Skip to content
Autonomous patrol and inspection robots · Johannesburg, South Africa
Menu

Where we are and what's next.

Done, in simulation

  • Autonomous mapping and navigation to within 2 to 4 cm, re-planning around obstacles.
  • Plain-language missions carried out by the robot.
  • Per-waypoint inspection with timestamped images and reasons.
  • Safety outcomes: the robot can always stop, a human always outranks the AI, the AI cannot do anything it hasn't been allowed to do.
  • Self-docking.
  • End-to-end proof gate passed four times in a row on two robot bodies.
  • Checks trained from the robot's own patrols and answering on the robot in about a millisecond.

Done, on the real robot

  • Mapped a real room on its own and reached every stop on a course within about 2 cm.
  • Plain-language instructions carried out on the floor, including going to an object it had learned, not just a named point.
  • People seen and tracked by the on-robot camera.
  • Lost-link behaviour: hold, return home, report what it did, in the dark, on battery.
  • A digital twin of the room built from a phone scan, used as the test course before every change ships.

Now

  • The control room: your existing cameras and the fleet on one screen, and a camera alert that sends the robot to look. Proven in the twin first, then on the floor.
  • The production sensor pair fitted to the first robot.
  • The full proof gate on the real robot.

Next

  • Site onboarding from a phone scan with no hand steps: scan in, map, twin and checks out.
  • Real-camera inspection checks; on-robot checks replacing cloud calls where speed or cost demands it.
  • A server on site that plans missions with no internet.
  • A three-robot fleet on the same software.
  • Pilot sites: logistics first.

Later

  • Industrial wheeled perimeter fleets with long-range heat detection and centimetre-accurate positioning, for mines and farms.
  • Cold-rated indoor bodies with temperature logging.
  • Site AI server deployments.
  • Multi-robot handoff, multi-floor tracking.
  • A legged body for sites with stairs or rough ground, on the same software.
  • On-robot language models at every site.
  • Robot arms.
  • Photoreal simulation for training inspection checks.

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.