How would these cameras identify spills? I would imagine it would be hard to detect a clear liquid spill from a ceiling camera, but maybe I'm wrong. I would imagine if ceiling cameras could do this job, they would, as it seems way less expensive and intrusive.
Polarization imaging [1] could probably do it. Granted, the sensors are expensive and lower resolution than one might like.
Another way is to measure surface specularity. Use one or more IR spotlights at different angles to the surface and strobe them on and off. A wet surface will have a different angular variation in reflectivity vs a dry surface, unless the floor is very shiny.
Even if it can't, it has wheels and, I assume, distance measuring sensors/encoders, etc. Slip is "easily" detected by monitoring wheel response to motion commands.
Of course they'll want to "expand the capabilities". With facial recognition, and their database of you and your family's shopping habits, it'll be happy to sidle up to you with helpful suggestions: "Your wife's phone says she's driving like she's really stressed this evening; why don't you pick up that bunch of flowers to help her feel better? And she'd love this wine with dinner. Wait, that's not the shampoo she usually buys; get her the one with coconut oil."
Unless its going to physically stock the shelves, motion is so expensive and difficult to maintain and cameras and other sensors are so cheap, I can't ever see a mobile robot being more cost effective than a sensor blanket.
It makes natural sense for the robot to move from identifying spills to cleaning them up. Lots of people say "well somebody just sees a spill and sends a human to clean them up" which doesn't work when the first person to detect the spill is the old lady who slips in it and breaks her hip.
If Marty is roaming the aisles at a reasonable speed (i.e. not so fast as to endanger shoppers!), there's still a substantial chance that the first "person" to detect any given spill is the elderly person with the fragile hip, while Marty ambles around the other side of the store.
So how many robots are needed per store if they're to have more effective coverage than asking the employees who roam the aisles tidying and re-stocking shelves, helping customers find things, fulfilling click-and-collect orders, etc., to also report spills?
Or perhaps these stores don't have any such employees on the floor -- they're just an array of unserviced aisles, with some staff at the checkouts?
>If Marty is roaming the aisles at a reasonable speed (i.e. not so fast as to endanger shoppers!), there's still a substantial chance
Still substantial but less than without it.
Most of the restocking and cleaning happens when the store is closed. Customers don't want to have people getting in their way stocking shelves and you don't want to be mopping floors (other than cleaning up spills) when customers are walking down them.
Cameras with wide angle lenses every 10 feet on opposing shelves. Still cheaper. Much less maintenance. Software assembles a seamless view of every shelf in the store in real time. No need to wait for the bot to wander by.
Now you've created the need to re-wire shelves for power + data, map the cameras to store positions, keep people from stealing them, maintain/replace them, etc. I'm not at all convinced that would be cheaper in fully burdened cost than a few roving robots.
Cameras every 10 feet on each side of an aisle. Now you're talking dozens and dozens of cameras. Definitely not less expensive than the glorified Roomba they have roaming the aisles that I've seen.
>Another said the robot is currently broken and stored on the charging dock, a $35,000 decoration.
Thats a lot of cameras you could buy right there. I was figuring around 100 for a medium sized store. (40 ft isles, 12 of them, 2 cameras every 10 feet). How much is a little wifi camera that can send back a few images per minute?
>How much is a little wifi camera that can send back a few images per minute?
I was discussing cameras with someone who derives a non-negligible portion of their income from streaming 4k video. The "gold standard" as far as they are concerned is a $60 Logitech webcam. Swap out the cord for a wifi chip, swap out the consumer housing for something that clips to a shelf and can be secured with a self tapping screw and power it with a cheap solar panel and I still can't see it coming in above $100.
