Don't Buy the Vehicles First: A Six-Step Path from Survey to Handover for AGV Projects
Release time:
2026-11-18
In plenty of factories, the AGV project does not stall because the vehicles are bad. It stalls at approval, when the whole thing is treated as an ordinary purchase: buy a few vehicles, draw a route, assume it will run. Reality arrives with the equipment — aisles get congested, systems refuse to talk to each other, nobody knows how to clear an alarm, and the line stops on day one.

An AGV is not a vehicle that drives itself; it is a moving logistics line (illustrative image, sourced online)
An AGV project is closer to an engineering retrofit of your material flow than a purchase. Below, the road is broken into six legs, from solution survey to go-live handover: what each leg involves, and how far you have to go before it actually counts as done. At Yikong we tend to keep this same spine in front of us — don't rush to compare vehicles; get the process right first.
01Survey: Before You Ask How Many Vehicles, Understand the Move
The first step is the easiest to skip, and the one you can least afford to skip. A survey is not a question of "how many vehicles do you need". It is about describing the move itself: what is being moved — pallets or totes, what weight and dimensions; how it moves — where the start and end points are, how often, how fast the production takt runs, whether empty pallets have to come back; and the environment it moves in — how wide the aisles are, whether the floor is level, whether forklifts and people share the route, whether it has to cross floors.

Aisle width, floor condition, mixed traffic — all of it is written on site, not in the meeting room (illustrative image, sourced online)
Once that is clear, the survey has to produce a takt calculation: how much has to move per unit of time, whether the peak holds, and how many vehicles that requires. For early surveys, Yikong usually recommends walking the real flow on site rather than relying on descriptions from a meeting room. A vague requirement turns into rework in vehicle selection, layout and traffic control. One extra day here can save a week of rework later.
02Solution Design: Three Decisions You Cannot Avoid
Design starts once the requirement is clear. Vehicle type follows the load: underride (latent) vehicles for pallet transport, towing units for line-side hauling, forklift AGVs for pallet storage and retrieval, heavy-duty platforms for oversized transfers. Don't just copy what the plant next door uses.
![]() Underride: slides under the load and lifts — the workhorse for pallet moves | ![]() Forklift AGV: takes and places pallets, made for high racking |
Navigation is the same story — the most advanced option is not automatically the right one. Magnetic tape suits fixed routes and cost-sensitive sites; QR codes give high accuracy and suit dense storage locations; laser SLAM offers flexibility where the layout keeps changing. The point is fit, not fashion.
![]() QR codes: floor markers for positioning — high accuracy, needs upkeep | ![]() Laser SLAM: maps the environment — more flexible when the layout changes |
Traffic control decides how vehicles avoid each other, how tasks are dispatched, and how the fleet connects to WMS/MES. Its capability determines not only today's throughput but whether you can add vehicles and stations later. At the design stage Yikong puts the integration question on the table early: interface list, data fields, exception feedback, all checked before anything else — isolated systems are how a lot of these projects end.
03Simulation: Make Your Mistakes in Software, Before Go-Live
Where possible, simulate first or run a small pilot. Run real takt data through the model and look at three things in advance: whether the aisles will jam, whether you have enough vehicles, and whether task allocation is sensible. Simulation is not animation — it only means something when it runs on real orders and real takt.

Find the bottlenecks in software first; it costs far less than trial-and-error with real vehicles (illustrative image, sourced online)
This step costs software time and saves the cost of stopping the line to fix things after go-live. Plenty of projects skip it to protect the schedule, then burn far more time and money trial-and-erroring on site with real vehicles.
04Site Works and Integration: One Vehicle, Then the Fleet, Then the System
Once the equipment is on site, don't roll everything out at once. Finish navigation commissioning on a single vehicle, then bring multiple vehicles in to test avoidance and dispatching, and only then connect to WMS/MES and get the data flowing. At this stage the weak point is usually not the vehicle — it is the interface, system to system.

During single-vehicle commissioning, the engineer checks parameters at the control cabinet (illustrative image, sourced online)
We recommend three levels — single vehicle, fleet, system — commissioned and accepted level by level. Keep records at each level and close the issues before moving up. The later a problem surfaces, the harder it is to isolate, and the more chaotic the site becomes. In integration, Yikong cares more about whether each level can reproduce a fault on its own than about getting lucky on the last day.

Data flowing between systems eventually has to be visible on one screen (illustrative image, sourced online)
05Trial Run and Training: Someone Has to Be Able to Catch It
A system that runs is not a successful project. During the trial run, watch stability and failure rates continuously, and settle the exception process: what the common alarms are, how to reset them, when to switch to manual, what to do in an emergency. Training should not stop at the shift leader — cover operators by shift: how to load, how to clear a fault, how to switch between automatic and manual. Put the operation cards where people actually work, so a problem doesn't stop production.

Training has to reach concrete actions: loading, resetting, switching to manual (illustrative image, sourced online)
Many projects fail not on technology but on "nobody knows how to use it, so the first fault stops the line". However good the system is, someone has to be able to catch it.
06Acceptance and Handover: Put the Numbers in the Contract
Acceptance needs explicit criteria: whether the transport takt is met, the failure rate, positioning accuracy, system availability. These are best agreed at contract stage and checked item by item against data at acceptance. Acceptance is not a signing dinner — it is the numbers talking.

Acceptance is verified item by item against data, not judged by how the site looks (illustrative image, sourced online)
Handover is more than equipment: layout drawings, operating manuals, spare parts list, warranty period and after-sales response mechanism. Keep the service channel open and the system will keep running. Yikong prefers to see handover as the real start of the project: equipment arriving is only the first half of the road; whether it keeps running reliably afterwards, and how fast someone responds when something goes wrong, is what the customer faces every day.
In the end, the worst pattern in an AGV project is to buy first and think later.
Understand the site, think the interfaces through, train the people, fix the standards — everything downstream gets easier. Survey first; design around vehicle, navigation and traffic control; simulate before go-live; commission level by level; train people during the trial run; write acceptance criteria into the contract. Do all six properly and go-live is only a matter of timing. Skip any one of them and rework or downtime will find you somewhere along the way.
Mob: +86-513-82104870
E-mail: jimmy.tang@shtongpu.com
E-mail: info@bicontrols.com
Address: Building 63, No.2699, Binjiang Street, Binjiang Avenue, Haimen District, Nantong City, Jiangsu Province
Sweep and pay attention to Baoyan Technology





