The robot that works next to you

A traditional industrial robot is fast, strong, precise and dangerous. It lives behind a fence with an interlock on the gate, because a machine that moves a twenty-kilogram tool at two metres per second will not notice a person. The fence is not a detail. It is often a third of the installation cost and the reason the machine cannot be moved.

A collaborative robot is built around the opposite constraint. It limits force and speed so that contact with a person is not an injury, which lets it stand on a bench next to that person with no cage at all. It is slower and weaker than its caged cousin, and it wins anyway, because it can be installed in an afternoon.

What the safety promise actually means

Collaborative is a property of an application, not of a robot. A cobot holding a blade or a hot glue gun is not safe regardless of how gently it moves. The relevant standard sets pressure and force limits per body region, and a compliant installation requires a risk assessment of the whole cell: the tool, the workpiece, the speed and what happens if the part is dropped.

In practice most deployments run in a mixed mode: full speed while nobody is near, reduced speed when a person enters a zone, stop on contact. A laser scanner or a light curtain does the detecting, which is cheaper than a fence and far more flexible.

Where they earn their cost

The jobs that pay back fastest are the ones people dislike and machines find easy: machine tending, palletising small boxes, screwdriving, dispensing adhesive, sanding, and quality inspection with a camera. Each is repetitive, ergonomically unpleasant and tolerant of a robot that takes twice as long as a person.

The pattern in small firms is worth noting: the cobot rarely replaces anyone. It runs the second shift on a machine that used to sit idle after four in the afternoon, which turns a capital asset that was 50 percent utilised into one that is 85 percent utilised. That, not labour cost, is usually where the payback comes from.

The three ways a cobot project fails

The part presentation problem. A robot needs to know where the part is. If parts arrive in a heap, someone has to fix that with a feeder, a fixture or a vision system, and that is frequently more expensive than the robot. Budget for it from the start.

The cycle time trap. A cobot at reduced speed can be slower than the person it was bought to help. If it sits in the critical path of a line, the line slows. Cobots belong in parallel work, or on shifts where the alternative is nothing happening at all.

The nobody-owns-it problem. The cell works for two months and then the product changes. If no one on the shop floor can reprogram it, it becomes an expensive shelf. The firms that get value train two of their own people properly, and that training is the cheapest line item in the whole project.

What it costs, honestly

A 10-kilogram arm is in the region of 25,000 to 40,000 euros. Add a gripper, a stand, safety scanners, fixtures and integration, and a realistic first cell lands between 50,000 and 90,000 euros. Maintenance is a few percent of that a year. Our automation payback calculator lets you enter your own numbers, and the slider that matters most is not the wage but the utilisation.

At two shifts, 250 days, a loaded labour cost of 38 euros an hour and 70 percent of a position covered, a 90,000 euro cell pays back in roughly a year and a half. At one shift and 40 percent coverage, it takes five, which is longer than most owners will tolerate.

What to watch

Watch for arms sold with a learned grasping policy included rather than a teach pendant, because that removes the part presentation problem, which is the real barrier. And watch the leasing offers: a monthly rate turns a capital decision into an operating one, which is how this technology reaches firms with ten employees.