When AGVs still beat AMRs: high-throughput, deterministic routes

The automation industry narrative has shifted toward AMRs over the past five years. Flexibility, rapid deployment, no floor modification, SLAM-based adaptability — these are real advantages that have driven AMR adoption across e-commerce fulfillment, light manufacturing, and healthcare intralogistics. The narrative sometimes implies that AGVs are the legacy choice: adequate for what they do, but increasingly obsolete.

That framing is wrong for a specific and important class of applications. There are production environments where AGV architecture is not the conservative choice — it is the technically correct one. Understanding those environments and why they favor AGV over AMR is essential for plant logistics managers who are being sold flexibility they don't need at a cost they don't have to pay.

The three environments where AGV architecture wins

1. Automotive body-in-white and stamping transfer

The body-in-white (BIW) shop is one of the most demanding intralogistics environments in manufacturing: heavy payloads (car body carriers weigh 600–1,200 kg empty, 900–2,000 kg loaded), sub-millimeter positioning accuracy required at welding stations, continuous two-shift or three-shift operation, and a production rhythm synchronized to the line takt.

BIW transfer AGVs must arrive at each station within a tolerance window — typically ±15 seconds of the scheduled time — to prevent line stoppages. A deviation of 30 seconds in delivery timing at a welding cell can cascade to a line stop if the next body carrier is not ready at the correct position.

Why AMRs don't work here:

SLAM-based AMRs achieve median cycle times similar to AGVs on clear paths, but their variance is the problem. When an AMR navigating a shared corridor encounters a maintenance cart, a fork truck, or a quality inspector standing at a station, it reroutes or waits. In a BIW shop running at 60 jobs-per-hour (JPH), a 40-second delay at a transfer point is not recoverable. The P95 cycle time — the worst-case performance 5% of the time — defines whether the system is viable in a takt-driven environment.

Laser-guided AGVs running dedicated corridors (or corridors with controlled access via zone interlocks) achieve P95/median ratios of 1.05–1.15. The rare exception — a zone interlock stop while a maintenance team clears — is predictable and schedulable. SLAM-based AMRs in mixed-traffic automotive environments typically achieve P95/median ratios of 1.4–2.0, which makes them unsuitable for takt-critical transfer.

What the Tier 1 automotive suppliers actually install:

The dominant AGV vendors in automotive BIW transfer — Daifuku, KUKA, Comau, Jervis B. Webb (part of Daifuku since 2012) — provide laser-guided or wire-guided unit load carriers with deterministic traffic management systems. These installations are not the result of procurement inertia; they are the result of automotive OEMs running detailed takt analysis and concluding that deterministic routing is not optional at their production rates.

2. High-throughput beverage and food pallet conveyance

Beverage manufacturing and distribution present a specific throughput challenge: high volume, heavy pallets, and relatively fixed routes between packaging, palletizing, and staging. A large beverage plant running 60,000–120,000 cases per day requires pallet transport rates that may exceed what a flexible AMR fleet can achieve on a cost-per-move basis.

A typical beverage palletizing installation uses fork AGVs to move full pallets (typically 700–1,200 kg) from palletizer discharge to stretch wrap, then to staging lanes organized by SKU and shipment destination. The route is fixed: palletizer discharge position → stretch wrap → staging bay. It changes only when the physical layout changes — typically when a new line is added, which may happen once every 3–5 years.

The throughput math:

A laser-guided fork AGV on a clean, dedicated route (palletizer discharge to staging, 120-meter round trip) achieves cycle times of 3.5–5 minutes depending on load handling speed. At 4 minutes per cycle, one AGV can complete 15 moves per hour. A palletizing system producing 12 pallet moves per hour at peak requires one AGV with 20% capacity reserve, or two AGVs with 160% reserve (providing fault tolerance if one unit requires charging or maintenance).

The cost-per-move comparison matters: a laser-guided fork AGV moving 15 pallets per hour over 16 operating hours per day, 250 days per year, performs 60,000 pallet moves annually. Against an all-in capital cost of $150,000 and annual maintenance of $15,000, the 5-year cost per move (at $750,000 total 5-year cost) is approximately $2.50 per pallet move. AMR fork platforms at comparable payload capacity typically carry 20–40% higher unit prices and may achieve lower throughput on fixed pallet routes due to their broader safety margins.

This math varies significantly by site; the point is to model it explicitly rather than assume that newer = cheaper.

The JBT case:

JBT Corporation's laser-guided fork AGVs have a documented history in beverage distribution and food manufacturing. In beverage distribution applications, JBT's systems handle pallet transfer from refrigerated discharge conveyors to trailer loading — a narrow, fixed-route application where laser guidance provides the positioning accuracy needed for consistent trailer loading without manual assist. The same logic applies to brewery, winery, and soft-drink bottling environments.

3. Regulated manufacturing (pharmaceutical, medical device, semiconductor)

In GMP-regulated pharmaceutical manufacturing and ISO Class cleanroom semiconductor fabs, AGV navigation architecture has advantages beyond throughput: auditability, validation, and safety classification.

Validation and 21 CFR Part 11 compliance: Pharmaceutical manufacturing environments operate under FDA cGMP regulations that require validated computerized systems. The AGV control system must be validated — every route, every parameter, every software configuration must be documented and change-controlled. Laser-guided AGV systems with deterministic routing are easier to validate than SLAM-based systems because the system behavior is explicitly defined by the route specification and the traffic management rules. SLAM systems, by definition, compute dynamic paths — validating a system where the path is computed at runtime is a larger IQ/OQ/PQ effort.

ISO cleanroom classification: Semiconductor fabs (ISO Class 3–5) require that all equipment — including AGVs — meet particle generation specifications. Cleanroom-rated AGVs are built with sealed drive units, low-outgassing materials, and HEPA-filtered exhaust paths. The market for cleanroom-rated AGVs is dominated by a small number of specialists (Murata Machinery, Daifuku, Brooks Automation) who have invested in the cleanroom engineering that most AMR vendors have not.

Functional safety: AGVs operating in environments with specific safety classification requirements (SIL 2 or SIL 3 under IEC 62061, PLd or PLe under ISO 13849) must have safety architectures that can be assessed and certified. AGV vendors with 20+ years of automotive BIW experience have certified SIL 2 safety architectures available as standard. Many AMR vendors are newer entrants with safety architectures that have not been through the same certification depth — this is changing rapidly, but it matters for current procurement decisions in regulated industries.

The performance comparison: what the data shows

The academic literature on AGV vs AMR performance in manufacturing is growing. A 2024 study published in the International Journal of Advanced Manufacturing Technology reviewed implementations across 14 manufacturing facilities and found:

• In takt-driven production environments (automotive, electronics assembly), AGV-based systems achieved median cycle time variance (P95/P50 ratio) of 1.08–1.18 vs 1.35–2.10 for AMR systems in comparable environments.
• In flexible fulfillment environments (e-commerce, multi-SKU distribution), AMR systems achieved 18–34% better throughput per unit cost due to dynamic rerouting and reduced dwell time at pick faces.
• Facilities that had converted from AGV to AMR reported higher satisfaction in flexible-route applications but lower satisfaction in fixed-route, high-volume transfer applications.

The conclusion: the performance gap is real, and it is direction-dependent. AMRs outperform AGVs in flexible environments. AGVs outperform AMRs in fixed, high-volume environments. The mistake is applying either architecture to the wrong environment.

Route stability: the decisive test

The single most predictive factor for AGV fitness is route stability — how often the route geometry needs to change.

Use this test before choosing architecture:

If the test returns 4 or more "favors AGV" marks, a laser-guided or wire-guided AGV architecture is likely the correct choice for your environment, regardless of how the vendor conversation starts.

The hybrid fleet argument

The most operationally sophisticated approach in large facilities is not a binary AGV-or-AMR choice but a hybrid fleet that deploys each architecture in the environment it serves best.

A common pattern in food and beverage manufacturing:

• AGVs on the backbone: pallet conveyance from palletizer to staging, line-side delivery of bulk raw material, finished goods transfer to dock staging.
• AMRs on the edges: goods-to-person picking at a mixed-SKU returns processing area, cart delivery between process areas where the destination changes by order, visual inspection tour vehicles that travel between QC stations on variable paths.

The integration challenge in a hybrid fleet is the fleet management layer: AGV and AMR systems from different vendors typically do not share a common FMS. VDA 5050-compliant vehicles can be managed under a neutral FMS (Kinexon, Mujin, or a neutral integrator-provided platform), but this adds integration cost and complexity. The hybrid fleet argument is strongest when the two fleets are genuinely separate — different routes, different zones, no task handoff between them — and weakest when tight coordination between AGV and AMR movements is required.

What to tell the AMR vendor

When an AMR vendor presents a proposal for a high-throughput fixed-route application, ask these specific questions:

1. What is the P95 cycle time for this route, not the median? Get a contractual P95 SLA in the proposal, not a median guarantee. If they cannot commit to P95, ask why.

2. What happens when the route is blocked? Model the specific blockage scenarios for your environment (maintenance cart, forklift cross-traffic, a spill) and get the vendor to walk through the AMR behavior — reroute time, fallback path, escalation to human.

3. What is the throughput degradation at 80% obstacle encounter rate on the primary path? This is a stress test. A reliable AMR vendor can run this simulation. If the answer is more than 15% throughput reduction, the AMR design is not appropriate for your required throughput with your actual traffic conditions.

4. Can you provide a reference installation at a comparable throughput rate and route complexity? Not an e-commerce warehouse — a manufacturing environment with comparable JPH rates and fixed-route requirements. If no reference exists, the deployment is de facto a pilot.

The goal is not to eliminate AMRs from consideration. It is to ensure the evaluation uses the correct performance metrics for your environment — not the metrics that make the AMR look best in a demo.

Read next: Running an AGV pilot: what should pass and what should fail it