This is NOT Sci- Fi

We had an order cancelled last week (well, reduced significantly so we cancelled it), but normally they're very good. Rarely any substitutions, and the variety is huge.

Too bad there won't be anything for them to pack.

That's a good question and I would love to hear his answer in his typically clear, concise prose.

That's a good video. I found myself mesmerized by the "hive's" activity.

Likely he would talk about the wheels or the lifting mechanism then onto a specific cart and it's duties. Then discuss the controlling cpu and the human programmer.

Dunno. Probably something like: What problem is this expected to solve and what new problems will it create?

I was on the phone yesterday for an hour, talking to several representatives of my credit card company, trying to untangle a problem that some anonymous badly thinking human created. What impressed me was how well these representatives were able to emulate a non-thinking robot.

I froze a few frames of the video and attempted to estimate how many robots were moving in the video. Pulling numbers out of thin air and on the low end, I would guess(tm) about 3,000 robots. If these robots were expected to function without a failure for 3 years (about 1,100 days), the number of mistakes and irritated customers would be about 3 per day. That's tolerable.

I've been in automated warehouses. In most cases, the lights are always turned off (to save electricity). Items are located by barcodes, RFID tags, location, GPS pseudolite, laser tags, etc. The prime safety directive is that no human is allowed anywhere near the robots while the system is running. That begs the question "how does one recover a failed and immobile mobile robot"? There are some options, such as marking its location in the computer memory and having all the other robots avoid it until the system can be shut down. That works well with only a few failed robots, but really badly if failed robots are blocking an aisle or area. So, everything has to be shut down to retrieve the robot, thus making the failure rate the product of the individual reliability of all 3,000 robots.

Let's see how that works with "five nines" reliability or 99.999%. Reliability = 0.99999^3000 = 0.970 or a failure rate of 0.030. At 365 days per year, that's

0.030 * 365 = 1 failure every 11 days or about 1 failure per month. That's also tolerable, except that "five nines" is rather optimistic for things that move, for computer software, and for a world with Microsoft updates. That's probably the best that can be done. It also doesn't include failures in associated services (cloud computers, power outages, unplanned disasters, accidents, and human intervention). Pulling another number out of thin air, my guess(tm) is a factor of 10 worse, which would be 1 failure per day, requiring a system maintenance shutdown every day. That's not so tolerable.

Incidentally, shutting down and restarting a robotic factory can also create its own list of potential problem. Orchestrated in a sequential manner is probably the safest and most reliable way to do both, but is also the method that takes the longest.

In this case, the hypermarket in question has an easy solution available. Install an overhead gantry and use it to lift and transport any robot that has failed. That should allow the system to tolerate failures without the need to shut down everything. Looking at the video, I don't see such a system, which suggests anything between excessive optimism, lack of concern for dealing with failures, or maybe some cost cutting.

Isaac Asimov might then ask how long is this hypermarket expected to last and how long until the Luddites arrive to throw a wrench into the works.

would need several gantries, and they would need to be able to reach past their ends because there are pillars holding the roof up.

perhaps send two tow-robots to carry the failed droid, but if it's stuck with it's reaching gizmo fully extended down a hole that's going to need to be retracted before even a crane can move it.

There does seem to be plenty of headroom for a specialised robot either side to lift the broken one clear.

I wonder who's making money from this activity?

The facility has a big parking lot. The purchasing and sales portions of the operation obviousl aren't as highly automated, though I expect the suppliers and purchasers end up doing most of the keyboard pounding for them.

I bet each part of the organization is its own LLC.

Easier to skim.

RL

Oops.. Feb 2019

July 2021

Spehro Pefhany wrote: ===================

** So they broke Asimov's third law ?

" ... third law is that a robot shall avoid actions or situations that could cause it to come to harm itself."

..... Phil

Thanks.

So much for the pre-cooked meals option. I asked for medium-well done and received incinerated instead.

I guess nobody will be licensing Ocado's patented: "Robotic fire extinguishing device and handling method"

While the factory burns, the lawyers for Ocado and AutoStore busy themselves with lawsuits: "AutoStore Files UK Action Claiming Ownership and Transfer of Ocado Patents"

"Ocado and AutoStore expand robot-assisted technology war to Europe" It makes me wonder how the fires really started.

One article mumbled something about the fire being caused by a crash of several robots. For a crash to turn into a conflagration, some kind of thermal runaway might be helpful, such as found in Li-Ion batteries with a cobalt bearing cathode. I thought the battery industry moved away from cobalt after the 2013 787 Dreamliner fire but possibly not: "If Cobalt Is So Bad, Why Are Some Companies Still Using It In Batteries?"

Or do we feel safer if the supermarket robots designed in self destruct mechanism is to spontaneously combust after going into thermal runaway?

The holes are 21 boxes deep (@1:15). The ceiling is only about 2.5m.

ooh . . . . . insurance. I get it now.

RL

Yes you're right - I'd thought the boxes were like a push-up stack, but it seems not.