[flyinglizard]

My team deployed lots of stuff on Kuka robots from across the world using a VPN with remote pendant software and few PTZ cameras around the robotic cell. It can be done.

The thing though is you still need physical space for integration but a lot of the developed can be fully remote.

[LeifCarrotson]

I've developed a lot of stuff for Fanucs with a similar setup, minus the cameras. A relatively small amount of the work is actually getting in the cell with the pendant and teaching the position registers. A large amount is structuring your programs, writing interprocess communication, setting up inter-machine communication, fault handling, zone exclusions, process monitoring, manual overrides, dry cycle controls, etc. etc. etc.

First, write the programs that set flags, work with IO, branch and jump, fault and recover, and so on. Stick in comments as appropriate that read eg. "Now move from PR1:Home to PR17:Infeed Magazine 2, approaching with the offsets from PR101".

We typically implement some sort of 'dry cycle' mode, where we don't actually check for part presence/energize vacuum generators/spool out wire/spray paint, so you can check for intermittent mechanical problems, loose wires or fasteners, etc. When starting development in this way, it's a sort of anhydrous or dessicated cycle, without any motion whatsoever!

When the robots finally arrive or when you actually go into the shop, only then should you need to move stuff. A benefit is also that you should have a much better idea of precisely where and in what sequence you need to move it, not just the RFQ's description of "the robot does some stuff and assembles the part".