Service Robot Acceptance Testing and Commissioning

Why Acceptance Testing Is a Contract Event, Not a Formality

Most procurement contracts tie payment to delivery. That is the wrong trigger. Delivery only proves the unit left the factory; it says nothing about whether the unit meets the specification on your floor, on your network, and against your workload. The correct trigger is acceptance, and acceptance needs a definition written down before the unit ships.

The commercial logic is simple. Once the final payment clears, your leverage is warranty response time. Before it clears, your leverage is the payment itself. A documented acceptance test converts that leverage into objective measurements: coverage rate, navigation success in your aisles, docking repeatability, noise, and uptime over a defined trial window. Each is a number, and a number is either inside the agreed tolerance or it is not.

This matters most where the robot has to earn its money in a live environment. The operational assumptions behind fleet sizing are tested here, and the method for checking them is set out in the throughput planning arithmetic. If the measured coverage rate falls short of the planning figure, the fleet is undersized before it ever starts work.

Factory Acceptance Test vs Site Acceptance Test

Split acceptance into two events with different purposes. The factory acceptance test (FAT) verifies the unit against the specification before it is crated. The site acceptance test (SAT) verifies the same unit against your actual building. A supplier who offers only one is offering you half a process.

Attribute	Factory acceptance test (FAT)	Site acceptance test (SAT)
Where	Supplier facility, before shipment	Your site, after installation
What it proves	Build quality and spec compliance	Performance in your environment
Key measurements	Dimensions, battery capacity, cleaning width, sensor function, firmware version	Coverage rate, navigation success, dock repeatability, noise, network handover
Environment	Controlled test track	Real floors, real obstacles, real Wi-Fi
Who witnesses	Buyer's QA or third-party inspector	Buyer's operations and IT
Typical duration	1-3 days per configuration	5-15 working days across shifts
Failure action	Rework before shipment	Remediation, then re-test

Insist on witnessing the FAT, or on a third-party inspection report with photographs, if you cannot travel. A FAT caught a mis-set cleaning pressure on a recent scrubber order before the container was loaded; the same fault found on site would have cost three weeks of shipping and customs to correct.

What to Measure, and the Thresholds That Decide It

An acceptance test without numbers is a conversation. Define each measurement, the instrument, and the pass threshold in the contract annex before delivery. The table below is a working starting point; adjust the tolerances to the specification you actually bought.

Test	Method	Sample threshold	Fail action
Coverage rate	Run a marked 500 m² area, measure cleaned footprint per hour	≥90% of stated figure	Re-map, adjust brush pressure, re-test
Navigation success	100 recorded runs through a defined route	≥98% without intervention	Sensor calibration, map rebuild
Dock repeatability	50 consecutive auto-dock attempts	≥95% first-attempt success	Dock alignment or charging-contact adjustment
Noise	Sound level meter at 1 m, working mode	Within stated dB(A) ±2	Firmware or hardware remediation
Wi-Fi handover	Walk the full route, log disconnects	≤1 drop per run	Access point coverage remediation on site
Battery run time	Full-charge continuous run to low-battery return	≥90% of rated hours	Cell check, capacity re-test
Safety stop	Deliberate obstacle and bystander approach test	Stops within specified distance, all attempts	Any failure is a hard stop, no partial credit

The safety row is the one with no tolerance. A navigation miss costs you time; a failed safety stop costs you an incident and a liability review. Treat any safety-stop failure as an absolute barrier to acceptance regardless of how well every other metric performs.

The Commissioning Sequence

Acceptance proves the unit works; commissioning makes it ready for daily work. Run commissioning as a fixed sequence, and do not advance a stage until the previous one is signed. The sequence matters because mapping faults look like navigation faults, and network faults look like mapping faults.

1. Physical installation. Dock placement, power supply, drainage if required, and floor preparation. Confirm the dock is level and on a stable surface; a dock that shifts under load will produce intermittent charging faults.
2. Network provisioning. Dedicated SSID or VLAN, agreed channel plan, and coverage verification along every route. The design method is covered in the fleet wireless design guide; the acceptance test only confirms what the design should already guarantee.
3. Mapping. Full-site scan, then a review pass where a human walks the map and marks no-go zones, glass, ramps, and dynamic obstacles. Map once, review together, then freeze the version.
4. Route and schedule setup. Translate the cleaning or delivery schedule into routes. This is where the shift structure from night shift scheduling meets the map.
5. Operational training. Supervisors and named operators learn the water tank, emergency stop, manual recovery, and fault logging. The certification structure lives in the staff induction plan.
6. Trial run and data collection. Two to four weeks of supervised operation, logging coverage, faults, and interventions. This data is the real acceptance evidence and the baseline for the KPI set in the KPI benchmarks guide.

The Documents to Collect Before Final Payment

Acceptance is a paperwork event as much as a technical one. Withhold the final payment until this list is complete and signed. Each item is a document your finance and legal teams will want the moment a dispute arises.

Document	What it must contain
Signed FAT report	Measurements, instrument IDs, photographs, deviations noted
Signed SAT report	Threshold table with actual vs target, witness names and dates
Commissioning log	Stages completed, sign-off initials, issues raised and closed
Firmware and software baseline	Exact versions shipped, so future updates start from a known state
Map version archive	The frozen site map and its date
Spare parts and consumables list	Part numbers, quantities, and reorder lead times
Training completion record	Names, levels achieved, dates
Warranty and service terms	Response times, coverage, exclusions

Hold the documents in one place and version them the way you version your map. When a fault appears six months later, the baseline versions are what make a warranty claim unambiguous rather than a recollection contest. The same discipline applies to the maintenance schedule that follows commissioning, set out in the preventive maintenance schedule guide.

Common Ways Acceptance Slips

Failures in acceptance follow a short list of patterns. Each is cheap to prevent and expensive to discover late.

• No written threshold. The contract says "as specified" but the specification has no numbers. Fix: attach the threshold annex at order time.
• Testing in a clean building. Empty floors pass everything. Fix: saturate the test area with real obstacles and real traffic.
• Payment before SAT. Withholding leverage disappears. Fix: structure payment as deposit, post-FAT, post-SAT.
• Network excluded. The unit is blamed for a Wi-Fi problem the site owns. Fix: include the handover test and agree who remediates what.
• No trial window. A single good day is not evidence. Fix: two to four weeks of logged supervised running.
• Training skipped. Faults go unlogged and the data is useless. Fix: complete induction before the trial clock starts.

Handled this way, acceptance becomes a short, decisive gate rather than a slow argument. The unit either meets the numbers or it does not, the remediation path is agreed in advance, and the fleet that starts work is a fleet that has already proven itself on your own floor.