What an AMR will not forgive: 4 things to sort out beforehand

A mobile robot does not create order — it reveals it. If flows, addressing and data on the shop floor are run on improvisation, an AMR will expose this faster and more starkly than any audit. That is why preparing processes for AMRs starts not with choosing a vehicle, but with sorting out four things that autonomy does not forgive: material flows, addressing, location data and process discipline. This article breaks each of them down and ends with a practical checklist.

A lesson from the BMW pilot: data first, autonomy second

When the news broke that BMW was testing humanoid robots on the line, the robot itself was the loudest story. The most interesting part of the story, however, is something else entirely. Before the manufacturer added a layer of autonomy, it spent several years standardising data models and feeding operations from digital twins, quality systems and intralogistics platforms. The order was unambiguous: first order in the data and the process, and only then the machine that works on that order. This was reported by the editors of Automatyka.pl.

This reverses the intuition with which many plants approach the automation of transport. The natural reflex is: let's buy a robot, and it will sort out the transport runs. In practice, it is the other way round. An AMR works on the same data that today circulates in supervisors' heads, on sheets of paper by the racks and in inconsistent spreadsheets. If that data is scattered, the robot will not build consistency out of it — it will execute it literally and quickly expose every gap. The four areas below are exactly the places where gaps hurt the most.

The first thing: material flows that can be described in one sentence

An AMR runs repeatable routes. Its value grows where a transport run looks the same today, tomorrow and in a month's time: goods receipt – set-down area, warehouse – line, WIP – next operation. The problem starts when the flow depends on someone's decision on the day — "today we go this way because it's more convenient". This kind of variability is invisible to people who know the shop floor, but for a robot it means there is no rule that can be executed.

Preparing this area means mapping the real flows — not the assumed ones, but the ones that actually happen. For each point A – point B relationship, three things must be established: what is transported, how often and in what time windows. Only such a map shows which routes are stable enough to automate and which first need to be standardised. Physically organising walkways and workstations often goes hand in hand with modular infrastructure — pipe and joint systems make it possible to quickly build and rearrange set-down stations and trolleys so that the flow is repeatable rather than improvised.

The second thing: addressing, or a single language for locations

A robot does not "guess" where rack 12 or the buffer zone by gate 3 is. It needs an unambiguous address. In many plants, addressing exists, but in several parallel versions at once: one designation in the WMS, another on the rack label, yet another in the shift's everyday speech. A person will connect these variants from context. An AMR will not.

Sorting out addressing means bringing all these layers down to a single standard: the same location identifier in the system, on the floor and on the rack. The label must be legible from a distance and from different angles, because the robot and the operator look at it from different perspectives. This is where it is worth using floor and rack labels designed for the production environment — durable, high-contrast and consistent throughout the hall. Addressing is the foundation, because every subsequent layer depends on it: the route, the set-down point and the material status all refer to the location.

The third thing: location data as a single source of truth

This extends addressing into the digital dimension. A label on a rack is not enough if three different location registers are running in the background and do not agree with one another. Scattered data is the most common pitfall of AMR deployments — the robot retrieves a location from the system, travels according to it, and on arrival finds a different reality because the register was not up to date.

Preparing this area means designating a single source of truth for locations and synchronising the others with it. This is not about implementing an expensive platform — it is about deciding which system is the master and about the discipline of keeping it up to date. Set-down points, buffer zones and routes must be reflected in the data before automation starts to use it. It is precisely this stage that took BMW years, and it is what distinguishes a deployment that scales without friction from one that turns into firefighting after a week.

The fourth thing: process discipline, or the "always the same way" standard

The hardest of the four, because it concerns people, not tables. An AMR requires the same operation to be carried out identically every time — the same set-down point, the same sequence, the same place for the pallet. The rule of "put it wherever there happens to be space" is convenient for the operator and fatal for repeatable transport. The work standard cannot exist only in a written procedure; it must be enforced on the shop floor and visible at a glance.

In practice, this means clearly defining zones: where the set-down area ends, where the walkway begins, where the buffer lies. The faster such a zone can be defined and rearranged when the process changes, the easier it is to maintain discipline without permanently freezing the hall layout. Mobile solutions such as ZonePro barriers make it possible to organise zones quickly and repeatably, so that the "always the same way" standard has a physical form and not just an entry in a document. Process discipline is a test of readiness: if a new person cannot understand the rules of the hall within a few minutes, the robot will not "understand" them either.

Preparing processes for AMRs — where to start

The four areas do not have to be completed at the same time. A sensible order runs from data to execution: first establish a single language for locations, then base the flows on it, next synchronise the digital data and finally consolidate it all with discipline on the shop floor. The checklist below helps to assess where the plant really stands. Every "no" is an area to be sorted out before an AMR takes to the route.

• Internal transport flows are described as point-to-point relationships, with frequency and time windows.
• The routes intended for automation are stable, not changed from day to day by a shift's decision.
• Each location has a single identifier, consistent in the system, on the floor and on the rack.
• Location labels are legible from a distance and from different angles.
• There is a single master source of truth for locations, and the other registers are synchronised with it.
• Set-down points and buffer zones are reflected in the data, not just in practice.
• The work standard for repeatable operations is written down and enforced, not merely declared.
• Set-down areas and walkways are clearly defined and can be rearranged quickly.
• A new employee understands the rules for traffic and setting down material within a few minutes.

If most of the points are a "yes", the plant is ready to talk about vehicles. If "no" dominates, investing first in sorting out the data and the process will pay back faster than any robot — because an AMR accelerates an organised process, whereas in chaos it only makes the chaos more visible.

Summary

There are four things an AMR will not forgive: undescribed flows, inconsistent addressing, scattered location data and a lack of process discipline. These are not technology issues — they are organisational foundations that exist regardless of whether a robot appears on the shop floor. Sorting them out has value in itself, and when automation is deployed it becomes a condition for success. The BMW sequence is the best guide here: data and process first, autonomy second.

Does preparing processes for AMRs require expensive IT systems?

Not always. What matters is deciding which system is the master source of truth for locations, and the discipline of keeping it up to date. Many plants start by aligning existing registers and addressing, and only then consider expanding their intralogistics platform.

Where do we start if our addressing is chaotic?

With a single standard for the location identifier and bringing all the layers — the system, the floor markings and the rack markings — into line with the same version. Addressing is the foundation, because routes, set-down points and material statuses all refer to the location.

How can we check whether a process is repeatable enough for an AMR?

A simple test is to describe the transport run in one sentence that will be true every day. If you have to add "it depends", the process still needs standardising before it is put on the robot's route.

Sources

• Automatyka.pl, "BMW testuje humanoidy" (BMW tests humanoids) — https://www.automatyka.pl/wiadomosci-i-komunikaty/bmw-testuje-humanoidy-249626-10