AGV vs AMR: the architectural choice that decides your next 10 years

A mid-sized automotive stamping plant outside Stuttgart spent 18 months piloting autonomous mobile robots on their press-to-assembly lineside transfer route. The AMRs navigated well in testing. In production, two things broke the program: the narrow transfer corridors — 1.4 meters — were just wide enough for one AMR to pass but not for two to choreograph without stopping, and the unpredictable obstacle patterns from forklift cross-traffic meant the AMRs' SLAM re-routing was triggering on average 3.4 times per hour per unit during peak shift. Throughput was 23% below the contractual baseline. The plant pulled the program and re-tendered on laser-guided AGVs running fixed loops.

Nobody made a bad technology choice. They made a bad architecture choice — selecting a flexible, dynamic system for a route that was fixed, narrow, and throughput-critical. That distinction is the most important thing a plant logistics manager can understand before signing an AGV or AMR contract.

The fundamental architecture gap

AGVs and AMRs are not two tiers of the same product. They are built on opposite assumptions about where intelligence should live.

AGVs externalize navigation to the infrastructure. The vehicle is predictable because the route is predetermined. The robot doesn't decide where to go — it follows signals from guide wire, magnetic tape, QR codes, or laser reflectors placed by engineers. The control system schedules flows, arbitrates intersections, and dispatches units. The vehicle is a dumb terminal with a reliable motor.

AMRs internalize navigation to the robot. LiDAR, cameras, and SLAM (Simultaneous Localization and Mapping) let the vehicle compute its own position in real time, reroute around obstacles, and adapt when the layout changes. The vehicle is intelligent; the route is soft.

This architectural split determines cost, risk, flexibility, and the switching cost if you change your mind.

Navigation physics: what each system actually does

AGV guidance methods

Wire guidance was the original and remains the most deterministic. A conductor buried 25–50mm below the floor surface carries an alternating current; the AGV tracks the electromagnetic field. Accuracy is ±5mm. The floor modification is permanent and adds $50–$150 per linear meter in conduit and installation cost. Changing a route means cutting the floor again.

Magnetic tape is cheaper to install ($5–$20 per linear meter) and sits on the floor surface or in a shallow 10mm channel. Accuracy is ±10mm. It degrades under heavy forklift traffic and requires periodic inspection and replacement of worn sections — typically every 12–36 months depending on traffic volume.

Laser guidance (LGVs — Laser Guided Vehicles) uses a rotating scanner that triangulates off reflective targets mounted on walls and racking. No floor modification required. Accuracy is ±5mm. Reflector placement and layout changes are managed in software. This is now the dominant guidance method for new installations in high-value manufacturing environments because it combines deterministic accuracy with reasonable layout flexibility.

Natural feature laser (contour navigation) uses the environment itself rather than reflectors, similar to AMR SLAM but still running fixed routes defined by the WCS (Warehouse Control System). It blurs the line between AGV and AMR — accurate to ±25mm and capable of operating in environments where reflector mounting is impractical.

AMR guidance

AMRs build an environmental map on first deployment (2D or 3D LiDAR + camera fusion), maintain localization via continuous SLAM, and navigate dynamically between assigned waypoints. Route logic lives in software and can change without touching hardware. The system adapts to obstacles — but "adapts" means either waiting, rerouting, or escalating to human review, each of which consumes time.

Accuracy varies by sensor suite: ±25mm on commercial warehouse AMRs, ±10mm on high-precision industrial platforms. Neither matches laser-guided AGV precision at high speed.

The performance envelope

Understanding the performance ceiling of each architecture is essential before any procurement decision.

Cycle time and throughput

AGVs on closed-loop routes achieve predictable, schedulable cycle times. A laser-guided unit on a 120-meter loop in an automotive body shop can be scheduled to ±3 seconds across a full shift. That predictability is what makes AGVs integrable into lean manufacturing pull systems and JIT line delivery — if your production line needs a part trolley every 8 minutes, an AGV can be tuned to deliver one every 8 minutes ± 15 seconds, all day, every day.

AMRs in mixed-traffic environments achieve median cycle times that are often comparable to AGVs, but with a variance that matters. A route that averages 4 minutes can spike to 7 minutes when an AMR encounters a blocked aisle and reroutes. In high-volume pull systems, variance at the carrier level propagates to variance at the line — and line stoppages are where the real cost appears.

The relevant metric is P95 cycle time, not median. In controlled environments with low obstacle interference, AMR P95 can be close to median. In dynamic manufacturing floors with mixed manned-forklift traffic, AMR P95 can be 2–3x median. AGV P95 on a closed loop is typically 1.1–1.3x median.

Speed

AGVs on clear corridor routes typically run at 1.2–1.8 m/s loaded. AMRs in shared-space environments typically run at 0.8–1.2 m/s because safety-rated speed limits apply when humans share the space. If your route is forklift-only or requires formal segregation anyway, the speed penalty doesn't apply — and a laser-guided AGV may run faster simply because the route geometry is optimized at installation.

Payload

Heavy-payload AGVs are designed from the ground up around load handling: fork AGVs, unit load carriers, tugger trains pulling multi-cart loads of 1,000–10,000+ kg. The guidance system is separate from the load-handling mechanism. AMR platforms in the heavy-payload segment exist (several manufacturers offer 1,500 kg AMRs) but are less mature and carry higher per-unit prices for equivalent payload capacity.

Infrastructure commitment: what you're actually buying

The AGV-vs-AMR decision is primarily a question of where you want to carry infrastructure debt.

AGVs front-load infrastructure cost. Wire trenching, reflector installation, traffic control hardware, integration engineering — most of this is paid before the first pallet moves. A typical laser-guided AGV system installation budget for a 10-unit fleet runs $800K–$2M all-in, with $200K–$600K of that in site preparation and integration. The vehicles themselves may represent only 40–60% of the total project cost.

AMRs shift infrastructure cost to software and integration. Map commissioning, WMS/WCS integration, safety zone configuration, and fleet management software licensing are the dominant upfront costs beyond the vehicles. A 10-unit AMR fleet installation for a comparable throughput scenario often runs $600K–$1.5M all-in — potentially lower in raw dollars, but with infrastructure debt that is harder to quantify: the ongoing cost of re-commissioning maps after layout changes, software licensing, and the fleet management platform lock-in.

Neither is definitively cheaper. The cost structure is different.

The switching cost trap

The most underappreciated factor in AGV-vs-AMR decisions is the cost of switching if the architecture turns out to be wrong.

For AGV systems: if you installed wire guidance and want to change a route, you cut and repour concrete. If you want to switch to a different vendor's AGVs, you likely replace reflectors, retune the traffic management system, and recommission the entire integration. The guide infrastructure is partially reusable for the same navigation technology; it's a write-off if you switch navigation types.

For AMR systems: the maps are proprietary. Virtually every AMR platform stores environment maps in a vendor-specific format that is not portable. If you switch vendors, you re-map the facility from scratch. The fleet management software — which carries task assignment logic, integration with your WMS, and zone configurations built over months — is not portable either.

VDA 5050 is the communication standard developed by the German automotive association (VDA) and the mechanical engineering association (VDMA) to let AGVs and AMRs from different manufacturers operate under a common control interface. It's supported by a growing list of vendors including Jungheinrich, Linde, KUKA, and several AMR manufacturers. Specifying VDA 5050 compliance in your RFP is the most practical lever available today to reduce proprietary lock-in. It doesn't eliminate switching cost, but it separates the fleet management layer from the vehicle layer.

Decision framework

The question is not "AGV or AMR" in the abstract. It is: what does your route look like in three years?

Use this test before you spec a system:

Choose laser-guided AGV when:

Route geometry is fixed or changes fewer than 4 times per year
Throughput requirement has a hard SLA (JIT line delivery, time-critical intralogistics)
Payload exceeds 1,000 kg consistently
Traffic is forklift-segregated or the route is a dedicated aisle
Cycle time variance > 20% would cause downstream production impact
Environment has narrow corridors (< 2m) with high passage frequency

Choose AMR when:

Layout changes more than monthly (seasonal SKU rotation, e-commerce reconfiguration)
Mixed human-robot foot-traffic is unavoidable and segregation is impractical
Throughput SLA can tolerate ±15–20% variance
The use case spans multiple work zones with variable pick points
Deployment speed is a priority (days to commission vs weeks)
You are running a pilot and need to defer infrastructure commitment

Evaluate hybrid fleet when:

The facility has both fixed backbone routes (line-side delivery, staging loops) and flexible pick-face operations
A single AGV vendor cannot cover the full use-case portfolio
You have budget to manage two fleet management systems and their WMS integrations

The decision you cannot unmake

Changing navigation architecture mid-program is expensive in direct cost — reflector removal, floor repair, recommissioning — and in opportunity cost: the throughput lost during the transition window, the WMS/WCS re-integration, and the retraining of maintenance staff.

The Stuttgart plant mentioned at the top of this piece spent 18 months on an AMR pilot before switching. They lost the pilot investment and the re-tender timeline. In a JIT automotive environment, 18 months of non-optimal lineside logistics has measurable consequences.

The architecture decision is where to spend your diligence budget. Everything else — vendor selection, fleet sizing, maintenance contracts — is recoverable. The navigation philosophy commitment is not.