The AGV Preventive Maintenance Checklist: Skip These and Your Fleet Slows Down

Here's what dozens of AGV warehouses taught me: a slowing fleet needs an AGV preventive maintenance checklist, not new parts. The complaint is always the same: the vehicles were lightning fast at go-live, and two or three years later they start "phoning it in." Slower speeds, more mission aborts, longer charging times. Nine out of ten managers assume the fleet is aging out and needs new parts. In my experience, eight out of ten cases are solved by something far cheaper than new parts: the checklist. Here is the one I hand out on site.

The principle: AGVs don't break, they wear out

Most people treat an AGV like a household appliance — working means fine, broken means replace. Think of it instead as a human knee: the cartilage wears down silently, and by the time it hurts, you are already late. A fleet that "slows down" is never a sudden event. It is five wear channels leaking at once:

Channel 1: rising battery internal resistance. Most modern fleets run lithium packs, and the BMS quietly logs every deep discharge. Think of it as the battery's arteries narrowing — the same stored energy can no longer deliver the same peak power, voltage sags under acceleration, and the controller derates speed to protect the pack. Deep cycling is the number-one killer; a fleet run on shallow opportunity charging will show dramatically better pack health at year three than one run to empty every shift.

Channel 2: drive wheels wearing down. AGVs measure distance by wheel revolutions (odometry). Lose 2–3 mm of wheel diameter and the stopping point starts drifting. Worse, when the left and right wheels wear unevenly, the vehicle pulls slightly and the navigation stack spends its time correcting — so it never reaches full speed.

Channel 3: dusty laser scanner lenses and navigation cameras. Warehouse dust is a slow poison. A film of dust on the lens makes the LiDAR's view of the world jitter. The symptoms: emergency stops with nothing in front of the vehicle, localization jumping around. It is like driving with dirty glasses — would you go fast? Neither will the vehicle.

Channel 4: oxidized charging contacts. Blackened, carbon-fouled contacts raise contact resistance, so less energy actually makes it into the pack — and you get "fake full" charges: the display says 100%, and twenty minutes of runtime later it reads 70%. Every fleet has one or two "power hogs." Eight times out of ten, it is the contacts.

Channel 5: drifted navigation maps and reflectors. Racks get moved, reflectors get clipped by forklifts, floor lines get repainted — but the map is still the one from three years ago. The vehicle follows an old map through a new world, and everything feels awkward.

A textbook case: an auto-parts DC I visited in 2023, 18 laser-guided AGVs, year three of operation. The dispatcher said the fleet was "getting dumber." One afternoon of diagnosis found exactly two root causes: charging contacts that had not been cleaned in eight months — black as coal, vehicles charging three hours for two hours of runtime — and drive wheels worn unevenly by nearly 3 mm, with stop positions drifting almost 10 cm. Two technicians, one afternoon: clean the contacts, swap the wheels, recalibrate. Mission aborts dropped from twenty-plus a day to single digits the next morning. Not one major part replaced.

The dividing line: what your team does vs. what belongs to the OEM

A checklist only works if the work is split correctly. My dividing line is simple: if it doesn't break a seal or open the drivetrain, your team does it; if it touches the "heart," call the OEM.

And here I will take a shot at a fashionable pseudo-need: vendors now sell "AI predictive maintenance platforms" with per-vehicle annual fees and gorgeous dashboards. For a fleet under 50 vehicles, that money is pure tax. The onboard controller already logs everything — abort counts, charge durations, cell voltage spreads. Export it, drop it into a spreadsheet, pivot by vehicle. The two vehicles with the most aborts are the two that need maintenance. The data is already in your hands; what has always been missing is someone looking at it once a week.

The checklist itself, by frequency

Print this AGV preventive maintenance checklist and tape it to the wall — a checklist in a drawer is the same as no checklist.

Every shift / daily (5 minutes at handover):

• Walk around: dents, loose panels, dragging cables — log anything you see
• Listen: bearing failure always announces itself with noise first
• Look at the charging contacts: shiny is healthy; black or green gets cleaned the same day
• Check wheels for wrapped debris: zip ties and stretch film are wheel killers

Weekly:

• Wipe lenses: clean, lint-free cloth, circular motion. Note — circular, not back-and-forth. Linear wiping drags grit across the lens and scratches it; scratches hurt the scanner worse than dust does
• Test every e-stop: press it on each vehicle, confirm it stops and reports
• Safety field test: block the scanner zone with a carton and confirm the vehicle stops

Monthly:

• Measure drive wheel diameter with calipers, log it, watch the three-month trend — the fastest-wearing vehicle gets its wheels first
• Check cell voltage spread in the BMS: a pack whose spread keeps widening gets flagged early
• Verify reflectors, QR codes, and magnetic tape: shifted, damaged, or blocked
• Spot-check fasteners: AGVs vibrate all day; loose screws are the norm, not the exception

Quarterly / semi-annual:

• Bearing lubrication or replacement (follow the OEM manual — don't invent your own intervals)
• One full charge–discharge calibration so the BMS relearns the pack's true capacity
• Map review: re-survey if the rack layout changed; even if it didn't, validate localization accuracy on two sample vehicles
• Measure actual charger output voltage and current — don't trust the panel display

One decision framework for aborts, three tiers — don't call the OEM for everything:

L1 — handled on the floor (supervisor, under 10 minutes): dirty lens, dirty contacts, wrapped debris, blocked reflector — wipe, clean, move, resume. L2 — scheduled repair (within the week): wheel wear beyond limit, one vehicle topping the abort chart three days running, abnormal cell spread — slot it into the weekly plan, swap when parts arrive. L3 — OEM call: controller alarm codes you can't find, LiDAR data anomalies, pack faults — don't open it yourself, pick up the phone.

Run the numbers: maintenance is a cost, downtime is a bigger one

Another site, a 3C warehouse in 2024: 24 lithium AGVs on opportunity charging — topping up whenever idle. The dispatcher, trying to be helpful, made a rule: "no vehicle leaves until 100%." So vehicles sat at full charge on the dock all night and ran the next day. The BMS data told the story: after eighteen months, these packs had lost usable capacity far faster than packs run on shallow cycling. We changed the rule to "charge when you can, roll at around eighty percent," and three months later the capacity curve had visibly flattened. What kills lithium fastest is not use — it is sitting fully charged doing nothing. Like a person who overeats and never moves.

The math is simple. Running an AGV preventive maintenance checklist costs you a couple of hours of part-time inspection a week plus consumables — essentially rounding error. One unplanned stoppage — say two dead vehicles blocking a 20-vehicle line for two hours — means piled-up orders, idle labor, and overtime to catch up. Tens of times the maintenance cost. Whatever you "save" by cutting the maintenance budget won't cover one major stoppage's overtime bill. The OEM will never do this math for you.

One uncomfortable truth to close on: most AGV fleets are not "used to death." They die of neglect. A checklist on the wall with an empty execution log is the same as no checklist.

Whether the money is well spent comes down to a single number: mission aborts per vehicle per day. Get it under 0.5 and the checklist has paid for itself. If it won't come down, the checklist is hanging on the wall with nobody executing it — that is a management problem, not a technical one.

By Da Liu, Automation Engineer