Seems easy enough to me to have a 6-8 hour “human shift” where the bots are shut down for maintenance, and then run the bots the rest of the 24 hour cycle.
Checkups only reduce the likelihood of something going wrong, not eliminate it completely. You would still have times where you gotta shut everything down to work on the robots.
If something mucks up (and there can and will be sensors to determine this)
Sensors themselves can go wrong. While you can absolutely have some fail safes, one of the problems about these things working in the real world is that nothing is a simple binary solution, and failures can look like something that just won’t plain show up in software.
What do I mean? Let’s create an example. Robot A has moved very close to Shelf A while looking for a given object. The robot sees it on Shelf B, just behind shelf A, because the part of Shelf A the robot is at is empty. However, the robot is so close to the shelf and the camera is positioned in such a way that it cannot see Shelf A, It believes it has a direct path to Shelf B. So what does it do? It plows forward directly into Shelf A, destroying the entire shelfing unit and destroying hundreds of thousands worth of merch.
Think this is far fetched? Teslas have caused crashes and killed people by this same flaw.
If something mucks up (and there can and will be sensors to determine this) you suspend that section of production until humans can come unfuck it, maximum one day of downtime for a given area unless major overhauls are required.
The problem is that this doesn’t scale well. What do I mean? Let’s say that the chance of a mobility breakdown requiring a total warehouse shutdown in order to safely fix in a year is 1%. Keep in mind this is very optimistic.
This works well enough for 1 robot in the warehouse. You get max 3.65 days of downtime from this little trooper. However, unless this warehouse is basically someone’s shed, this isn’t a realistic number. A decently sized warehouse is going to have something like 40 or 50 of these things and distribution centers? They’ll have hundreds. That means your facility downtime goes from a few days to entire chunks of the year.
Sure, with larger facilities, you could isolate given areas, but you can’t just get away with isolating one shelf or one area, you have to create a safe path for technicians to get there. That potentially means shutting down areas that weren’t affected simply because they’re the only or main way through.
And god help you if an attacker gets on your network. The typical security around these kinds of devices is comically shit. They basically rely on an attacker not getting onto the same network as the devices, something a physical Intrusion will get around pretty darn easy.
Checkups only reduce the likelihood of something going wrong, not eliminate it completely. You would still have times where you gotta shut everything down to work on the robots.
Sensors themselves can go wrong. While you can absolutely have some fail safes, one of the problems about these things working in the real world is that nothing is a simple binary solution, and failures can look like something that just won’t plain show up in software.
What do I mean? Let’s create an example. Robot A has moved very close to Shelf A while looking for a given object. The robot sees it on Shelf B, just behind shelf A, because the part of Shelf A the robot is at is empty. However, the robot is so close to the shelf and the camera is positioned in such a way that it cannot see Shelf A, It believes it has a direct path to Shelf B. So what does it do? It plows forward directly into Shelf A, destroying the entire shelfing unit and destroying hundreds of thousands worth of merch.
Think this is far fetched? Teslas have caused crashes and killed people by this same flaw.
The problem is that this doesn’t scale well. What do I mean? Let’s say that the chance of a mobility breakdown requiring a total warehouse shutdown in order to safely fix in a year is 1%. Keep in mind this is very optimistic.
This works well enough for 1 robot in the warehouse. You get max 3.65 days of downtime from this little trooper. However, unless this warehouse is basically someone’s shed, this isn’t a realistic number. A decently sized warehouse is going to have something like 40 or 50 of these things and distribution centers? They’ll have hundreds. That means your facility downtime goes from a few days to entire chunks of the year.
Sure, with larger facilities, you could isolate given areas, but you can’t just get away with isolating one shelf or one area, you have to create a safe path for technicians to get there. That potentially means shutting down areas that weren’t affected simply because they’re the only or main way through.
And god help you if an attacker gets on your network. The typical security around these kinds of devices is comically shit. They basically rely on an attacker not getting onto the same network as the devices, something a physical Intrusion will get around pretty darn easy.