Tell the fleet what to do. In English.
Simple to run
The person who owns the risk, a site manager or a control-room operator, configures the fleet directly and reviews what it's doing. No engineer in between. No ticket. No release cycle.
- Describe the patrol. "Go to the loading dock, face the roller door and take a picture." The platform turns it into robot actions and remembers what happened. In testing, a three-step instruction was carried out within 2 to 3 cm, three runs out of three.
- Change it in a sentence. Add a checkpoint, drop a zone, change the alert rule, from the control room.
- It knows what is on your site. The fleet learns the objects on a site, not just a list of named points, so "stand in front of the generator" and "see what that is" work without an engineer marking them up first. Proven on the real robot in a mapped room.
- Start from a template. Industry packs give the fleet a head start on the patrols your industry runs.
Takeover, two ways
- In plain language. "Go back to the east gate, get closer, show me that lock." The robot does it, with every safety check still in place.
- By joystick. Direct, millisecond control for the operator, on a path that bypasses the AI entirely.
Both are logged as corrections, so the fleet learns from exactly the moments it fell short.
Safe by design
Three outcomes, built into the architecture and the same on every robot in the fleet.
- The robot can always stop. Braking, balance and obstacle avoidance run on the robot in milliseconds, in proven controllers, with no AI and no internet in the loop. If the robot stops hearing from its operator or its site, it stops within about a second. An operator can still drive it.
- A human always outranks the AI. The emergency stop, then the operator, then the mission, then the AI, in that order. The emergency stop cuts power regardless of any software.
- It cannot do anything it hasn't been allowed to do. The AI can only use the robot functions it has been granted. Asked to "open the roller door and chase the intruder", it refused before the robot moved. "Never pursue" isn't a policy document; it's a function the AI doesn't have.
Rules of engagement
Never pursue. Alert first. Protect the guard. Log everything.
Plain-language control, takeover and the safety outcomes are proven in simulation, apart from the hardware emergency stop, which is validated on the first real robot.
Ready to see robots patrolling your site?
Before any hardware is bought, we can build a simulated copy of your site and show the fleet patrolling it at night while you configure the checks in plain language.