"AMR Selection: Payload, Navigation, and Battery Life — Three Hard Specs"

The Phenomenon: Every AMR Spec Sheet Looks Great, Every Buyer Gets Confused

I have been reading AMR vendor selection guides lately, and one impression keeps coming back: the spec sheets keep getting prettier, and the buying decisions keep getting harder. Rated payloads span from 500 kg to the 1,500 kg class; the navigation row lists "laser SLAM," "vision-laser fusion," and "QR-code assisted" in rotation; battery life is almost universally rated at 8 hours or more. Every vendor's numbers look sufficient — it feels like you could close your eyes and still not go wrong.

But in my field conversations I keep hearing the same kind of story: three months after deployment, the problems surface all at once. An e-commerce warehouse in eastern China brought in a fleet of underride AMRs in late 2024, rated at 1,000 kg; once 800 kg racks were loaded, the chassis pitched visibly under acceleration, and the operator had to cap loads at 650 kg — a third of the fleet's capacity written off overnight. A 3PL warehouse in the south chose laser navigation, only to discover that expansion joints and reflective epoxy flooring made the LiDAR lose localization constantly; commissioning dragged on for two extra months.

The premise: AMR selection comes down to three hard specs — payload, navigation, and battery life. But the "hardness" of each is not in the spec sheet; it is in how well the spec matches the scenario.

Selection Spec 1: Payload — Between Rated and Usable Lies a Dynamic Duty Cycle

The first distinction is between rated payload and usable payload. Rated payload is measured under ideal conditions: flat floor, constant speed, centered load. Real operations involve constant acceleration and braking, turns, and small ramps, and the load's center of gravity is rarely perfectly centered. Across vendor technical documents, the industry convention is to derate usable payload to roughly 70–80% of the rated figure — standard practice in mobile robotics, inherited from the AGV era.

The second dimension is dynamic loading. Two vehicles both rated at 1,000 kg, one accelerating at 0.5 m/s² and the other at 0.8 m/s²: the faster one subjects cargo to far greater inertial forces under braking — a higher risk of load shift or tipping for tall cartons or irregular items. So payload cannot be read as a bare kilogram number; it must be read together with acceleration, turning radius, and emergency-stop distance — jointly, they determine whether "this payload" can actually run safely at your takt time.

A rough but practical framework: take your heaviest single handling unit (including pallet or cart tare weight), multiply by 1.25–1.4, and treat the result as the minimum rated payload to shortlist. If the math says 600 kg, don't pay for a 1,500 kg flagship — each payload class up drives chassis weight, battery capacity, and price upward.

Selection Spec 2: Navigation — Three Routes, No Best One, Only Best Fit

The three mainstream navigation methods fit in one table:

A caveat: these accuracy figures are commonly cited orders of magnitude in public industry materials; they vary by vendor and conditions, so verify against the latest vendor specifications and on-site testing.

The table hides a critical trade-off: precision and flexibility cannot both be maximized.

For a manufacturing line-side warehouse whose layout barely moves, QR codes are the safe choice: most stable on accuracy, though codes get crushed by forklifts and covered in dust — a permanent maintenance cost — and every route change means re-laying and re-calibrating. For an e-commerce DC with churning SKUs and wave patterns, the flexibility value of laser SLAM far outweighs occasional jitter — jitter can be mitigated with engineering measures (reduced speed, reflective markers), but the rigidity of a system needing reconstruction for every layout change cannot be bought away at any price.

Selection Spec 3: Battery — The Underestimated Charger Ratio

The most common battery number — "8 hours per full charge" — is measured at standard load and speed; heavy loads, low temperatures, and frequent stop-start cycles all cut into it. More importantly, endurance itself is rarely the bottleneck — the charging strategy is.

Two mainstream approaches. Automatic charging: the AMR drives itself to a charger at low battery; a full charge takes roughly 1.5–2.5 hours (order of magnitude, varies by vendor). Unattended, but the vehicle is down while charging, so the fleet must be oversized — the rule of thumb being that if charging occupies 20% of the duty cycle, keeping N vehicles online needs roughly N × 1.25 vehicles. Battery swapping: the module is exchanged in minutes, so utilization is extremely high, at the cost of spare batteries (typically 1.2–1.5×), a swapping station, and full lifecycle management of the battery pool.

The counterintuitive takeaway: many buyers agonize over the endurance number, then discover the real cost driver is "the extra vehicles bought just to cover charging." A 20-vehicle project with a poorly calculated charging strategy might need 24–25 vehicles to guarantee 20 online — those extra units, plus electrical work and floor space for chargers, can easily be the most underestimated line item in the whole project. Conversely, a two-shift warehouse with a clear 2-hour break window can complete automatic charging inside that window with zero redundancy. Once again, AMR selection comes back to the scenario.

Attribution: Selection Failures Are Scenario Failures

The three specs point to one root cause: selection failures are almost never about misreading parameters — they are about never auditing the scenario. Four factors: aisle width caps vehicle footprint and turning radius — with a 2-meter main aisle, a 1,500 kg-class vehicle may not turn at all; floor conditions determine whether a navigation method is viable — always demand an on-site survey before signing, not a quote from a floor plan; human-machine traffic intensity eats throughput through safety slowdowns — build in margin; and business volatility sets the sizing logic — peak volume can run 3–5× average, so "how do we handle the peak" must be in the plan at selection time, not discovered later.

None of these four appears on a vendor spec sheet. They can only come from the buyer's own knowledge of the warehouse. AMR selection is therefore less a technology decision than a scenario audit — the more thorough the audit, the more meaningful the spec sheet becomes.

Trend Call: Navigation Technologies Are Converging

Based on industry reports and vendors' public roadmaps, my observation is that navigation routes are converging, with laser SLAM becoming the default option, on three grounds. Supply side: purely QR-code or magnetic-tape navigation keeps shrinking in new launches; laser SLAM or laser-vision fusion is standard on 2024–2025 models. Demand side: layout-change frequency keeps rising — quarterly re-slotting is moving from exception to norm, structurally disadvantaging any method requiring construction work. Technology curve: LiDAR costs have fallen steadily — a 32-line unit now costs roughly what a 16-line unit cost a few years ago, directly pulling down laser-SLAM system cost.

But convergence is not the endgame. My call is that the next two to three years settle into a "laser SLAM primary, QR-code assisted" pattern: QR codes for fine positioning at dock points demanding extreme precision, laser SLAM for open-area travel — a hybrid already appearing in leading vendors' products. On the timeline, I expect purely magnetic-tape/spot navigation to exit mainstream selection by around 2027 — not a technical verdict, but the economic result of construction costs and flexibility demand acting together.

An Open Question

Back to the opening question: every spec sheet looks great, so how do you choose? My advice is counterintuitive — don't look at the spec sheet first. Answer four questions instead: What is your heaviest handling unit? How many times will you change the layout in three years? What is your floor like? How many times the average is your peak? The answers will narrow the three hard specs to a very small shortlist on their own.

And a longer-term question: as navigation converges and hardware commoditizes, where will vendors' real differentiation land? My early read points to the fleet management system — multi-vehicle coordination, depth of WMS and conveyor integration, and self-recovery from exceptions. But that is another topic, for another day.

By Dr. Chen, Supply Chain Researcher