By Dr Nicola Tomatis, CEO, BlueBotics
By the time a mobile robot project reaches commissioning, many of the decisions that will shape its performance have already been made. Fleet size, routes, charging points, and pick/drop locations have all been agreed.
But if those assumptions prove wrong, correcting them on-site can mean extra engineering time, additional vehicles, or changes to an already-built layout.
Simulation lets AGV and AMR integrators test these decisions earlier. Starting with the site layout and expected mission flows, a team can assess whether the proposed fleet is large enough to complete the required work, where traffic is likely to build up, how charging affects vehicle availability, and whether routes or workstations need reconsideration.
Much of the planning can therefore take place before vehicles reach the site. Fleet size, routes, charging and mission flows can be refined while changes are still relatively easy to make, leaving on-site commissioning to focus on physical integration and final adjustments.
What a spreadsheet can miss
For system sizing, a conventional calculation may be enough for a very small project. With fleets of two or three vehicles following simple routes and predictable missions, there is usually more room to absorb an estimate error.
As fleets widen and interactions increase, however, calculations based on mission volumes, distance, travel speed and charging assumptions become much less reliable.
A spreadsheet can calculate distance and apply an assumed travel speed. It is much harder to reproduce vehicles slowing through curves, waiting in shared areas, interacting at intersections, or leaving service to charge. These factors affect both throughput and the number of vehicles required.
Take two vehicles with similar technical specifications. One may take five seconds to collect or deposit a pallet, and another may take 10. A third might complete the same operation quickly but be unable to raise or lower its forks while moving.
Charging times introduce another variable that is not obvious from maximum speed, payload, or positioning accuracy alone.
Across hundreds or thousands of missions, these small differences add up. In simulation tests, our team discovered that otherwise similar vehicles produced markedly different outcomes once pick-up and drop-off behaviour was modelled accurately.
In one comparison, one vehicle configuration would have required around 30 percent more vehicles to complete the same job. In a separate test, adding only a one-second delay at each crossing was enough to create a substantial difference across the system performance in the simulated scenarios.
An undersized fleet may fail to complete the required work. An oversized fleet increases capital cost and requires additional charging, parking, and operating space.
A 10 percent error in a three-vehicle project may be manageable. In a proposed fleet of 100 vehicles, it can mean finding the budget and floor space for another ten.
Fleet size is not the only factor. A small fleet in a tightly constrained layout can still justify simulation because traffic and movement may be difficult even with only a few vehicles. In a more open environment, that tipping point may come slightly later.
Once a project reaches roughly five to ten vehicles, however, the value of simulation increases rapidly; with fleets of tens or hundreds of vehicles, the number of routes, pick/drop points and vehicle interactions makes relying on simple calculations alone increasingly risky.
Layout decisions shape performance
Fleet sizing is only part of the job. Simulation can also show whether the proposed material flows themselves need to change.
A project may begin with a CAD drawing showing where vehicles should travel, where they should collect and deliver loads, and whether routes are intended to be single-direction or bidirectional. The arrangement may look perfectly reasonable. Running the expected missions can reveal something different.
A bidirectional route may create unnecessary vehicle interactions as traffic meets in the same aisle. Making it single-direction may improve flow but increase empty travel.
A charging point might be too far from where vehicles normally finish their work. Several busy pickup positions may place too much traffic in the same part of the building.
Consider, for example, an operation with hundreds of missions per hour concentrated through only two pickup points.
The vehicles may have enough capacity, but concentrating so much activity in two places can create a bottleneck. Spreading the work across three or four positions may significantly improve the flow.
That is a relatively simple change while the facility is still being planned. It becomes much harder once machinery, wiring, and other services are already in place. If a traffic or layout problem is only discovered after go-live, correcting it may mean disrupting or stopping part of a live production operation.
Simulation allows routes, traffic rules, charging positions, and workstation locations to be tested before those decisions become physical constraints and before changes carry the risk of costly downtime.
Accuracy matters more than appearance
The reliability of a simulation depends on how closely it represents the vehicles, fleet-management logic and operating conditions being planned.
A vehicle represented simply as a point travelling at a constant speed may be enough to illustrate a basic material flow. It does not account for the vehicle’s physical dimensions, behaviour through curves, acceleration and braking, pick-up and drop-off times, or charging characteristics.
Nor does it reproduce the way the fleet-management system assigns missions, routes vehicles and manages traffic interactions.
In one customer simulation reviewed by our team, involving around 120 vehicles, each vehicle had been represented as a point travelling at a constant speed regardless of curves or traffic conditions. With a fleet of that size, even simple modelling assumptions can significantly distort the result.
A useful simulation therefore needs to reproduce the characteristics that influence real operation. At BlueBotics, our ANT vehicle simulator tool simulates the vehicles and their behaviour exactly, with our fleet management software operating just as it would in the live installation.
The fleet manager responds to the simulated vehicles the same way as it would physical ones, giving the project team a much closer representation of how the complete system will behave, typically with more than 95 percent accuracy against the real installation.
No simulation can reproduce every event in a working factory. Unexpected obstacles, emergency stops, equipment failures, and human activity will always introduce some variation. An appropriate margin still needs to be allowed for these real-world conditions.
Preparing more before going on site
Some work will always remain for commissioning. Pickup and drop positions need adjusting, alignment with conveyors and racks must be checked, and payload transfers must be tested with real loads. The project map needs to be aligned with the environment recorded by the vehicles on site.
Ideally, engineers should not arrive at this stage still deciding whether the fleet is large enough, how the traffic should work, or where charging should take place. Elements like routes, mission logic, traffic rules and charging strategies can all be tested beforehand.
In one project for a packaging producer in Japan, simulation was used to validate traffic management across a seven-vehicle, multi-brand AGV fleet comprising two pallet trucks and five unit load roll transporters.
The modelling helped identify potential deadlocks and allowed the proposed layout to be adjusted before commissioning. On-site commissioning subsequently took around two weeks and was largely limited to mapping the site, fine-tuning node positions and resolving one remaining deadlock.
The project team estimated that commissioning could otherwise have taken twice as long or more, as traffic issues would have needed to be identified, redesigned and retested on site. Eighteen months later, the system remained in operation with vehicle uptime above 99 percent.
In our case, we can transfer the project developed in simulation to the production environment in one click, rather than recreate it from scratch in different software. Once the on-site map and physical positions are adjusted, commissioning can concentrate on vehicle set-up and fine-tuning.
The same approach remains useful after the system is running. A new route, higher mission volume, or fleet change can be tested virtually before altering an installation that is already in production.
That reduces the risk of introducing an unforeseen traffic or performance problem into a live system, where correcting it may require disrupting production and incur costly downtime.
The aim is to arrive on site with fewer unanswered questions. Instead of discovering whether the basic concept works once the vehicles are already on the floor, the team can spend its time fine-tuning a system that has already been tested and avoid solving preventable design problems in a live production environment.
About the author: Dr Nicola Tomatis is CEO of BlueBotics, a developer of autonomous navigation and fleet management technology for AGVs and AMRs. He has led the company since 2003 and joined its board of directors in 2015. Tomatis holds an MSc in Computer Science from ETH Zurich and a PhD in Robotics from EPFL in Lausanne. He received the IEEE Early Career Award in Robotics and Automation in 2008.
