AGV Positioning with RFID: Landmark Tags & Accuracy Guide

Ask any AGV integrator what actually goes wrong on site, and the answer is rarely "the motors failed." It is drift. A laser SLAM vehicle that knows where it is to ±10 mm after a fresh mapping session can wander by several centimeters after a few hundred meters of travel, after a floor repair, after a season of changing light and moving pallets. Inertial and odometry estimates accumulate error by definition. RFID positioning solves this in the simplest possible way: it gives the robot an absolute, physical truth to touch every few meters.

This is not the same proposition as RFID identification on a production carrier. Here the tag is not the object being tracked — the tag is the map.

The core idea: tags as absolute reference points

In an AGV positioning layout, passive RFID tags are installed in or on the floor at defined coordinates: intersections, Library-borrowing-machine-touch-query-intelligent-terminal-all-in-one-machine.html target='_blank'>workstation approaches, charging stations, elevator landings, buffer lanes. Each tag carries a unique ID that the fleet management system has already bound to a real coordinate in the site map.

A reader mounted on the underside of the AGV passes over the tag and gets a read. The controller looks up the ID, learns "I am at node X," and — critically — resets the accumulated drift of its odometry, IMU or SLAM estimate back to zero. Between two tags the vehicle navigates by its own means; at each tag it re-anchors. The longer the run, the more valuable that re-anchoring becomes.

The second use is station-level decision making. A tag is not only a coordinate, it is an instruction: node A means turn left, node B means decelerate, node C means stop, dock and hand off to the robot cell. Because the tag is writable, the instruction set can be revised without touching the floor.

Why RFID survives where vision struggles

Compared with optical codes and magnetic tape, RFID's advantage is that it does not need line of sight and has no moving or optical components. Tags are typically embedded in the floor, covered by a wear layer, or recessed into a protective housing. That means:

• Dust, oil mist, metal swarf and washdown do not block a radio read the way they defeat a camera looking at a QR code.

• No consumable surface to maintain. QR codes get scuffed by forklift wheels and need cleaning and replacement; magnetic tape lifts, tears and is destroyed when the route changes.

• Passive and battery-free. A tag has no power source, no maintenance interval, and a service life measured in years of continuous traffic.

• Speed tolerance. AGV-grade readers are specified to read reliably at vehicle speeds of 2–3 m/s, which matters on long haul routes where stopping to scan is unacceptable.

Choosing the frequency: LF, HF or UHF

This is the decision that most determines system behaviour, and the three bands behave very differently.

The practical rule: use HF (or LF) where you need a crisp, repeatable position event — docking, turning, stopping — and use UHF where you need early warning or Asset reading — detecting that a vehicle is entering a zone, or reading pallet tags while the AGV drives past. Many real installations run both: HF underneath for positioning, UHF on the mast or load deck for material identification.

The reason is field geometry. A tightly controlled read window is a feature, not a limitation: it is what makes "the tag fired, therefore I am within ±20 mm of the node" a valid statement. A UHF reader with a multi-metre field cannot make that claim from a single read, which is why precision AGV work has historically gravitated to the lower bands.

Getting more than a node ID: RSSI, phase and tag arrays

A single tag gives a discrete correction — excellent, but only at tag spacing. Research and advanced deployments push further:

• RSSI-based ranging. Signal strength gives a rough distance estimate. Used alone, it is noisy in industrial environments — metal racking, moving vehicles and multipath all distort it — but filtered through a Kalman or weighted-centroid model it becomes usable for metre-level to sub-metre estimation.

• Phase-based methods. The phase of the backscattered signal carries far finer distance information than RSSI. Multi-antenna robot platforms using synthetic-aperture or phase-difference techniques have demonstrated centimetre-class tag localisation in laboratory and warehouse conditions.

• Dense tag arrays. Instead of sparse landmarks, lay tags on a regular grid (triangular or cellular spacing) so the vehicle is almost always over or near a tag. Published implementations of RFID-array floors report average positioning accuracy in the sub-5 cm range, with dedicated "stop zone" tag clusters giving tighter docking repeatability.

• RF-SLAM. The frontier: using a moving RFID Reader to build a map of unknown tag positions and localise the robot simultaneously. Published results put robot localisation around 5 cm and tag position estimation around 10 cm, without any prior tag survey — attractive because it removes the most tedious part of a deployment, the manual coordinate survey.

These methods are real but they carry caveats: they assume reasonably open environments, consistent tag orientation, and calibration effort. For most production AGV fleets, the honest Engineering answer remains sparse HF/LF landmarks for absolute correction, plus the vehicle's own SLAM or odometry in between.

How it fits with SLAM, magnetic tape and QR

RFID is rarely a whole navigation system; it is the anchor inside one. The current mainstream architecture is hybrid:

1. Laser or visual SLAM handles free-path routing, obstacle avoidance and route flexibility.

2. RFID landmarks provide drift-free absolute correction at intersections and, most importantly, the final approach to a dock.

3. A mechanical or vision fine-alignment step handles the last few millimetres if the process demands it.

This split matters because SLAM accuracy is environment-dependent and degrades in long corridors, glass-walled areas and highly dynamic spaces. Docking repeatability is where defects are born — a lift that engages a rack 15 mm off-centre, a conveyor handoff that jams. RFID tags at the dock convert a probabilistic estimate into a deterministic one, and vendors commonly quote ±5 mm class docking performance when a landmark tag defines the stop point.

Conversely, RFID alone is poor at continuous path following. It tells you where you are at discrete points, not how to curve smoothly between them. Treating it as a full replacement for SLAM leads to sparse, jerky motion; treating it as a correction source leads to a system that is both flexible and repeatable.

Deployment checklist

• Tag placement. Define nodes at every routing decision and every dock. Avoid placing tags directly against large steel plates or inside reinforced floor joints without testing; use on-metal or spacer-mounted tags where the floor is steel.

• Mounting height and orientation. Reader-to-floor distance is the single most sensitive mechanical parameter. Keep it consistent, and protect the reader from impact and washdown.

• Reader interface. Industrial readers should speak RS-232/485, CAN bus or modbus, and industrial Ethernet (Profinet, EtherNet/IP, Modbus TCP) for direct PLC integration.

• Speed budget. Confirm the reader's specified maximum pass speed exceeds your fastest AGV with margin — a missed read at a fork is a routing failure.

• Map binding. Maintain a single authoritative ID-to-coordinate table in the fleet manager; duplicated or hand-edited copies are a classic source of "the AGV stopped at the wrong station."

• Read signal strength monitoring. Good readers expose RSSI per read. Logging it turns tag-survey mistakes and local interference into diagnosable data rather than mystery misses.

• Redundancy. For safety-critical stops, do not rely on a single tag. Use a tag plus an independent sensor (mechanical limit, proximity, or encoded landmark) so that a damaged tag degrades gracefully instead of causing an overrun.

The bottom line

RFID will not replace SLAM, and SLAM has not made RFID obsolete. They solve different halves of the same problem: SLAM answers "how do I get there flexibly," RFID answers "am I really here." For fleets that must dock precisely, run for years without re-mapping, and work in dirty, metallic, washed-down industrial environments, a floor of passive tags remains one of the cheapest forms of certainty an AGV can buy.