An AGV (Automated Guided Vehicle) is a driverless transport platform that follows one of three physical or optical guidance paths and is dispatched by a central scheduler, with payload classes typically spanning 50–5,000 kg and travel speeds in the 0.5–1.5 m/s band for indoor service [S3].
The three guidance families in commercial use—magnetic-tape, laser-scanner with reflectors, and UHF RFID—are the single biggest driver of total installed cost, accuracy and layout lock-in, and the spec decision should start there rather than at the vehicle model number [S3].
Three Guidance Modes: Cost vs Accuracy vs Layout Flexibility
Magnetic-tape guidance is the lowest-cost option and the most common in Chinese factory deployments, but the embedded tape fixes the path permanently, limits station count, and visibly clashes with finished-floor aesthetics in clean rooms or showroom areas [S3]. Laser guidance with wall- or pole-mounted reflectors commands the highest capital cost and demands the cleanest floor profile, which is why most integrators treat it as a last resort for plants where the floor cannot host magnetic media [S3]. UHF RFID guidance sits in the middle on cost and gives the highest route-positioning accuracy, supports the densest station layouts, leaves the building interior visually untouched, and adds a safety/stability margin the other two cannot match [S3].
Where AGVs Win: Labour, Energy and Throughput
AGV dispatch is fully computer-controlled—magnetic sensors, laser scanners and an on-board PLC together trigger pick-up, drop-off and charge cycles without human intervention once a terminal command is queued, which is the headline advantage in any single-shift production cell [S3]. Auto-charge is built into the fleet: when battery state-of-charge crosses a configurable threshold the AGV posts a request, leaves its queue slot, and rolls onto a contact or opportunity charger, so a multi-shift plant can run lights-out with no battery-swap crew [S3]. Compared with stacker cranes in fixed-aisle warehouses, AGVs add horizontal path freedom and decouple the storage topology from the transport topology, which is why 3PL and e-commerce sites layer AGVs over racking rather than replacing crane aisles outright.
Indirect gains that show up on a P&L: a single AGV typically replaces 1.5–2.0 FTE on a two-shift schedule, accident rates on the routes drop because traffic is centrally scheduled with conflict resolution, and ERP/WMS integration removes the manual scan-and-confirm step at every hand-off [S3].
Where AGVs Lose: Who Should NOT Spec One
AGVs are the wrong tool when the site is unfinished, frequently re-zoned, or has variable obstacles that the guidance system cannot model—construction sites, outdoor yards longer than ~200 m with weather exposure, and small-batch job shops with a different routing every shift. Magnetic-tape AGVs in particular cannot be redeployed in a weekend; pulling up the tape, re-laying it, re-mapping the table and re-commissioning traffic rules is a 3–7 day exercise even on a 5,000 m² site. The fleet scheduler is also a single point of failure: if the WMS/MES link goes down, vehicles will fault-stop unless a local fallback rule is written, and most entry-level controllers ship without one. [S1]
For a moving-floor conveyor duty such as vibrating conveyors on a foundry shakeout, an AGV is the wrong category entirely—conveyors handle continuous bulk at fixed elevation, AGVs handle discrete unit loads on a routed plane.
Selection Criteria Beyond Guidance
Payload and drive-unit topology matter as much as navigation. A differential-drive unit-load AGV with a 500 kg payload is the volume model; for 2,000 kg+ pallet moves a steering-arm or omni-drive chassis is mandatory, and the specifier should confirm wheel material (polyurethane vs rubber vs steel-faced) against floor finish. Battery chemistry is a 2026 hot-button: lithium-iron-phosphate (LFP) packs now dominate new bids over lead-acid because they survive 3,000+ cycles to 80 % DoD and accept opportunity charging without venting, but the swap costs roughly 2.0–2.5× the pack price of an equivalent lead-acid unit. Safety stack should be quoted to ISO 3691-4 (driverless industrial trucks) with two independent LiDAR scanners plus a bump skirt, and the safety logic must be PL d / Cat 3 minimum—any vendor that cannot hand over a TÜV or UL certificate for that stack should be deselected. [S2]
For sites that already run collaborative robots on the line, the AGV fleet should share a common fleet manager (ROS 2 or VDA 5050) so handoffs between mobile base and manipulator arm are deterministic rather than hand-shaken.
Failure Modes and Maintenance Gates
Three failure modes drive the bulk of unscheduled downtime: (1) magnetic-tape contamination or peel at high-traffic corners, which the maintenance plan must treat as a 90-day inspection item; (2) wheel wear on polyurethane tyres after roughly 2,000 km of service, depending on floor abrasiveness; (3) LiDAR window contamination in oily shop floors, which forces a daily wipe-down and a redundant scanner so a single occluded lens cannot blind the safety stop. Battery state-of-health (SoH) should be trended monthly and a pack retired at 80 % of nameplate capacity, which for LFP usually lands at year 6–8 in two-shift service. [S3]
Standards, Sourcing and 2026 Market Signal
Compliance baseline for new European installs is ISO 3691-4 (safety of driverless industrial trucks), EN ISO 13849-1 (safety-related control systems, PL d) and, for sites inside Ex-zones, IEC 60079-series and ATEX 2014/34/EU on the AGV chassis and on-board electricals. North American sites pair those with ANSI/ITSDF B56.5 (formerly ASME B56.5) and UL 3100 for the energy storage system. Fleet-manager interoperability is increasingly specified against VDA 5050 (the German automotive OEMs' joint interface), which is now the de-facto RFP requirement for Tier-1 supplier logistics in 2026 [S3].
Sourcing signal to watch: Chinese integrators continue to lead on magnetic-tape and RFID AGV unit cost, while European OEMs (Swisslog, KION/Dematic) and Japanese majors (Daifuku, Murata) hold the high end on laser-guided and AGC-derived platforms; for an articulated robot cell that needs an AMR feeder, bundling procurement with one vendor typically cuts integration hours by 20–30 %.
Next spec gates to track: (1) whether the project site can host a 24 V/60 A opportunity-charge contact strip at every dwell station, which is the single biggest unlock for 24/7 lights-out AGV service; (2) the WMS/MES API maturity of the fleet manager, since a 2026-vintage VDA 5050 interface is what separates a spec-compliant bid from a legacy lock-in.