Autonomous mobile robots (AMRs) are the fastest-moving segment of intralogistics, with the global AMR market projected to reach $14.4 billion by 2030 at a reported CAGR through 2030 [S5]. In a 2026 specification audit, an AMR is a self-navigating wheeled platform that localises via onboard sensors, builds its own map, and plans paths around obstacles — distinct from an AGV, which follows fixed magnetic/QR guides and stops on obstruction.
As of 2026-06-25, OEM directories list 11 manufacturers in the heavy-load AMR category with 24 products and 16 manufacturers with 20 battery-powered AMR models [S3][S4], giving buyers a wide but fragmented shortlist. The fastest way to cut that shortlist is to lock down four spec lines first: payload, navigation, safety standard, and fleet software.
Payload Classes: Light-Transfer vs Heavy-Pallet AMRs
AMR payload spans roughly 5 kg (small tote/cart units) to 1,900 kg (heavy-pallet units), and the 2026 product directories cluster the market into two distinct bands. Light-transfer AMRs (Mobile Industrial Robots, Locus Robotics, OTTO 100/750 class, Hikrobot) handle totes, cartons and small carts typically under 1,500 kg; heavy-load AMRs (OTTO 1500, ek robotics, CeiliX, QENVI, NORCAN) start at roughly 1,500 kg and are built for full pallet, dolly and rack transport [S2][S3].
The OTTO 1500 specs 1,900 kg payload, 2 m/s (7.2 km/h) maximum travel speed, and 1,837 × 1,283 × 351 mm footprint — a useful benchmark for the heavy end [S6]. When payload exceeds 1,000 kg, also verify floor flatness spec (typically ≤ 3 mm/m per ISO 3691-4 duty class), drive-wheel material (polyurethane vs rubber on steel), and whether the platform supports conveyor-top, lift-deck or roller-top modules. For warehouses weighing the AMR business case against fixed storage, the Pallet Rack vs AS/RS 2026 trade-off comparison frames the upstream density question an AMR fleet then serves.
Navigation Stack: LiDAR-SLAM vs Magnetic/QR-Tape
Modern AMRs run LiDAR-based SLAM with 2D safety scanners, often fused with IMU and wheel odometry, mapping a facility without modifying infrastructure; the Intel Open Edge AMR reference stack publishes pre-validated sensor ingestion, environment modelling and SLAM modules on ROS 2 Humble [S5]. Older AGV-style navigation (magnetic tape, QR/ArUco floor fiducials, reflective strips) still ships on some transfer units, and it remains cheaper but inflexible once the route needs to change.
Selection rule from the field: pick LiDAR-SLAM when the route graph changes more than once a quarter, the facility has dynamic pedestrian traffic, or the ceiling/walls cannot host reflectors. Pick magnetic/QR-tape navigation only for stable, dedicated lanes in clean factory cells with low pedestrian density. Confirm minimum scan-plane height clearance (most 2D safety LiDAR sits at 150–200 mm) and the SLAM drift spec (a competent 2D LiDAR-SLAM stack holds < 10 mm position drift in a 50 × 50 m map; vendor-published numbers should be checked against facility layout). Buyers moving from a fixed-path articulated robot cell to a mobile fleet should expect a different safety zoning model — the robot's stop distance scales with payload, not just speed.
Safety Standards and Compliance: B56.5, R15.08 and ISO 3691-4
Heavy-duty AMRs sold into North American industrial trucks are typically CE Marked and listed to ANSI/ITSDF B56.5 (driverless industrial trucks), ANSI/RIA R15.08-1 (mobile robot safety), ISO 12100, ISO 13849-1, and ISO 3691-4 (driverless industrial trucks — safety) [S6]. EU sites additionally need EN 1175-1 and EN 60204-1, with EMC compliance to FCC Part 15 Subpart B and ICES-003 / ICES-002 for North America [S6]. The 2026 OTTO 1500 datasheet is a clean reference line for what a heavy-pallet AMR compliance stack should look like.
Ask the vendor for the PL (Performance Level) rating on every safety function — emergency stop, field reduction, personnel detection — under ISO 13849-1. PL d, Category 3 is the typical minimum for shared-space human/AMR operation; lower than that should be limited to fenced cells. Also ask for the safety scanner field shape in millimetres, not just "360° coverage", because the field defines the floor footprint you need to keep clear around the platform. For battery and charging safety, look for UL 3300 or IEC 62619 references on lithium chemistries if the unit ships with Li-ion; the AGV robot encyclopedia page covers the older fixed-path cousins and clarifies which safety clauses the two categories share.
Battery, Runtime and Charging Topology
Lithium-ion is now the default chemistry for new 2026 AMRs, with the OTTO 1500 specifying 10 hours runtime on a 90%→10% duty cycle and a 60-minute 10%→90% charge window [S6]. That 6:1 runtime-to-charge ratio is a useful baseline: anything worse means opportunity charging will chew into shift availability.
For a single-shift operation, target ≥ 8 hours runtime; for two-shift or three-shift, the AMR must support opportunity charging (autonomous dock contact or wireless) and hot-swap battery modules. Verify the published charge cycle count (typically 3,000–5,000 cycles to 80% capacity for LFP cells), the operating temperature window (most Li-ion AMRs derate below 0 °C), and whether the vendor supports opportunity charging contacts on multiple sides of the platform. Battery-powered AMR directories list 16 manufacturers with 20 products as of 2026-05-29 [S4] — the field is mature enough that lead-time pressure should be on runtime evidence, not on basic availability. Sites that are also weighing lift-automation should cross-check with the VLM/AMR pairing: a Modula vertical lift module handing totes to a MiR AMR is a documented 2026 reference integration pattern [S1].
Fleet Management Software and WMS/MES Integration
Most 2026 AMR platforms ship with vendor proprietary fleet managers — OTTO Motors' dispatcher, MiR Fleet, Hikrobot RCS, OMRON's Flow Core — and they expose REST, MQTT or OPC UA interfaces up to the WMS/WCS/MES. The Open Edge AMR reference stack publishes map building, SLAM, and action planning on ROS 2 Humble with reference algorithms [S5], which lowers the integration ceiling for buyers who can run ROS 2 in production.
Integration checklist: confirm support for the WMS/WCS message set you already run (SAP EWM, Manhattan Active, Blue Yonder), verify the fleet manager scales beyond 50 units in a single map, and ask for traffic-management behaviour at intersections (priority rules, deadlock recovery, queue length). OMRON's 2026-05-18 AMR application guide (OL-450S) is one example of vendor-published hands-on integration documentation [S5]. A common 2026 error is buying a fleet manager that does not natively model human workers as moving obstacles in the path planner — verify this on the demo floor, not in the brochure. Buyers already running an AMR robot cell should plan fleet-manager upgrade paths before vendor lock-in.
Decision Matrix: AMR Class vs Application Fit
Match four criteria to your operation: payload, environment, traffic, and throughput windows. Light-transfer AMRs (< 100 kg) suit picking, sortation, and tote handoff in clean distribution centres; they need little floor prep and integrate fastest with goods-to-person shelving. Mid-payload (100–1,500 kg) units cover rack transport, sub-pallet flows, and line-side delivery, and they tolerate semi-dynamic layouts. Heavy-pallet (1,500–1,900 kg class) AMRs replace forklift trunk loops in 24/7 manufacturing, require flat-floor certification, and demand the most rigorous safety case under ISO 3691-4 and ANSI/RIA R15.08-1 [S3][S6].
Who an AMR is for: warehouses with stable SKU counts and repeatable pick paths, manufacturers with 24/7 line-side delivery, and 3PLs consolidating cross-dock flows. Who an AMR is not for: low-volume facilities (< 8 hours/day utilisation), sites with very poor floor condition (cracks > 5 mm, standing water, oil sheen), and operations needing heavy lifting above roughly 1,900 kg — those still need forklifts or crossed-roller guide-based lift decks, not a single AMR. If cost-per-pick is the headline metric, model it on full utilisation including charging windows; an underused AMR fleet is a worse ROI than a single forklift, regardless of spec sheet.
Two trackable signals to watch through H2 2026: the rollout of R15.08-2 (the Part 2 amendment covering heavier mobile robots and updated safety-rated soft limits) and the appearance of multi-vendor fleet interoperability profiles under the VDA 5050 interface standard. Buyers specifying in mid-2026 should request the vendor's VDA 5050 conformance statement up front to avoid stranded assets when the fleet mix expands.