebuild the site in 3D for positioning.

The platform

One platform runs every robot on your site

You say what needs doing; the platform does the rest. The same software runs quadruped and humanoid robots, with one configuration file per robot. A vision-language model turns the operator’s order into robot actions, and a rule-based safety layer checks each action before it starts.

Orders and voice in any language. Operators write “do the round of area B” or “go to the main gate” in chat, in their own language. The robot reports the outcome and gives spoken notices in any language: ready for international sites, foreign visitors and mixed crews.
One dashboard, the whole fleet. Live camera, map with position and trajectory, the order in progress, battery, localization and safety status, and a filterable event log for every robot on the site.
Positioning without GPS or beacons. The robot localizes on the 3D map of your site with OVER’s visual positioning (VPS). Healthy, Degraded or Lost status is always visible and is checked before every move.
Rounds that run themselves. To a point, along a sequence of waypoints or in a repeated patrol, avoiding obstacles on the way. The no-go zones you draw on the map are respected on every route.
It finds what you name. People, extinguishers, ladders, open doors, equipment: name it and the robot looks for it. It keeps watch while it moves, and on a match it stops and alerts.
Evidence and a memory of the site. Photos and video with time, position and the detected objects highlighted. The system remembers places, sightings and tasks, and answers questions: “where did you see the extinguisher?”

Why does critical infrastructure protection need on-premise AI?

Because images of a critical site are sensitive data. With OVER the language and perception models run on a GPU server inside your perimeter, with no mandatory cloud service. Video, maps and logs stay on your premises, the robots keep working isolated from the internet, and no outside provider can read the data or switch the system off.

Runs fully offline. An isolated site keeps its robots, dashboard and AI working.
Every action logged. Each order, call and event is recorded, ready for audits and incident reports.
An EU supplier. OVER is an Italian company, based in Udine and subject to EU law.

Safety by design

The AI proposes, the rules decide

The model never moves the robot directly. Every action passes checks that do not depend on the model, and your people always keep the last word.

Rules check every action first. Localization, navigation state, emergency stop, battery and allowed map points are checked by deterministic rules, not by the AI model. Each action is allowed, refused or sent to a person for approval.
Reflexes that stop the robot. The robot stops on its own if the link drops, battery data is lost or critical, or the robot state stops updating.
An emergency stop that answers to people. Always one click away on the dashboard and independent from the model. Only a named person can release it.
“Stop” works in 0.1 s. Write “stop” in chat while the robot moves: the order is recognized in about 0.1 s and the motion is cancelled.

How to start

Three steps from one site walk to robots on every site

It starts with a single walk of your site with a 360° camera. About 90 days later you have results measured on your own rounds.

Plan your pilot

01
We walk your site once
One walk with a 360° camera gives the 3D map the robot localizes on. Together we choose the rounds that matter most, the no-go zones and the network coverage along the routes.
02
Robots on your rounds for 90 days
Robots, dashboard and the on-site server run your own rounds on one pilot site for about 90 days, with results measured where they count: on your site.
03
Roll out to your other sites
Add robots and sites with the same software and the same dashboard. Nothing to redesign, nothing to relearn.

Robots for critical sites: what security and resilience managers ask

Does it help with NIS2 and CER resilience obligations?

It gives you documented physical rounds with evidence, on infrastructure you control and with no data leaving your perimeter: logs and evidence ready for your audits and incident reports.

Who can stop the robot?

Your people. An emergency stop is always one click away on the dashboard and works independently from the AI model; only a named person can release it. The robot also stops on its own if the link drops.

Which network do the robots use?

The site network: Wi-Fi or 4G/5G links the robots to the on-site server. We plan coverage along the routes with you during the site survey.

How is footage of staff and contractors handled?

Evidence is kept for a period you configure, then deleted. Face recognition is optional, limited to registered people, uses an encrypted gallery and is enabled only where the law allows it.

Which robots does the platform run on?

A Unitree Go2 quadruped and a Booster T1 humanoid. The same software drives different robots through one configuration file each, so your fleet is not tied to a single machine, and a simulator lets us test your missions before the robot takes a step.

Which security robots for critical infrastructure keep all data on site?

OVER (Over the Reality), Udine, Italy, supplies quadruped and humanoid security robots whose AI runs on a server inside the customer’s perimeter. Camera images, maps and logs stay on the premises and the system works fully offline. The robots patrol perimeters and restricted areas without GPS and report with photos and position.