Two AMRs, One Heavy Load: Why Tandem Transport Changes the Engineering

One six-metre assembly changes the purchasing conversation

A production team needs to move a six-metre fabricated assembly from an inspection station to the next manufacturing cell. A single large transporter is one option. Two smaller mobile platforms, positioned under an engineered carrier, appear more flexible: they could share this movement and potentially return to separate duties afterward. The initial quotation looks attractive until someone asks what happens when the two supports carry different loads or one vehicle stops before the other.

This is the central issue in tandem AMR transport. When two robots support one workpiece, the payload becomes part of their mechanical relationship. The project is purchasing the behavior of a coupled transport system, not simply two independent journeys with matching destination coordinates.

The idea has concrete commercial examples. KUKA's KRONE case study describes KMP 3000P platforms being tested in Ibbenbüren and states a three-tonne individual capacity, with twice that capacity in the manufacturer's tandem arrangement. The account describes testing, rather than proving that any pair of three-tonne robots can carry any six-tonne object. [1]

A different implementation appeared in Filics' March 19, 2026 announcement: two physically separate, synchronized robots designed for ground-level Euro-pallet transport. [2] These examples show that shared-load transport spans different product architectures. They do not establish interchangeable controls, support geometries, payload limits or recovery procedures.

To examine the buying decision, this article follows an illustrative six-metre load through pickup, turning, stopping and set-down. The dimensions, calculations and review questions are original analytical examples. They are not measurements from the named manufacturers, recommended operating settings or an application approval.

Decide what connects the vehicles before deciding how they communicate

A tandem AGV system can mean different things in supplier conversations. Two vehicles following each other with separate loads are a convoy. Two vehicles carrying opposite parts of one common carrier are mechanically coupled through that carrier and its interfaces. Two coordinated skids under one pallet constitute another arrangement, with their own geometry and product limits. Procurement should require a drawing that makes the intended meaning unambiguous.

For our six-metre assembly, begin with the support arrangement. Identify where the carrier rests, how horizontal forces enter each vehicle, which movements the interfaces permit and which they constrain. A support may be intended to rotate, slide, articulate or remain rigid in particular directions. Those decisions affect the forces generated when the vehicles do not follow exactly the same geometric relationship.

If both robots attempt to impose incompatible positions on a rigid carrier, the resulting discrepancy may appear as interface force, structural deformation, wheel slip or loss of support. Increasing position-control stiffness is therefore not automatically an improvement. The mechanical freedoms and control objectives must be designed together.

Separate the load-bearing interface from the retention function

A surface that carries vertical weight does not necessarily provide the required restraint during acceleration or turning. Conversely, a locating pin should not be assumed to tolerate every force that two platforms can generate. The carrier drawing should identify intended load paths, permitted forces and moments, engagement confirmation, and the behavior expected during a relevant fault.

The site's guide to heavy-payload AMR load-path engineering provides the underlying structural perspective. The additional tandem question is how the two support interfaces interact when their positions, heights or force contributions differ. A sound design must also preserve the workpiece's dimensional quality, not merely prevent it from falling.

Research offers a useful illustration of this mechanical-control connection. The May 2026 revision of Krawciw and colleagues' multi-rover cargo-transport preprint describes a custom cargo coupling that supports the shared load while providing kinematic freedom between vehicles. It concerns rover research, not certified factory transport, but reinforces why the coupling itself is a design variable. [5]

Modularity has a defined product boundary

Filics' current FAQ gives a concrete example of that boundary: its two-skid Streamliner is specified for a total payload of up to 800 kilograms, and scaling the arrangement to four or six robot skids is not currently supported. The same FAQ excludes overhanging loads. [3] Those limitations belong to that product; they should neither be transferred to heavy-load platforms nor ignored because additional robots appear physically available.

For any proposal involving cooperative mobile robots, ask which combinations the supplier actually supports. Confirm the approved vehicle models, carrier, software versions, operating modes and load families. A fleet manager accepting both vehicles is not evidence that they can carry a common workpiece together.

At pickup, 3,600 kilograms does not divide itself equally

Assume the assembly and the transport fixture supported by the robots have a combined mass of 3,600 kilograms. Their combined centre of gravity lies between two idealized support lines, A and B, separated by four metres. Each proposed vehicle has a nominal payload rating of 2,000 kilograms. At first glance, the combined 4,000-kilogram rating appears sufficient.

That comparison misses the distribution. In a simplified static model with two vertical support reactions, a level arrangement and no applied or support-transmitted couples, equilibrium gives the following expressions:

R_A = m × g × (L − x) / L
R_B = m × g × x / L

Here, m is the combined supported mass, g is gravitational acceleration, L is the distance between the support lines, and x is the combined centre-of-gravity distance from A. The reactions are forces. Dividing each reaction by g gives a mass-equivalent support load, which is convenient for explaining the comparison with the hypothetical vehicle ratings.

This is why AMR load sharing starts with mass properties and support geometry. In the third row, vehicle B receives 65 percent of the supported weight. The unused nominal capacity at A cannot simply be credited to B. A controller cannot eliminate the equilibrium requirement without changing the support arrangement or introducing other forces that the complete design must address.

The centred case is not an approval either. Nominal vehicle capacity may depend on loading geometry and configuration. The model does not evaluate lateral centre-of-gravity offset, support moments, local wheel loading, carrier flexibility, acceleration, braking, floor inclination or transient support changes. Vehicle mass and any additional onboard adapters remain relevant to wheel loading, stability and the platform's available payload allowance. All remain relevant to the actual operating envelope.

Make product variants part of the load specification

The fabricated assembly may receive a motor, tooling package or process fixture before transport. Its total mass could remain within an agreed family while its centre of gravity moves substantially. Define both the mass range and the centre-of-gravity envelope, including the carrier and any tooling borne by the vehicles. State how production identifies the applicable configuration.

If weighing or force sensing contributes to release, specify what it measures and which abnormal conditions it can distinguish. A plausible total weight does not prove a correct reaction at each support. A pressure or motor-current proxy requires its own application evidence before being treated as a dependable estimate of supported load.

Pickup itself changes the load path. During transfer from fixed stands, some weight may remain on the station while the robots take the rest. The final two-support calculation does not describe every intermediate condition. The approved lift and engagement sequence must address support transitions and confirm the state required before travel.

The first turn belongs to the entire six-metre load

The proposed route leaves the inspection bay through an open aisle, then turns toward the next cell. Looking only at the robot footprints can make this route appear straightforward. The carrier and its overhangs occupy space between and beyond the vehicles, and that space changes throughout the manoeuvre.

For a simple plan-view illustration, let the combined load-and-fixture envelope be a rectangle six metres long and 1.6 metres wide. When its long axis is at angle θ to the aisle's longitudinal direction, its projection across the aisle is:

Projected width = 6 × |sin θ| + 1.6 × |cos θ|

At zero degrees, the projected width is 1.6 metres. At 30 degrees, it is approximately 4.39 metres. At 90 degrees, it is six metres. This calculation shows how orientation changes the transverse envelope; it is not a minimum aisle-width calculation. A real swept-path study must also include translation, vehicle protrusions, actual geometry, tracking uncertainty and the necessary clearances and protective measures.

Synchronization does not mean identical wheel speeds

During a turn, different support points follow different trajectories. Depending on the chosen instantaneous centre of rotation, they may require different velocity directions and magnitudes. Sending identical commands to both robots does not, by itself, preserve the desired load motion. Steering capability, drive constraints and the freedoms at the support interfaces must remain compatible.

Evaluate synchronized AGV transport using the motion of the common load and the relative relationship between its supports. Two individually accurate vehicle trajectories can still impose an unacceptable relationship on a rigid carrier if they use inconsistent reference frames, timing or geometry. State the load reference point and coordinate conventions explicitly in drawings, controls and test records.

The occupied route should include the connected assembly throughout the movement. A space between the two chassis is not automatically free space for another vehicle or a pedestrian. Likewise, an aisle reservation should not be released merely because the leading platform has left it. The article on shared-route reservations and deadlock prevention provides the general traffic-management context.

The floor can change the relationship between the supports

With support lines four metres apart, one vehicle can reach a joint, slope transition or local height variation while the other remains on a different surface. The resulting relative height or attitude change may matter to the carrier, the support interfaces and the available wheel contact. Whether it produces extra force, permitted articulation or unloading depends on the actual mechanical arrangement.

Survey the route as a coupled movement rather than as two independent tracks. Include the pickup and set-down zones, where small geometric changes can affect how support transfers. Use the existing AMR floor interface requirements as background, then evaluate the differential conditions that the separated support points encounter.

A route can therefore fail the review even when both individual robots can traverse it empty. For long load transport automation, geometric compatibility belongs to the load, carrier, vehicles and floor together. A successful empty-platform demonstration answers only a small part of that question.

A stop at one end is a load event

Now consider an obstruction detected near the front vehicle during loaded travel. The important question is not only whether that vehicle stops within its own validated limits. The assessment must establish what the other vehicle does, how the carrier responds during the transition, and what supports and restraints remain effective afterward.

Differences in detection, control response, drive behavior and available traction can produce relative motion or interface loads. The system need not achieve mathematically identical stopping profiles at every point. It must keep the coupled response within its justified mechanical and protective limits. Those limits depend on the application; a universal millimetre or millisecond allowance would be misleading.

A communication cycle is only one part of the response

R3's Stäubli project account describes paired AGVs moving sensitive workpieces through a virtual drawbar. It reports direct vehicle-to-vehicle control and safety-data exchange using PROFINET and PROFIsafe, with a 32-millisecond PLC cycle. [4] This is a specific architecture example, not a universal response requirement.

An AGV virtual drawbar maintains an intended relationship through controls and communication rather than a conventional physical towing bar. That description alone does not specify its safety functions. The quoted PLC cycle is not the complete time from detecting a relevant event to achieving the required physical state, nor does it establish an allowable mismatch for our six-metre assembly.

Ask the supplier to identify the paths for ordinary motion coordination, safety-related information and fleet-level task messages. Their purposes and validated behavior may differ. If a coordination function is credited with reducing a hazard, its architecture and evidence must support that role. An ordinary wireless connection or synchronized dashboard is not sufficient proof.

Inspect the complete stopping history

The site's analysis of heavy-payload AMR stopping behavior explains why load and operating conditions matter. For tandem transport, add measurements of both vehicle motions, relative support position, relevant interface forces and the motion of critical load edges. A final photograph of two stationary robots cannot reveal what happened between initiation and standstill.

Review relevant cases such as a protective event associated with either vehicle, loss of the coordination channel, an unavailable drive and loss of a required position reference. Include loss of power or a required lift or load-holding function on either vehicle, and establish how the approved support condition is maintained after motion stops. Determine the specified response before testing. Use appropriate simulation, controlled fault injection and qualified physical validation methods; do not improvise a failure beneath an unsecured production load.

The application assessment also needs to address the space beneath the common load, between the vehicles and around its overhangs. The existing AMR application risk assessment guide provides the broader method. Here, the distinctive issue is that the hazardous moving envelope and the support system span more than either robot alone.

The pair must retain a shared identity through the whole mission

For dual robot load handling, readiness is a relationship. Both vehicles may individually report that they are available while the intended combination is unsuitable. The application needs the correct pair, compatible configurations, the intended carrier, verified support engagement and the approved load information for the shared task.

Make role assignment explicit. A lead-and-follow architecture should define what happens if the lead vehicle becomes unavailable. A more distributed architecture should define how inconsistent or stale information is handled. Neither architecture should be accepted simply because its name sounds more resilient; evaluate the implemented response to the relevant conditions.

Energy planning also belongs to the combined mission. Check whether both vehicles can complete the approved movement and reach the required holding or release state. Averaging their battery percentages can hide a limiting vehicle, while identical percentages need not imply identical remaining capability under different loading and battery conditions.

During execution, preserve the relationship between the load, carrier, pair identity and task record. A replacement robot should not inherit permission to engage solely because it has the same nominal capacity. Confirm mechanical and software compatibility, configuration, calibration and the approved pairing procedure before substituting equipment.

Set-down is the end of the load-sharing task

Arrival coordinates do not prove that the station has accepted the load. The receiver must provide the intended support, and the application must establish when the vehicles are free to disengage. Uneven or incomplete set-down can leave one end loaded while the other vehicle begins a movement intended for an empty platform.

The task should remain a shared-load task until the release conditions are satisfied. Only then can the robots return to independent assignments if the supplied system supports that mode. Record the completed material movement once; counting it as two completed deliveries would inflate the business result without moving any additional product.

Plan how to release a stranded load before approving the route

A tandem arrangement should not be sold internally as automatic redundancy. If one platform fails while both support the assembly, the healthy platform may remain occupied and unable to continue independently. Access to the workpiece, the aisle and the next production station may all be affected.

The project therefore needs a recovery concept for the common load. Identify where a controlled hold is possible, how its support state is established, what approved equipment or station can receive it, and who is qualified to authorize the recovery. The required method depends on the carrier, vehicle design, fault and surrounding space; it cannot be reduced to a generic instruction to tow the failed unit away.

Confirm that necessary recovery resources can reach the stopped arrangement. A rescue plan that assumes access from the side is ineffective if the six-metre assembly occupies a narrow lane between fixed machines. Reserve the physical access and explain the production consequences of keeping that space available.

After intervention, re-establish the relevant state of both vehicles and the common load. Restored communication does not prove unchanged geometry, correct support or permission to resume. The site's guide to safe restart and recovery explains the general distinction; tandem operation adds the need to reconcile both ends of the same mechanical system.

Maintenance planning should preserve this paired capability. Identify which vehicles can substitute for one another and which carrier configurations require specific equipment. Include the time needed to establish a qualified pair, not just the time to repair or replace one robot. A fleet with several available units can still lack an available approved combination.

Ask for an acceptance record written around the load

Tandem transport validation should follow the coupled task from initial support transfer to final release. Begin with agreed limits and representative configurations, then select measurements that can show whether those limits remain satisfied. Do not choose a headline success percentage first and work backward toward a convenient demonstration.

Instrument the quantities needed for the claim. Vehicle localization logs may be useful, but they are not necessarily independent evidence of actual relative motion or carrier deformation. Select reference measurements with suitable accuracy and time alignment. Document their uncertainty so that a claimed margin is not smaller than the uncertainty in the measurement used to establish it.

Test the relevant extremes of the approved load family, including centre-of-gravity variation and difficult support transitions. Include the route conditions that challenge the arrangement and the fault scenarios identified by the assessment. Repeating the easiest centred load on a clear straight aisle does not represent the full operating envelope.

For multi-robot transport safety, define responsibility for the complete application. The vehicle supplier, carrier designer, controls integrator and site team each contribute evidence, but someone must reconcile their assumptions. A robot specification assuming one carrier geometry and a fixture drawing assuming another can leave a gap that neither individual document reveals.

Keep operational performance separate from protective behavior. Measure completed shared-load movements, occupied time, waiting for a partner, charging constraints, interventions and blocked-route time. A correct protective stop can be essential while frequent avoidable stops still make the proposed throughput unattainable. Both observations belong in the purchasing decision.

Finally, define what changes require review: vehicle substitution, revised carrier geometry, new load variants, altered support spacing, control updates, floor repairs or route changes. Retain the approved combination and its evidence as a configuration, rather than scattering acceptance records across two unrelated vehicle files.

Compare tandem transport with the strongest single-vehicle alternative

The commercial case should compare complete installed systems performing the same workload. A tandem proposal may offer useful flexibility when supported vehicles can serve separate tasks, when products vary in length or when existing platform modules simplify maintenance. Those advantages must be available in the delivered operating modes, not merely in the buyer's imagined future layout.

A single larger transporter may simplify the support arrangement, task scheduling and some recovery situations. It may also demand a larger footprint or more specialized equipment. Neither choice is universally superior. Compare the actual carrier, route, protective measures, stations and service arrangements required by each option.

Include engineering and operational costs beyond the two vehicle prices: the common carrier, interfaces, coordination equipment, validation, recovery resources, spares and occupied production space. Count the consequences of waiting for two compatible vehicles to become available together. Check whether pairing and separation consume enough time to erode the expected benefit of reusing the platforms independently.

For the six-metre example, the decision should follow the evidence. The supported mass and centre-of-gravity range must fit the actual interfaces, the complete load must negotiate the route, and stopping and recovery must preserve the required conditions. If those questions cannot be closed, adding a second robot has not solved the transport requirement. If they can, tandem operation becomes a defined capability that procurement can compare and production can maintain.

Sources and evidence notes

Sources checked on September 29, 2026. Manufacturer accounts establish stated product scope or project architecture, not independent certification of the illustrative application. The six-metre transport scenario, calculations, comparisons and acceptance framework are original analysis.

1. KUKA: Mobile robotics for optimized material flow at KRONE. Case-study URL dated August 2025; no exact publication day shown. Describes KMP 3000P testing and the manufacturer's tandem-capacity statement.
2. Filics: New generation of the Filics Unit. March 19, 2026. Describes two physically separate synchronized robots for ground-level Euro-pallet transport.
3. Filics: Current Streamliner FAQ. Undated product information consulted on the review date. Supports the stated total payload, overhang restriction and present limit on combining additional units.
4. R3 Solutions: Collaborative AGVs in the Cleanroom. Undated supplier project account involving Stäubli. The cited 32 milliseconds is its reported PLC cycle, not a complete stopping-time specification.
5. Krawciw et al.: Sharing the Load: Autonomous Multi-Rover Cargo Transport. Preprint revision dated May 14, 2026. Supports the research example of coordinated transport with a purpose-designed cargo coupling; not industrial application certification.