Can AMRs and AGVs Operate on Mezzanine Floors? A Retrofit Decision Guide

Can mobile robots work safely on an existing mezzanine?

Yes, autonomous mobile robots (AMRs) and automated guided vehicles (AGVs) can operate on mezzanine floors when the structure, running surface, edge protection and material-transfer system are suitable for the complete application. An existing platform is not automatically suitable because it already carries shelves, people or pallet trucks. Approval needs to cover the configured vehicle, its load, its routes and the conditions in which several vehicles may occupy the same area.

For buyers evaluating AMR mezzanine floors, the first decision is whether robots will remain on one level, exchange loads with another level, or travel between levels themselves. That choice changes the structural review, lift requirement, fleet size, recovery arrangements and installed cost. It should be settled before a robot quotation becomes the project budget.

This article follows that decision through an existing-facility retrofit. It separates published industry examples from an illustrative planning calculation and explains which questions belong to the structural engineer, robot supplier, lift specialist and system integrator.

Why the discussion has moved from floor space to connections between floors

Recent industry material puts the interfaces between equipment and buildings into sharper focus. On January 26, 2026, Autoquip published a multi-level storage case study involving AMRs carrying pallets on vertical reciprocating conveyors, or VRCs. Its account describes mechanical restraint during travel and coordinated release through the wider control system. The case demonstrates an implemented architecture; it does not establish a standard throughput or return on investment for other sites.

On February 6, 2026, AMRA published TARS/AMRA-300:2026, its guidelines for robot-friendly environments. The public scope covers physical and digital building conditions, designated robot work areas, prohibited areas and human-robot interaction. That wider infrastructure perspective is particularly relevant when a robot route reaches a mezzanine opening or crosses a building-service boundary.

The fall-protection issue also has dedicated commercial attention. Mezzanine Safeti-Gates introduced RobotGate in February 2025 for robotic applications at VRC and elevator shafts. Its published arrangement keeps the shaft-facing gate closed until the lift is present and aligned. This is one supplier's solution, not a universal product requirement.

Together, these developments support a practical procurement question: can a particular building sustain the complete material flow? They do not support the assumption that adding autonomous navigation makes every existing elevated platform ready for automation.

Choose what crosses the floor boundary before choosing the robot

Picture one pallet arriving at ground-floor receiving and needing to reach a workstation upstairs. There are three substantially different ways to automate the journey. The distinction is useful for both AMR projects and AGV mezzanine floors; the navigation method alone does not decide which material-flow architecture is best.

Keeping robots on one level can simplify a first deployment

A packaging mezzanine might have enough repetitive horizontal movement to justify local automation without changing how materials arrive upstairs. This arrangement lets the buyer define a smaller project boundary. However, the business case must retain the labor and waiting time associated with the existing transfer process. A quotation that prices only the robot route should not be presented internally as a fully automated receiving-to-workstation solution.

Transferring only the load separates fleets but creates handoff work

With one fleet on each floor, the vertical equipment can move a load independently of the robot that delivered it. The receiving robot then collects it. This can avoid committing a vehicle to the entire vertical journey, but it introduces transfer positions and coordination between two fleets or fleet zones.

The buyer should ask what happens when the upper-floor pickup position is occupied. Does the lower robot wait while holding its pallet? Can it deposit the load in an approved buffer and resume work? Who owns the task while the pallet is between floors? These operational answers determine whether the apparent simplicity of separate fleets survives a busy shift.

Taking the robot upstairs preserves one mission but consumes lift time

A robot that accompanies its pallet can maintain responsibility for the load through the journey. The system nevertheless has to manage access to the vertical resource and its destination exit. The Autoquip case shows that this approach is possible. It does not imply that any goods lift can accept any autonomous vehicle.

Compare the three arrangements using the same destination demand and operating hours. Otherwise, one bidder may price horizontal transport while another includes the entire inter-floor process. Their headline robot prices will look comparable even though their delivered scope is different.

Read the existing platform as a structure, not a square-meter allowance

The most consequential question about mezzanine floor load capacity is what the documented rating actually covers. A uniformly distributed load describes a loading condition spread over an area. A loaded robot applies forces through individual wheels at changing positions. One description cannot simply replace the other.

Cogan's mezzanine design guidance distinguishes distributed loads, point loads and pallet-jack loads, and describes how forces pass from the deck through the supporting structure to its base. For a retrofit, this means the running surface and the steel frame are parts of the same review. A stronger overlay does not, by itself, increase the capacity of every supporting member or the slab below the columns.

Start with gross vehicle mass

Consider an illustrative purchase proposal with a 700 kg chassis, a 150 kg top module, a 100 kg carrier and a 1,000 kg product load. If those masses are separate and not already included in another figure, the configured moving mass is 1,950 kg. The product's mass alone represents barely more than half of it.

This arithmetic is only an inventory of mass. It is not a structural approval. The project still needs wheel positions and the relevant wheel reactions. If the supplier's vehicle figure already includes the top module, adding it again would also be wrong. Request a component breakdown with a clear boundary around every stated mass.

Ask the supplier to identify the loaded configurations that govern AGV wheel loads, including the effect of permitted load positions and vehicle maneuvers. Do not assume that four wheels each carry exactly one quarter of the total. The underlying payload definitions and stability questions are covered in our guide to payload and center-of-gravity limits.

Include where vehicles can accumulate

A route drawing showing one moving robot can conceal a queue of three loaded robots waiting outside a lift. It can also omit the receiving pallet, charging stations, a maintenance cart or stored goods added beside the aisle. The engineer needs credible simultaneous loading arrangements, not only a vehicle specification and a total floor area.

For procurement purposes, mark travel lanes, waiting positions, charging locations, transfer buffers and maintenance positions on the same drawing. Ask the engineer which combinations are acceptable and which require restrictions. If a solution depends on keeping a particular area clear, that condition must become an operational responsibility with a named owner.

Separate strength from serviceability

A structure may require assessment for both load-carrying capacity and movement under load. Mezzanine floor deflection matters to robot operation because transfer alignment depends on the relative positions of the deck, station and vehicle. Vibration or local movement may also require attention even when the structure satisfies its strength checks.

Hi-Level's March 2026 load-rating guide identifies concentrated loads, dynamic effects and vibration as relevant considerations for automation. It does not supply a universal acceptance limit for every robot. The project engineer and equipment suppliers need to agree on the applicable limits and measurement conditions for this installation.

When original drawings, alterations or material details are missing, put investigation into the feasibility budget. The appropriate next step is a competent structural assessment, not a loaded demonstration drive used as a substitute for engineering verification.

Walk the wheel path, including the last meter before each transfer

Once the structural concept is accepted, inspect the actual running surface. An installation of robots on mezzanine floors has distinct local interfaces: panel joints, lift thresholds, fixed transfer stands, charging contacts, repair patches and changes in surface material. A smooth-looking aisle does not establish that every interface is suitable.

For robotic mezzanine flooring, obtain the specified assembly rather than just a product name. ResinDek's MD product information, for example, gives different mobile-robot rolling-load figures for different supporting B-Deck configurations. The useful lesson is that flooring performance depends on the declared construction. A rating detached from its support arrangement is incomplete purchasing information.

Ask the floor supplier to identify the panel or deck, supporting arrangement, joints, fasteners, finish and installation requirements included in the proposal. Ask the robot supplier to confirm compatibility with the actual wheel material and intended duty. Neither party should have to infer the other's assumptions from a general sales brochure.

The detailed AMR floor requirements should then be checked along the intended routes. On a mezzanine retrofit, give particular attention to the transition between a relatively flexible floor and a separately supported lift landing or station. The loaded condition can differ from the unloaded survey.

Plan for repair as well as commissioning. If a damaged panel closes the only route to the upper-floor packing area, who can replace it, where is a spare kept, and what temporary material flow remains available? These questions belong in the purchase decision because maintenance access and alternative routes may be much more limited upstairs.

A useful handover drawing marks each inspected transition and the associated maintenance responsibility. This makes later repairs traceable without expecting the maintenance team to reconstruct the integrator's original route assumptions.

Treat every edge and opening as a separate application risk

An obstacle in a ground-floor aisle and an unprotected opening on an elevated platform have different consequences. A navigation map that tells a robot where to travel does not establish the performance of a fall-prevention system. Likewise, a pedestrian guardrail should not be assumed to restrain a loaded industrial vehicle unless that duty has been evaluated.

The OMRON LD-Series Integration Guide explicitly warns that poor localization can undermine map-based workspace restrictions and calls for physical barriers where hazards require them. That is manufacturer guidance for the identified product family, not a universal barrier specification. The broader purchasing lesson is to distinguish route planning from the protective measures credited in the application assessment.

Review perimeter edges, stairs, lift shafts, pallet-loading gates and temporary maintenance openings separately. Include the possibility of falling cargo and the people or equipment below. An opening that is normally closed may still be exposed during cleaning, recovery or maintenance, when ordinary production assumptions no longer apply.

The robot supplier, guarding specialist and integrator should define which protective measures address each scenario, how they interact, and how their performance will be verified. The evaluation should address the actual vehicle and load, rather than borrowing a barrier dimension from an unrelated robot manual or treating painted floor markings as physical restraint.

At lift landings, the project also needs a clear distinction between a request to open and confirmation that access is safe. The RobotGate example illustrates the relevance of lift presence and alignment. It does not establish that a particular installation has completed its system-level safety validation.

Use the existing guides to AMR protective-field engineering and application-specific AMR risk assessment for the wider methods. The mezzanine review must add its elevated edges, openings, falling-object exposure and recovery conditions to that work.

Specify a complete lift interface, not just an elevator call

For AMR elevator integration, receiving a successful call acknowledgment is only one part of the journey. The system must also coordinate arrival, access, occupancy, travel, exit and release of the shared resource. A valid message from a fleet manager does not by itself prove that the physical entry conditions are acceptable.

Confirm what equipment is being supplied

The terms goods lift, freight elevator and VRC are sometimes used loosely during early discussions. They should not be treated as interchangeable procurement categories. Establish the equipment classification, intended use, applicable local requirements, permitted occupants and maintenance obligations with the lift specialist. A material-only device must not acquire an assumed passenger or rescue function simply because a robot is stranded on it.

Then check the complete configuration: mass, platform dimensions, loaded swept envelope, approach direction, doorway clearance, threshold geometry and landing alignment. A payload rating can be sufficient while the doorway or turning space is not. Link the interface survey to the wider AMR site survey and lift-interface checks.

Assign ownership of states and faults

IDEC Factory Solutions describes inter-floor AMR installations involving collaboration with the elevator manufacturer, an external control interface and a coordinating control panel. Its examples show why the building equipment supplier belongs in the project early. A robot vendor's ability to exchange signals is not the same as permission to modify an existing elevator.

For the buyer, a practical interface review asks who provides and verifies each status: lift available, correct level reached, access condition satisfied, vehicle fully positioned, destination clear and transfer complete. Define what each party sees when information is missing, contradictory or delayed. The safety-related implementation and required performance must be established by the responsible specialists.

Do not leave restart ownership implicit. After a power interruption, the lift controller, robot and warehouse system may each hold part of the transaction history. Someone must define how they reconcile the physical load location and task state before normal service resumes. Otherwise, an operator may be asked to improvise a recovery across several supplier interfaces.

Protect the receiving side from becoming invisible

A lift can arrive correctly while its exit is blocked by a pallet or another robot. The delivery mission therefore needs a destination acceptance condition, not only permission to enter at the originating floor. Include upper-floor queues and emergency access in the layout review.

For factories sharing vertical equipment with people or other material flows, agree on operating modes and access rules with the appropriate specialists. Record how production priorities interact with maintenance and other building functions. Neither the robot project nor a throughput target should silently override those requirements.

Calculate the vertical bottleneck before adding more robots

The strongest commercial argument for multi-level warehouse automation is dependable material delivery. It is therefore useful to calculate the capacity of the inter-floor connection before discussing how many additional vehicles the fleet can accommodate.

Consider an illustrative two-level system moving one loaded robot upward at a time. Assume that reservation, entry, travel, exit and the empty lift's return to its starting condition together take 120 seconds. In this simplified case, the theoretical lift-only upper bound is 3,600 divided by 120, or 30 completed upward transfers per hour.

Now assume, for planning only, that 45 minutes of each hour are available to that flow after agreed interruptions and competing use. The equivalent capacity becomes 2,700 divided by 120, or 22.5 transfers per hour. A sustained requirement for 24 upward transfers per hour would exceed that planning capacity. Buying another robot would not resolve this particular constraint.

These figures are invented to explain the method. They are not vendor performance data or a recommended utilization target. The example assumes one load per cycle, includes the lift's return, and excludes a productive downward load. Empty-robot return trips are also excluded from this optimistic upper bound. A sustained-operation model must account for returning vehicles to the originating level and any additional lift occupancy this requires. A real bid should model the site's directions, batching possibilities, peak arrival pattern and equipment sequence.

Compare alternatives against the same demand trace. Transferring only a pallet might release the delivering robot sooner. Two lifts might provide additional capacity or recovery options. An upstairs buffer might absorb a short interruption but cannot correct a sustained capacity deficit. Each option has a different space, structure and investment consequence.

Ask the integrator to demonstrate queue behavior during a realistic demand peak and explain what happens after an interruption. Average cycle time alone can conceal a system that completes enough work over a shift but misses time-critical deliveries to a production line.

Plan the day a robot cannot leave the mezzanine

Failure recovery deserves its own drawing. Consider a vehicle that stops beside a transfer opening, a damaged pallet that cannot be collected automatically, or a lift unavailable while the only serviceable robot remains upstairs. The operational question is how materials and maintenance personnel regain controlled access without inventing a new handling method during the incident.

Request supplier-approved recovery arrangements for the complete loaded vehicle. Identify any required towing equipment, lifting provisions, service clearances and trained personnel. If a recovery device will enter the mezzanine, its loading and access requirements also need approval. The rescue plan cannot depend on bringing an unspecified heavy truck onto a platform reviewed only for normal robot traffic.

Separate equipment recovery from the continuity plan for production. A spare robot upstairs may restore local movement but cannot repair a failed lift. Additional buffer stock may sustain a workstation temporarily but occupies capacity and space. A manual fallback may be viable only during staffed hours.

Procurement should ask the operations manager to choose the required continuity outcome, then have bidders price a consistent response. This prevents one proposal from assuming immediate service attendance while another includes spare equipment and a maintained fallback route. Those are materially different offers even when their normal operating demonstrations look identical.

Buy a defined operating scope and test the difficult interfaces

A well-scoped request for quotation describes where the robots will operate, what they will move and which surrounding systems are included. It also identifies the information still missing. Requiring bidders to declare exclusions is particularly valuable when structural work, guarding, lift modifications and fleet software come from different suppliers.

For acceptance, use the approved operating range rather than an empty-vehicle exhibition. Include the specified loaded configurations, relevant travel directions, critical transfer approaches and simultaneous traffic cases. The test plan should explain which conditions are verified by calculation, inspection, measurement or controlled functional testing.

Fault-response demonstrations must be designed and supervised by the responsible specialists. They should establish the agreed response without exposing a robot or a person to an uncontrolled open edge. Safety validation is not an invitation to create a hazardous failure on a live platform.

Keep the mezzanine-specific evidence with the broader AMR site acceptance testing record. Capture the approved route revision, load family, lift configuration and recovery arrangements so that later operational changes can be compared with a known baseline.

When the retrofit is ready for a purchase decision

A positive feasibility decision should identify an approved operating scope and the work needed to deliver it. For example, a project might be suitable for a dedicated upper-floor fleet after specified floor repairs and a transfer-station redesign. Another site might support loaded robots but lack a practical vertical connection. Those outcomes call for different investments.

Keep unresolved items visible in the commercial comparison. Missing wheel-load information, undocumented structural alterations or uncertain lift-modification permissions are open project dependencies. Do not hide them inside a generic contingency and assume that commissioning will settle them.

The purchase decision becomes more reliable when the buyer can follow one load from origin to destination, explain every change of equipment responsibility, and describe how service resumes after an interruption. That is the level at which a mezzanine automation proposal can be compared with a ground-floor alternative or a different material-handling architecture.

Sources and evidence boundaries

This article combines primary-source industry examples with original procurement analysis. Supplier case studies establish what those suppliers describe, not independent performance guarantees. The numerical planning example is hypothetical. Source pages were reviewed on October 8, 2026.

1. Autoquip: Integrating VRCs and AMRs for Multi-Level Automated Storage & Retrieval, January 26, 2026. Application example involving loaded robots and vertical transport.
2. AMRA: TARS/AMRA-300:2026, February 6, 2026. Public introduction and scope of the robot-friendly-environment guidelines; no unpublished clause requirements are inferred.
3. Mezzanine Safeti-Gates: RobotGate announcement, February 11, 2025. Supplier description of robotic shaft-opening protection.
4. Cogan: Mezzanine Design Loads & Capacity Planning. Load categories and the supporting structural load path.
5. Hi-Level: Understanding Mezzanine Floor Load Ratings, March 18, 2026. Structural considerations including concentrated and dynamic loading.
6. ResinDek: MD Flooring Panels. Product-specific assembly and rolling-load information; not a universal flooring specification.
7. OMRON: LD-Series Integration Guide, I680-E-02. Product-family guidance on localization, physical barriers and integration responsibilities.
8. IDEC Factory Solutions: Automating Inter-Floor Transport Through AMR and Elevator Collaboration. Integration examples and supplier coordination.