ALS lights a warehouse aisle only when a person is actually in it, telling a human from a robot where a plain motion sensor could not.

ALS (Automatic Lighting System) is one of Optense's products, and the most focused of them. It came out of the same relationship that produced ICS. After we piloted with Unilever, the engineering team at one of their personal-care facilities came to us with a small, specific problem, and ALS is the answer to it. It watches a warehouse aisle and turns the lights on only when a person is genuinely there. I built it in about three weeks, and it leaned heavily on two things I already had in hand, a working knowledge of Modbus from ICS and a person-detection model from SCM.

The facility had already tried to automate its aisle lighting with a motion sensor, and it kept getting fooled. The warehouse runs AGVs, Automated Guided Vehicles, driverless robots that move goods around a warehouse floordriverless robots that ferry goods around the floor, and a plain motion sensor cannot tell one of those from a person. So the lights would switch on for an empty aisle every time a robot rolled through, which defeats the point of automating them at all. What they needed was something that could answer a narrower question. Is there a human here, specifically.

ALS is, at heart, a single-purpose person detector. A camera watches the aisle, the model looks for people and nothing else, and when it finds one, ALS sets a state on a relay over Modbus A long-standing protocol for communicating with industrial sensors, relays, and controllersto switch the lamp on. When the aisle empties, a thirty-second debounce A short delay that holds a state briefly before it changes, to stop rapid on-off flippingholds the light on before it turns off, so a worker who steps out of frame for a moment is not left in the dark. That is the whole loop. The restraint is the point. It does one job, and it does it without guessing.

I did not have a hard time building most of it. The one genuine challenge was distance. The warehouse is big enough that the camera has to spot a person up to forty meters away, and at that range a human barely registers as a smudge of pixels, far too small for the model to catch reliably. I solved it with resolution and tiling. Splitting a frame into a grid of smaller pieces so a model can detect small, distant objectsI run the camera at a high resolution, then slice each frame into a grid of smaller tiles and run detection on each tile on its own. A person who is tiny in the full frame becomes a reasonable size inside their tile, so the model can actually see them. Stitched back together, the aisle is covered end to end.

ALS does two things for the facility, and safety is the one that matters most. A lit aisle is a safe aisle for anyone walking it, and the system makes sure the lights are on whenever a person is present. The second is energy. I do not have exact figures for the site, and a very rough estimate puts a single aisle's lighting at around 768 kilowatt-hours, so lighting only the aisles that are occupied adds up across a whole warehouse. ALS also sits naturally alongside ICS. It is really vision deciding an action over Modbus, which is exactly the pattern ICS is built to run, so in time it could live inside ICS rather than stand on its own. For now it is the simplest thing Optense makes, and sometimes the simplest thing is the one that just works.