A robot that has stalled in an aisle is an expensive, immovable object, and the instinct of the person nearest to it is to keep trying. Restart it, nudge it, restart it again. Sometimes that works. Often it burns an hour while orders queue up behind it. In robotics operations the remedy is a stop rule, meaning a cut-off agreed before any fault occurs so that everyone knows when retrying ends and what follows. What comes next is a working audit for your own site: eight checks, each scored pass or fail.
The scope here is operations and process, such as handling stuck units, failing missions and recurring faults. It does not replace the manufacturer’s safety instructions or your site’s safety procedures, which always come first.
Check 1: Is there a retry limit for a stuck unit?
Pass if a written number exists, for example two restarts, after which the unit is taken out of service. Fail if the practice is “try until it works”. Without a limit, repeat restarts hide the real fault and tie up a person who could be doing something else.
Check 2: Is there a time limit per fault?
Pass if each fault type has a maximum on-the-spot troubleshooting time, such as ten minutes for a routine stall, before a technician is called. Fail if nobody can say how long is too long. The limit should be short enough that downtime stays small and long enough for a genuine first check.
Check 3: Does a trigger exist for pulling a unit from the fleet?
Pass if the rule names a signal, such as the same fault occurring a set number of times in a shift, that sends the unit to the maintenance bay. Repeating faults are information. A unit that has stalled three times in a morning is not unlucky; it is telling you something, and the rule makes sure someone listens. The principles behind reliable fleets are discussed in the overview of what running a fleet of guided vehicles involves.
Check 4: Is the sensor-fault response defined?
Pass if a degraded or inconsistent sensor reading has a defined action, usually to stop the mission and flag the unit instead of letting it carry on. Fail if operators decide case by case. Understanding what the inputs mean helps here, as covered in the explainer on how sensors feed industrial robots.
Check 5: Is there a named owner for escalation?
Pass if each shift has a named role, not a vague “maintenance”, with a channel and a response expectation. Fail if escalation means asking around. When a retry limit is hit, the next move should be a message to a specific role with a short note: unit, location, fault, steps taken.
Check 6: Is the hand-off note standard?
Pass if every escalation uses the same four fields: unit identifier, what was seen, what was tried, and current state. Fail if notes are free-form messages. A standard note cuts the first ten minutes of every technician visit.
Check 7: Is there a fallback for the work?
Pass if a unit taken out of service triggers a defined reassignment of its tasks, to other units or to manual handling. Fail if work simply waits. A stop rule that pulls a robot but leaves its tasks orphaned moves the problem instead of solving it.
Check 8: Are the rules reviewed?
Pass if, each month, someone reads the log of triggered limits and adjusts the numbers. Fail if the rules were written once and forgotten. Look for limits that never fire, which are too loose, and ones that fire constantly, which are too tight or point to a deeper fault.
Scoring the audit
Count the passes. Seven or eight means your operation already has good habits; document them. Four to six means you have gaps worth closing this quarter, starting with checks one, two and five because they deliver the quickest improvement. Three or fewer means unplanned downtime is likely costing more than it needs to; pick two checks and fix those this week.
Fixing a failed check without a project
Each failed check can be fixed in about an hour. Write the number or the role on a laminated card at the control desk, brief the shift once, and begin logging each time the rule triggers. Resist the urge to build a large procedure. A one-line rule that everybody remembers beats a long manual that nobody opens.
An example of the audit in use
Consider a mid-sized warehouse with a fleet of guided vehicles and three shifts. A first run of the audit scores five passes. Retry limits and escalation owners exist; the rest do not. The most telling gap is check seven: when a unit is pulled, its tasks sit unassigned until a supervisor notices. On the audit sheet, the supervisor writes a single line: “pulled unit’s open tasks go to the nearest free unit; if none is free, the shift desk is told within two minutes.”
Three months later the audit is repeated. Seven checks pass. Average stall time has not been measured with precision, but the shift desk reports fewer long queues behind stuck units, and technicians report that notes arrive in the same format every time. The remaining failures, a missing monthly review and an undefined sensor response, become the next quarter’s two small tasks.
Common mistakes when setting limits
Two mistakes appear often. The first is setting a limit so tight that it fires on every minor hiccup, which teaches people to ignore it. The second is setting it so loose that it never fires. A good starting point is to look at the last month of stalls, find the typical time to resolve a routine one, and set the limit at about twice that. Adjust after the first review, and keep the changes small.
One more habit from outside robotics
Operators researching unfamiliar topics can borrow the idea. A directory such as BrainSurgeryHospitals is described as a health-information resource with hospital and doctor listings, set out under guides, tests and conditions. A site arranged that way is easy to use on a budget: choose the part you need, read it, and close the window instead of drifting.
Run the audit with your shift leads, keep the scores, and repeat it in three months. The trend matters more than any single result, and it shows quickly whether your limits are doing their job.