One robot, two buyers

A distribution centre at night has two sets of people who want to know what is happening on the floor.

The security manager wants to know that the yard gate is closed, the trailers are sealed, the high-value cage is locked and nobody is somewhere they should not be. The operations manager wants to know that the stock in each rack location matches what the warehouse management system says is there.

Usually those are two separate jobs, with two budgets, two teams and two sets of walking. A patrol robot already covers the whole floor every night. MoLogica’s design puts both jobs on the same robot, on the same circuit.

The security job

Logistics depots and warehouses are MoLogica’s first market, and the security problem has two layers.

Gate and yard. Trucks waved through. Seals logged as intact when they are not. Quiet back doors after dark.

Floor. Distributed internal theft in ones and twos, in dispatch, returns and the high-value cage, that rarely shows on any camera anyone is watching.

The robot answers both with randomised patrols and per-waypoint inspection. At each defined checkpoint it stops, faces the object and returns PASS or FAIL with a timestamped image and a reason:

Zone	Typical checks
Yard and docks	Trailer locked, seal intact, right truck in the right bay, roller door down
Floor	Cage locked, shelving intact, aisles and dispatch on randomised circuits
Perimeter	Gate closed, padlock on, fence intact

Routes are randomised after every dock, so nobody can time the gaps. Every verdict comes with the image it was made from, which gives the security manager a record produced by the robot rather than by anyone on the floor.

The operations job

The same pass through the racks is also a stock count, if the robot reads what it drives past. The design has four steps:

1. Read what’s there. The optical camera reads rack location labels, pallet labels and carton barcodes continuously as the robot moves. A bolt-on industrial barcode scanner handles dense or small labels.
2. Know what should be there. The site server integrates with the warehouse management system.
3. Reconcile. The system flags empty slots shown as full, the wrong SKU in a slot, short counts and pallets in the wrong bay.
4. Fill the gaps with vision. Where a barcode is hidden or a pallet is wrapped, vision estimates fill state and product type.

Exceptions go to the people who fix them by morning. Instead of a periodic cycle count that takes people off other work, the floor is checked on every patrol.

Why one robot works for both

The two jobs fit together better than it first appears.

Same route. Security wants every aisle covered on unpredictable routes. A stock audit wants every rack location seen regularly. Randomised routes are designed to pass every location over a night, just not in a fixed order.

Same sensors. The night-first optical and near-infrared cameras that photograph a cage door also read a rack label. The barcode scanner is an upgrade for logistics sites with dense labelling. Nothing about the security kit gets in the way of reading labels.

Same evidence. A security verdict is an image, a result and a reason. A stock exception is an image, a location and a discrepancy. Both land in the same log, in the same open format.

Same brain. Both are checks at waypoints, configured through the mission layer. The person who owns the risk sets them in plain language: “check the high-value cage, then read the labels in the returns aisle.” The industry pack for logistics supplies the checks, mission templates and rules.

Who pays for what

This changes the conversation inside a business. Security often carries the case for a patrol robot. Operations then gets a nightly cycle count on the same circuit. That is a lot easier to justify than two separate programmes, and it gives two departments a reason to want the robot to work.

It also changes how the robot is judged. A security patrol is often measured by what does not happen, which is hard to see. A stock reconciliation produces a list of exceptions every morning, which is easy to see. Both come from the same patrol.

The lone night picker

The two jobs meet most directly on the floor at night. A lone picker working the late shift is both an operations task and a security exposure. The logistics design includes a robot shadowing lone night pickers, with everything logged. The same robot that counts the stock is present while it moves.

How the checks get better

Inspection verdicts in MoLogica’s current testing come from a large cloud model judging each image. The design moves routine checks onto a small model that runs on the robot. In an overnight simulation run, the robot drove itself to 538 positions over four hours, collected and labelled images and trained a small model to 95% validation accuracy, answering in about 1 millisecond on a GPU. That run proved the pipeline on simulated images. Real checks for a depot will be trained on the robot’s own camera, on that depot’s racks, seals and cages.

Peak season

Distribution centres have busy periods. The service design includes short-term additional robots for holiday periods and stock-take, handed back afterwards. Because every robot runs the same software image and sensor kit, an extra body is flashed with the same image instead of being set up from scratch.

What is proven and what is not

Capability	Status
Randomised patrol with per-waypoint inspection verdicts	Proven in simulation
Self-docking and rotation	Self-docking proven in simulation
Plain-language mission configuration	Proven in simulation
Rack and pallet label reading	Designed
WMS integration and reconciliation	Designed
Lone-picker shadowing	Designed
Logistics industry pack	Designed; first generic checks proven in simulation

Where this stands

Patrol, inspection verdicts with timestamped images, docking and the training pipeline are proven in a simulated depot. Label reading, WMS reconciliation and the logistics pack are designed and not yet built. There are no live sites. The first physical robot is on the bench, and logistics is where pilots start.

Book a demo to see a depot patrol and its inspection log in simulation.