Before you order an AMR for high-bay storage — check 3 things

The decision to buy a mobile robot is usually made at hardware level: load capacity, speed, battery life, unit price. Yet the success of a deployment is decided by something that is not on the datasheet — the condition of the warehouse the robot is going to enter. Before you order an AMR for high-bay storage, check three things that determine whether automation will speed up your hall or merely expose its mess faster.

Automation does not fix a warehouse — it reads it

This is well illustrated by the high-profile case of the distributor Southern Glazer's, which rolled out drone-based inventory counting across nine distribution centres. After around 5,000 flights, the drones detected nearly 35,000 inventory discrepancies, and the company moved from quarterly counting to a two-week cycle (Modern Materials Handling).

It is worth pausing on that figure. The 35,000 discrepancies did not arise the moment the drone started flying — they had been in the warehouse all along; nobody could see them between one stocktake and the next. Automation neither created the problem nor solved it. It did one thing: it shortened the time after which a discrepancy becomes visible, from a quarter to two weeks. This is the essence of thinking about robotising high-bay racking — the machine is a sensor, not a cleaner. The clearer the space you give it, the shorter the response time to what it detects.

From this perspective, buying an AMR without preparing the hall is an investment in a very expensive diagnostic tool. The robot will drive, scan and report — and leave you with a list of exceptions that the process cannot handle. That is why the order matters: first sort out what the robot is to "read", then let it in.

The first thing: zone boundaries and traffic logic

AMRs move according to a map and rules, not according to an operator's intuition. In high-bay storage, where aisles are narrow and vertical movement intersects with horizontal movement, ambiguous hall geometry turns into downtime and collision situations.

Before the first robot enters, walk the hall with three questions in mind. Where are the crossings, i.e. the places where the AMR route crosses pedestrian or truck traffic? Where are the buffer zones for setting down and picking up, and do they have a physical, visible boundary? Do the aisles have a fixed, respected width, or do they narrow by half when someone leaves a pallet "for a moment"?

A firm, unambiguous boundary on the floor works here as a rule understood both by people and by the robot's traffic logic. Durable floor marking tapes defining walkways and buffer zones, together with mobile barriers organising the space in contested areas — for example ZonePro mobile barriers — reduce the number of exceptions before automation even arrives. Less ambiguity in the geometry means fewer stops in the work cycle.

The second thing: legibility of locations

A robot taking inventory of high-bay racking reads locations from different angles and distances — from floor level, from a height of several metres, at an angle, often in the uneven lighting of the storage zone. A label that is legible to a person standing a metre from the rack is not necessarily legible to a camera looking up at a sharp angle.

What to check before automated scanning

• Consistency of the addressing system. Every location described according to a single logic, with no local exceptions "because it was easier that way".
• Visibility from multiple angles. The label positioned so that it can also be read from the level at which the robot's sensor looks, not just the operator.
• Durability and contrast. Markings resistant to scuffing by trucks, with no faded codes that automation will read as an error.

This is the most frequently underestimated element of preparation. Well-organised rack and floor labels are not cosmetic, but a data layer that the robot uses on every pass. Illegible addressing will not stop a deployment, but it will turn every real discrepancy into a wave of false alarms that someone has to verify manually.

The third thing: responding to exceptions

Automation generates exceptions — that is its normal mode of operation, not a failure. A drone or AMR reports: "location A-12-4 should hold 6 pallets, there are 5". The question is what happens in your hall in the minute following such a signal.

If the answer is "well, someone will check at some point", automation will not work, however good the hardware. Discrepancies will build up faster than the process can close them. That is why the third element of readiness is a simple, written exception-response scenario:

1. Who verifies the reported discrepancy — a specific role, not "the warehouse".
2. Within what time — a response on the same day or within the current shift.
3. How the problem zone is marked, so that it is visible to the whole shift until it has been resolved.

A physical, unambiguous marking of the zone under investigation closes the loop between the signal from automation and the human response. Elements that are visible regardless of lighting work well here — barriers and rack guards fencing off the disputed location, frames for corrective documents or photoluminescent markings in zones with variable light. The point is that a discrepancy detected by the robot should not disappear into the system, but remain visible until you close it.

Readiness checklist for AMRs in high-bay storage

Before you sign the order, walk the hall with this list. If your answer to any of the questions is "it depends", you have work to do there before deployment, not after it.

• Do the aisles, crossings and buffer zones have a fixed, physically marked boundary?
• Is location addressing consistent throughout the hall, with no local exceptions?
• Are rack labels legible from the sensor's perspective as well, not just the operator's?
• Is there a written rule: who checks an exception, within what time and how the problem zone is marked?
• Is the zone under investigation visible to the whole shift until the discrepancy has been closed?

The order is simple and consistent. First, unambiguous boundaries and locations, then automation. An AMR will not fix an illegible warehouse — but it will very quickly show where your mess is. It is better for it to show this in an organised hall, where every discrepancy has an owner and a response time, than in a hall where 35,000 reports land in the system without a response.

FAQ

Does an AMR for high-bay storage make sense without first organising the hall?

The robot will drive and scan regardless of the state of the warehouse, but without clear boundaries and locations it will mainly return false alarms and exceptions that the process will not close. The result is an expensive diagnostic tool instead of real automation. Preparing the space is a condition for a return on investment, not an add-on.

Where should you start when preparing a warehouse for robotisation?

With three layers: traffic geometry (zone boundaries, walkways, crossings), legibility of locations (consistent, durable addressing visible from multiple angles) and the exception-response process (who, within what time, how the problem zone is marked). Each of them can be improved with marking and barriers before the first AMR even enters the hall.

Sources

• Modern Materials Handling, "Southern Glazer's ramps deployment of Corvus Robotics drone inventory system to nine DCs" — mmh.com