Quadrupeds

The robot is the cheap part of an inspection deployment. What decides whether it earns its keep is the sensor package, the route and what happens to the recordings.

Start with what you are trying to detect, then work backwards.

A thermal camera finds a hot bearing or a failing connection. A gas sensor finds a leak. A microphone finds a change in a pump that a person walking past would not notice. Visual inspection finds gauges, valve positions, spills and open doors. Each of those is a different payload, a different weight and a different amount of the platform budget, and the D1 EDU carries an effective 5 kg in total. Decide the package first, because the package is what the platform has to serve.

Then the route.

An inspection round is worth automating when it is repetitive, frequent and dull, and when a human doing it is either expensive or exposed. Once a shift around a substation, a compressor house or a tank farm qualifies. Once a month does not: the robot will be uncharged and out of date when you want it. Walk the route yourself first and time it, including doors, gates and anything that needs a person to open it, because those are what the round will actually be limited by.

What the machine produces is data, and somebody has to look at it.

This is where inspection projects quietly fail. A quadruped that walks a perfect round and writes 400 images a day into a folder nobody opens has automated the walking and left the inspecting undone. Agree before deployment what gets flagged automatically, what threshold raises a ticket, which existing system the ticket lands in and who reads it on a Tuesday morning.

Recordings of a workplace also mean people.

A camera on a moving platform will capture staff, and how long the footage is kept, who can view it, whether faces are blurred and what the works council or staff representatives have been told are decisions to make before the first round rather than after somebody asks. It is straightforward when it is designed in and awkward when it is discovered.

Finally, the boring logistics that decide reliability: a charging point on the route or at the end of it, a way for the robot to get through a door it cannot open, and a plan for what happens when it stops halfway. Published ranges run from about 6 km to 13.5 km, which sounds ample until the round includes stairs and a return leg.