my tray count says empty but there's a sleeve there

Count saves after retreat. Sleeve was down, gripper hadn't finished moving away.

That explains this case. Moving the save earlier only moves the awkward gap somewhere else. Ask your programmer for named transfer states and a recovery screen that can pause for a physical check when the last action is uncertain.

So 'released' isn't automatically proof of 'in the pocket'. Got it.

Exactly. It can be a command the program issued, rather than confirmation of where the sleeve ended up. What feedback do you actually have on the tool and tray? That decides how much recovery can be automatic.

We saved an 'in progress' flag in our training job, which at least stopped restart pretending everything was fine. Still needed a human to settle where the part was... not glamorous, but nobody loaded an occupied pocket twice.

Finger position only. No pocket sensing. Cheers for the distinction, Luis.

Don't let someone turn that human check into a yes button people hit to get running. Show the pocket number and what needs checking, with the cell in its established safe inspection condition. Sorry, 'human check' sounds simpler than the handover actually is.

Our screen just says resume. That'll need changing before anyone else gets this job.

Give the programmer that list of cases before they build the screen. Include a restart where nobody knows who last touched the tray. The recovery decision needs to restore an agreed state, not merely dismiss a warning.

Made the list. Missing sleeve and occupied pocket currently get the same answer, which looks daft now.

It's a useful discovery while this is still a test station. Keep those as separate cases in the handover. If the physical state cannot be established, stopping for someone qualified to resolve it is a valid recovery outcome.

Did your programmer say whether those saved states survive a controller restart? Ours survived stopping the job but vanished after power off, which made for a rather short celebration.