Deploying autonomous mobile robots (AMR) – how to start without overpaying or stopping production
An AMR pilot starts with one area, one route and one measurable transport task. The full deployment has four stages: pilot, scaling, optimisation and preparation of maintenance. The largest risk rarely lies in the technology. Usually the blocker is the plant infrastructure, the traffic organisation or an unprepared team. Below is the order of the steps, together with everything you need to measure before the first robot starts.
Table of Contents
When AMR make sense and when to drop the idea
AMR pay off where the material flow is repeatable and occurs many times per shift. The flow must also tie up employees in tasks that rules can describe. For occasional transport, or transport that needs constant human intervention, a forklift with an operator is cheaper.
Eight values to measure before the decision. Each one either justifies the project or rules it out:
| What to measure | Why it decides |
| Number of trips per shift | the main multiplier of the saving; below a certain threshold the robot stands idle |
| Distance covered by the operator | the second multiplier; a short route gives a small saving per trip |
| Duration of a single transport | the basis of the baseline and of the comparison after deployment |
| Type and mass of the load | determines the robot type and the load handling method |
| Route variability | a fixed route points to AGV, a changing one to AMR |
| Number of people tied up in transport | shows how many working hours you can release |
| Errors and downtime | a cost that is usually not counted, and often exceeds the cost of the hours |
| Traffic conditions in the plant | pedestrian and forklift traffic determines the real throughput of the route |
The profitability threshold comes from the data of the specific plant, not from a universal figure. The pre-deployment analysis defines two figures: the minimum distance and the minimum number of trips per shift. At those figures, AMR start to compete on cost with an operator.
If the analysis shows the flow is too small, several alternatives remain. These are a layout reorganisation, a shorter route, a conveyor, or a solution with more fixed travel logic. Advising against automation at this stage is cheaper than a suspended pilot.
AMR or AGV – when the route is fixed
If the vehicle travels the same unchanging route at all times, automated guided vehicles (AGV) also belong in the comparison. AMR gain the advantage where routes, pick-up points or the plant layout will change. The system then has to react to obstacles and to the current situation in the space. A full comparison of both technologies is a separate subject. For a first deployment, what counts is the match to the actual material flow.
Pilot – one area, one route, measurable goals
The first pilot should be simple: one route, for example warehouse → production, and one repeatable type of load. The task of the pilot is not to show every capability of AMR. The task is to answer one question: does the technology work in the conditions of this plant, and does the result justify scaling?
The baseline must be established before the robot starts. Without a baseline you cannot prove later that the investment improved the process.
| KPI | Unit | How to measure |
| Transport time of one pallet | minutes, with the spread stated | measurement on the route before and after the start |
| Transport time per shift | hours | working time records |
| Delivery address errors | percentage of transports | error register |
| Downtime from missing material | count and duration | production reports, WMS/MES |
| Acceptance of the solution by the workforce | no unit | weekly conversation with the operators |
Every one of these indicators needs a value measured before the robot starts and the same value after deployment. Without the “before and after” pair, none of them is evidence.
Transport time
Measure the spread as well as the average. The robot is often valuable not because it is faster, but because the time to complete the task becomes repeatable.
Time released to employees
Cutting the number of trips is not yet a benefit. The benefit is what employees do with the time they gain. Examples are quality control, machine operation, changeovers, order picking, material replenishment, and tasks that need operator experience. The difference between transport hours before and after deployment needs a stated destination. Otherwise the saving exists only in the presentation.
Number of errors
Four errors are worth measuring before deployment: material delivered to the wrong place, transport skipped, material missing at the workstation, and load damaged. Each needs a baseline value and a result after deployment. Only then can you see whether automation reduced the problem or merely moved it.
ROI
ROI is not an assumption written into a presentation before the project starts. ROI is the result of comparing the investment cost with the actual savings. Those savings are working hours previously spent on transport, the cost of errors, the cost of overtime, higher throughput, and less downtime from missing material. The calculation method is below.
Employee acceptance
The fifth indicator has no unit. In the first weeks, do not ask operators whether “the robot works”. Ask in which situations the robot makes their work harder. The most common concerns are safety, organisational change, and the effect of automation on jobs. The team must therefore know three things before the start. Why AMR are being deployed, which tasks they will take over, and how the operators’ work will change.
Automated or traditional warehouse – which to choose? Check it!
Scaling the fleet – order of steps, not number of robots
A successful pilot is not a reason to order a dozen robots at once. It is safer to add robots one area or one shift at a time. After every step, check how the system behaves under a higher load. Every scaling step is another experiment, not a purchase.
Six problems that are invisible with a single robot:
- traffic conflicts between robots,
- queues at the pick-up points,
- overloaded charging stations,
- competition for the same drop-off locations,
- a higher number of interactions with pedestrians and forklifts,
- insufficient throughput of the infrastructure.
A larger fleet needs a fleet manager: a fleet management system that controls tasks, priorities, robot status and the utilisation of individual units. The number of robots itself follows from the actual transport demand, the travel time, the loading and unloading time, and the charging time.
What the quotation does not show – infrastructure and integration
The robot is one element of the deployment, and often not the largest. Before the purchase, you check whether the plant is ready for autonomous transport.
| Audit area | What to check |
| Floor | condition and flatness, thresholds, differences in level |
| Transport routes | aisle widths, doors, gates, lifts |
| Charging | space for charging stations, available power supply |
| Network | availability and quality of Wi-Fi coverage across the whole route |
| Traffic organisation | pedestrian and forklift traffic, conflict points |
| Process points | location of pick-up and drop-off points |
The area most often skipped is the IT infrastructure. If AMR are to work with WMS, ERP or MES, five points must be settled before the project starts:
- which system is the source of the transport order,
- how the order reaches the fleet system,
- how the robot receives the pick-up and delivery point,
- how the system confirms that the task is complete,
- what happens in the event of an error or an interrupted transport.
Add to this the effort on the plant’s IT side. Even when the supplier performs the integration, plant employees provide system access, tests, configuration and acceptance of the solution.
For the person who decides on the budget, the cost structure in percentages is more useful than the price of the device. It shows what share of the budget goes to robots, what share to integration, and what share to preparing the infrastructure.
Safety and formalities
AMR reduce part of the manual transport and forklift traffic, but safety does not arrive with the purchase of a robot. Before each route starts, you assess the actual traffic conditions and the places where conflicts can occur.
The minimum before the start: a risk assessment of the route, marking of the required zones, and preparation and delivery of training. Add clear rules for sharing the space between robots, pedestrians and other vehicles.
Responsibility divides between two sides. The supplier is responsible for the technical solution and its documentation. The plant is responsible for organising safe work in the specific hall. Depending on the application, the relevant legal requirements and safety standards apply, including ISO 3691-4 where it applies to the solution in question.
Optimisation after the start – what to check and how often
The start of the robot does not end the project. In the first weeks after the start, the fleet is a system that needs tuning, not a finished solution. Three situations force a correction most often.
Peak traffic in the aisles. A test at low traffic does not reveal the problems that appear during a production shift. The fleet must be observed during the hours of heaviest pedestrian and forklift traffic. That is when the real problems with priorities, stops and throughput emerge.
Layout change. Moving racks, a machine or a workstation changes the travel conditions. Every major layout change is the moment to check the routes again.
New product. A new load format can mean different pick-up points, a different drop-off method, or a completely different route. These scenarios are checked before larger production starts.
A fixed review rhythm after the start: a weekly operational review and a monthly review of fleet parameters. You can correct speed, routes, task priorities, the charging schedule, waiting points and right of way rules.
System reports alone are not enough. A five-minute meeting with the operators once a week reveals problems that statistics do not show. Examples include a robot stopping regularly at one specific place, or operators bypassing the planned pick-up point.
Maintenance and competences in the team
With an AMR fleet, the downtime of a single robot is not always a problem. The downtime of the whole transport system affects production quickly. Two paths are therefore prepared before scaling: planned inspections and an emergency procedure.
Planned inspections. Scope and frequency follow the manufacturer’s documentation and the conditions of the specific plant. The list covers the condition of wheels and drive components, batteries, safety systems, sensors, mechanical parts, connections and communication, and the error history. For wear parts, the service life stated by the manufacturer is useful: the number of battery cycles and the mileage of drive components.
Emergency plan. The procedure answers four questions. Who notices the failure, who takes the first action, when the service team is called, and what production does until the problem is resolved. Add the response time recorded in the service contract, counted in hours rather than working days. The operator must know which actions are permitted and which are not.
More than one “robot champion”. If the competence for basic diagnostics sits with one person, that person’s holiday becomes an operational risk. At least two people must be trained, and they must be able to cover for each other. The training covers basic diagnostics, response to alarms, and safe stopping and restarting of the robot. It also covers use of the fleet management system and the escalation procedure to the service team.
What did not work first time
Three types of setback recur most often in deployments. A conflict with pedestrian or forklift traffic, a problem arising from the plant infrastructure, and a problem that appeared only after the number of robots increased. All three can be captured during the pilot, provided they are documented as they occur.
Documenting a setback is only useful when it contains four elements: problem → cause → solution → result. The entry “route optimised” says nothing. Compare it with this entry: “the robot stopped in the narrow passage by the packing area, because the protective field overlapped the drop-off location; the drop-off point was moved out of the protective field; the stops disappeared”. That is a ready procedure for the next area.
How to calculate the ROI of an AMR deployment
The annual benefits of automation are set against the full investment cost, not against the price of the robots alone.
| Benefit side | Data source | Investment side |
| working hours removed from transport | working time records | AMR fleet |
| fewer errors | error register | fleet management system |
| less overtime | working time records | integration |
| less downtime from missing material | WMS / MES, production reports | infrastructure |
| higher throughput | production reports | chargers |
| employees released to higher value tasks | process observation | training, commissioning, service |
If the amounts are confidential, budget shares are enough: robots, fleet management system, integration, infrastructure and remaining costs as percentages.
Simple payback period: investment cost / monthly value of the savings achieved.
The most important parameter is not the number of robots. It is the actual transport volume. For every project you check at how many trips per shift the investment starts to produce an economically justified effect. That single condition is more useful than the statement that AMR “can pay back quickly”.
Summary
Three quantities describe the optimal start of an AMR deployment. The length of the pilot, the number of robots at the start, and the assumed payback period. All three are calculated for the specific plant, not taken from a universal table. The order of the steps does not change. First prove the operation on one route, then increase the scale in stages, then optimise the fleet regularly. Maintenance is prepared from the beginning, not after the first failure.
Do you want to check which material flows in your plant are suitable for automation? The first step is a material flow audit and a feasibility analysis of an AMR deployment.
FAQ – AMR deployment
Where does an AMR deployment start?
With an analysis of the material flows and the selection of one repeatable transport task. Before the pilot, measure the baseline: transport time, number of trips, number of errors, and how many employees are tied up.
How long does the first AMR deployment take?
The time depends on the project scope, the infrastructure and the integration. The plan lists process analysis, infrastructure preparation, mapping, configuration, tests and the pilot itself as separate items.
How many robots do I need at the start?
The smallest fleet that allows a reliable test of the process. The number follows from the task volume, the cycle time, the charging time and the required availability. It does not follow from how many robots fit the budget.
How do I choose the area for the pilot?
One route and one repeatable task with a high number of trips. A good candidate has clearly defined pick-up and delivery points, and a measurable problem for automation to solve.
What goals should the pilot have and how do I measure them?
The minimum is four KPIs: transport time, time released to employees, number of errors, and ROI. Every indicator needs a baseline value measured before the robot starts.
What should I do if the pilot misses the set KPIs?
Establish whether the problem comes from the technology, the process, the infrastructure or a badly chosen KPI. Then apply a specific correction and repeat the measurement. A missed KPI is not yet a failed project.
How quickly should more robots follow a successful pilot?
In stages, adding another group of robots for one area or one shift. After each stage, check whether new bottlenecks have appeared: queues, traffic conflicts, or charging problems.
Do I have to rebuild the floor or the plant layout?
Not always. Before deployment, check thresholds, floor flatness, aisle widths, doors, lifts and charging points. Sometimes a small change to the infrastructure is cheaper than designing the whole solution around an existing problem.
What Wi-Fi does an AMR fleet require?
What matters is the quality of coverage across the whole route, not at selected points in the plant. Network availability and quality is one of the audit areas before deployment, alongside the floor, the aisles and the charging stations.
Can AMR cope with narrow aisles and pedestrian traffic?
This needs an analysis of the specific route: passage width, traffic intensity, and the places where conflicts can occur. The result is verified during the pilot and the risk assessment.
Which safety standards apply to AMR working among people?
Depending on the application, the relevant legal requirements and standards apply, including ISO 3691-4 where it applies to the solution in question. Before each route starts, you perform a risk assessment, mark the zones and deliver training.
Will AMR integrate with our WMS or ERP?
They can, but the scope of work depends on the IT architecture. Before the project, define the source of the orders, the method of communication, the confirmation of a completed transport, and the handling of errors.
How often does an AMR fleet need tuning after the start?
Intensively in the first weeks after the start, then in a fixed rhythm: weekly operational and monthly technical. The parameters are analysed again after a layout change, a rise in traffic, or the introduction of a new product.
What happens when a robot stops or meets an obstacle?
The procedure defines who responds and what the operator may do alone. It also defines when to report the problem to the service team, and how production works during the downtime. With a larger fleet, the service response time recorded in the contract is added.
How many people do I need to run and maintain the fleet?
That depends on the fleet size and the scope of responsibility. Whatever the scale, more than one person is trained. Redundancy of competence reduces the risk of downtime when the main specialist is absent.
How do I convince the workforce to work with robots?
Involve the operators during the pilot and collect their comments regularly, for example at a five-minute weekly meeting. Communicate clearly which tasks the robot takes over and where the time gained is used.
At what transport volume does the deployment pay back?
There is no universal number of trips. The threshold is calculated from transport volume, operator working time, the cost of errors, the investment cost, and the required fleet availability. Calculate it for the specific material flow, before the pilot.
