Slashdot Mirror
How Long Until We Have a Home Robot That Lives Up To the Hype?
moon_unit2 writes: You may have heard of "personal robots" such as Jibo, Buddy, and Pepper. One journalist recently met one of these home bots and found the reality less dazzling than the promotional videos. Whereas the Indiegogo clips of Buddy show the robot waking people up and helping with cooking, the current prototype can only perform a few canned tasks, and it struggles with natural language processing and vision. As the writer notes, the final version may be a lot more sophisticated, but it's hard to believe that real home helpers are just around the corner.
You mean like 3D printing?
... aren't good at dealing with shit just being anywhere in the house. They like things to be predictable. They're also really bad at identifying objects. I saw a thing in a lab where they had a robot that was doing a pretty good job of recognizing stuff. But are they going to be able to recognize the difference between a clean plate, a dirty plate, and a plate with food on it? And if they can't do that then they can't clear a table. Just a really basic thing you would want a home robot to do. Forget whether it has the arms to move any of that. If it can't tell the difference between these things then it can't clear a table.
When people say "personal home robot" what I think they're looking for is a robotic maid. Rosy the robot. Pick my crap up. Dust. Organize things. Clean. Make me food. Clean up. etc.
The roomba etc are about as close as we've gotten to that. And the roomba has so many fucking problems.
I've decided to stop wasting my time responding to AC trolls/sockpuppets... so if you want a response from me... login.
Give it some time.
As any AI researcher will tell you, we know how the brain works and Geoffrey Hinton's recent paper is nothing short of a breakthrough, and will lead to us having strong AI programs real soon.
We have IBM's Watson, a program that actually understands the information it's processing and will be used to augment medical diagnosis, SIRI, a personal assistant application that actually learns, and MAKO, a program who can do anything on a PC!
IBM is already making neural network chips that implement the way the brain really works, a program the learns the same way that a child learns, and many, many more!
We have courses that teach you AI, and ... it's easy!
Give it some time! We need to let the AI mature like a fine wine, and filter down into consumer devices.
It's coming soon - it really is!
> progress needs to be made in natural language processing, machine vision, and human-computer interaction
Natural language processing on my phone is getting pretty damn good.
I've seen machine vision used on security systems that you might find interesting. The object recognition is quite something-- it catalogs every new thing it sees, tracks it while visible, and is pretty successful at remembering things. Even the ol' Xbone is pretty decent; paired with some Roomba features I would think it could mostly suffice for your basic home robot.
Human-computer interaction is also fairly advanced now. Again, my phone does pretty well with this (and is getting better). Then you have something like Watson, which can actually compete on a game show randomly exploring a huge array of general trivia... yes, it had specialized software written for this purpose, but it shows how a simple, formalized style of language can facilitate a wide range of inquiries.
- - -
The big problem is that all these technologies are proprietary, and the rights to their use are divided among a sea of entities who seem to be addicted to squabbling with each other over short-term monetary anxieties rather than cooperating (must... preserve... teh profitz!!).
Ferengis, the lot of them.
One day, someone will transcend all this, get these existing solutions working together, and build a proper robot.
Until then, every other attempt will seem like a turd in comparison with the features we already know exist *right now* but are locked away in war chests behind medieval fortress walls.
Right now, I'd say that we'll have a home robot that lives up to the hype in about twenty more years. Of course, twenty years from now I expect to be saying the same thing, but that's just because no matter how good we get, the hype will be even better. It's about the ultimate in constantly-moving goalposts.
Good, inexpensive web hosting
We'll have robots that live up to the hype just as soon as we have wives that live up to the hype.
The first one that makes me a sandwich wins.
My, isn't my karma burning nicely...
Left MS Windows for Linux Mint and never looked back!
Vote for Bernie in 2016!
Oh, Roombas are mostly fine. My issue with them more than anything is that they're stupid, don't interface with say a program running on a computer that could make them less stupid, and they've unacceptably high maintenance issues.
The vacuum robots are getting decent. I'd like them to be clever enough to actually have a map of the room and know where they are in it... etc But what really annoys me most about them is that they have too many plastic parts in them. Most of the plastic in the guts of the roomba should be metal... ideally steel. Grit and other assorted shit gets into the gears and that creates friction and the friction creates heat. My last roomba ate itself. It melted its guts out.
You can get after market metal guts to replace the shitty plastic modules that should be made out of stamped stainless steel plate. And that largely resolves the maintenance issues.
However, I still think they should be smarter or should interface with something that is smarter. Have the thing connect via wifi to your network... ideally in a non-mickey mouse way... and then have a more substantial computer do the heavy lifting for it. I'm talking about the sort of thing a Raspberry pi could handle without breaking much of a sweat.
I've decided to stop wasting my time responding to AC trolls/sockpuppets... so if you want a response from me... login.
Emphasis on *SOME*....
In general, keeping a home clean entails a whole lot more than just vacuuming an area that is already free of clutter. It entails keeping the area free of clutter in the first place... This requires that a robot know where everything in the house belongs when it is put away, and will automatically clean and put things away that are left unattended for a sufficiently long period of time. Obviously, it should also know how to do this in a manner that does not in any way jeopardize the health or well-being of the occupants.
"We say we'd be happy with a robot that can clean our homes...." show me one that actually *CAN* clean my home, and we'll talk. Really, an oversized hockey-puck that can only vacuum one floor, can't do stairs, doesn't always cope well with pet fur, and can't figure out that just because it doesn't fit into an area right now because of how things happened to be positioned doesn't mean it shouldn't be vacuumed doesn't cover even half of the job of vacuuming for a lot of people, myself included, and probably not even a tenth of the total job of keeping a place clean and tidy. Forget about expecting cooking or driving kids to school or babysitting them or whatnot.... You claim that people are going to give AI a moving target when it comes to the matter of a robot housekeeper, how about just hitting the fucking original desired target of actually just keeping a house clean?
File under 'M' for 'Manic ranting'
I don't get why vacuuming robots have to be flat disks. Sure, they don't necessarily have to be incredibly tall like an upright, but it seems like they sacrifice real cleaning power for appearances sake. I want a nice middle-ground, where the robot is bigger and more effect, and like you say, made out of durable materials instead of cheap plastic.
Do not look into laser with remaining eye.
Depends on the room. If I could have a bathroom-cleaning robot I would be very happy. My bathroom has a countertop with two sinks, a toilet, a shower with a glass wall and a glass door, and tub, and a couple of towel racks. It also has a scale, a small seat that can be used when sitting at the counter for certain hygiene procedures, a trash can, and some supplies on the counter, in the shower on a shelf, and on a shelf next to the tub. We do not add extra furniture to the bathroom and if we had a robot whose sole purpose was to clean the bathroom we could maintain a routine of where non-fixed objects belong so that they don't interfere with the robot's job.
Do not look into laser with remaining eye.
They have to be that way because they're so stupid. The ENTIRE robot has to fit under your furniture. It doesn't understand how to clean under something without going entirely under it.
As to the size of the things... a bit more size would be a good idea. I'd also like it if they didn't put in any sneaky backdoor business plan into the thing such as "well, you just paid 200-500 dollars for this vacuum that works about as well as 50 dollar vacuum. Lets continue the fuckery by drilling you for whatever the stupid filters and other replacement parts cost!"
The roomba needs more sensors. Ideally some optical ones and possibly some ultrasonic sonar. It needs to be able to connect to a proper computer where it can profit from more processing power. And the stupid bits and pieces need to not be a sides revenue scheme.
I've decided to stop wasting my time responding to AC trolls/sockpuppets... so if you want a response from me... login.
Don't think I'll get it though.
Oh, Roombas are mostly fine. My issue with them more than anything is that they're stupid, don't interface with say a program running on a computer that could make them less stupid, and they've unacceptably high maintenance issues.
I don't understand the stupidity complaint; they are about as smart as they need to be in order to clean the entire floor which is reachable by a vacuuming robot that is 2D constrained. Making them smarter would only be useful for something like "avoid this area"; what other uses do you foresee could be applied, other than "don't go there" and "frequency of operation for a given area"? I suppose you could have "delay operation on Saturdays for 2 hours, as I will be hung over", and other scheduling stuff...
As far as maintenance, there are 3D models for most Roomba parts you'd want to replace, as well as for a lot of modifications, if you are interested. Some of them are here: http://www.yeggi.com/q/roomba/... and there are other sites.
You could either just print replacement parts in plastic, or, assuming you had the right printing hardware, you could have the metal geas you are asking for instead of the plastic ones. NB: My take on the metal gears is that they would typically be a bad idea, since I would prefer the motor tear a stuck gear apart (and leave a live motor) than the motor tear itself apart over a stuck gear, leaving me trying to replace the motor.
They have no memory of the room even though they routinely operate in the same room.
They do not know where doors are, where given activities happen, etc and thus cannot know where the focus of any cleaning should be.
Their programming is too limited to allow for a larger more effective robot because they're too stupid to clean under things properly the way a human would.
As someone else in the thread said, they don't understand that some things they're touching actually should be left alone. One guy was talking about how his roomba just spread cat vomit all over the place because his cat will throw up... as cats do... and then the roomba will roll over it and wipe cat vomit all over the place.
There are an enourmous number or problems with the brain of the thing. And I appreciate that it isn't economical to put that kind of brain power into the roomba... so don't. Most of the robots you see coming out of DARPA these days have most of the brains outside of the robot itself. Its all software running on a laptop or something. And if required for the brain to be in the machine for some competition they just make a cradle for the laptop ON the robot and just put the laptop on the robot.
So there you go... Roombas don't have the brain power they should. They should have a detailed 3d map of the area they operate in, they should know where things get dirty both from logging done by the roomba itself and by what a human would program into it by saying "here are doors".
A bigger robot could do a better job cleaning. Anyone that uses a roomba knows that it takes it DAYS to clean a room and it only keeps rooms clean at all because the fucking thing is scurrying around every day doing about as much cleaning in a week as I would in 30 minutes once a week. That limitation limits how much the roomba can clean. It should be able to clean an entire house. My vacuum cleaner... the one I as a human use... can clean the entire house. But the Roomba can't do that. It can't navigate the house and say "clean this room today" and "that room tomorrow"... and its so inefficient in the way it clean any room that it has to clean the same room several times to actually clean it at all.
And as the man said... cat vomit... or anything gross... gets spread around everywhere.
I could go on... but if you tell me it is as smart as it needs to be, I disagree. The thing it doesn't do which really pisses me off is it can't navigate and doesn't have a map of the house. That's the dumbest.
I've decided to stop wasting my time responding to AC trolls/sockpuppets... so if you want a response from me... login.
LG's Hom Bot maps rooms. It uses a camera pointed at the ceiling to create a map, as well as the usual IR distance and bump sensors. Neato make one with lidar that does a similar thing.
The LG robot is pretty good. It learns the room and then doesn't bump into things as much. The only problem is that it only has one map, so if you say take it upstairs it has no idea where it is and is more or less as dumb as the rest of them. It is really quiet though and does an excellent job of cleaning.
The Neato is junk, unfortunately. The batteries die quickly due to poor power management, and the lidar doesn't seem to help it much.
The thing is, a cheap Lidl robot that only has a bump sensor is not that much worse, and costs a fraction as much. It may be dumb and doesn't have a brush roller, but it's persistent. Simply by covering the same area repeatedly it eventually lifts quite a lot of dust. So there are rapidly diminishing returns for intelligence and advanced sensors.
const int one = 65536; (Silvermoon, Texture.cs)
SJW, n: "Someone I don't like, and by the way I'm a fuckwit" - AC
They have no memory of the room even though they routinely operate in the same room.
I would call this a feature, rather than a drawback, since obstacles such as doors could be open or closed, and obstacles such as dining chairs may move around. Given that, a static map would be a detriment, rather than a benefit.
They do not know where doors are, where given activities happen, etc and thus cannot know where the focus of any cleaning should be.
OK, I already conceded "frequency of operation for a given area"; however for their mop-bot or for shag carpet, either is going to look funny if you don't do all of it, so I'm not sure that's an issue. As long as it does all of it, yes, it's doing unnecessary vacuuming, but no moreso than a cleaning lady would probably do.
Their programming is too limited to allow for a larger more effective robot because they're too stupid to clean under things properly the way a human would.
You mean "by using their arms to move the chair out of the way" and so on? Or do you mean that the robot should be like a cannister vacuum, with the head unit separate from the body unit, in order to get more suction?
As someone else in the thread said, they don't understand that some things they're touching actually should be left alone.
One guy was talking about how his roomba just spread cat vomit all over the place because his cat will throw up... as cats do... and then the roomba will roll over it and wipe cat vomit all over the place.
Yes, well, that's going to be a limitation of pretty much anything below just above dog intelligence. At dog intelligence, it would *eat* the cat vomit.
So there you go... Roombas don't have the brain power they should.
So far, we've got "cat vomit" as a valid argument for more brains. And better sensors, which are typically, at the level of cat vomit detection, going to be more expensive than "better brains".
They should have a detailed 3d map of the area they operate in
Totally not getting this one, unless you plan on putting a quadcoptor onto the thing... it's a 2d device.
they should know where things get dirty both from logging done by the roomba itself and by what a human would program into it
You potentially have a point with the "this needs extra cleaning", but the UI for the "this needs extra cleaning" should probably connect you to the "buy a more powerful model Roomba" Amazon page. This *could* be covered by the "frequency of operation over an area argument"; how would you design as "this area was dirtier than that area" sensor, or conclude that you need to do something about it? I could see a "this is where I eat lunch, so I need it cleaned twice a day instead of once a day because I am a slob" being useful.
by saying "here are doors".
Seriously: transient barriers, which argues for not having a memory of where the barriers have been historically; what if you leave a dining chair pulled out, instead of pushing it in? Should it remember where it was, or should it just adapt its operation?
On doors themselves: it does it no good to know this, unless you either equip all your doors with an "open enough for a Roomba to get through" sensor and transponder, or add automatic door openers so the Roomba can come through the door even though it's closed. This actually does not require as much intelligence as you want the thing to have -- or any at all, if your house system has "I know the Roomba runs from 11:45AM to 12:11AM, so open all the doors at 11:40AM, and then at 12:15AM, close all the ones that used to be closed before, so the human is not subtly disturbed when they come home".
A bigger robot could do a better job cleaning. Anyone that uses a roomba knows that it takes it DAYS to clean a room and it onl
Professional cleaners are very competitive. You can pay a team of 3 to come into your place 1 time / month to do all the things you can't get to.
my roomba usually is to stupid to find it's way back to the docking station, and likes to stop/have the battery die/eat a crippling cable right under the bed or other hard to reach places. i'd really like it to have some cable avoidance and at least a rudimentary knowledge of where it is and how to get back - maybe using wifi beacons that are not as useless as the infrared "virtual wall" ones. and a little intelligence to guide it out of tight corners - it got a knack for finding a way into places that are barely wider than itself, but never finding out again.
1. As to floor mapping being a bad idea because furniture moves. The general goal here is to permit the unit to navigate the house. The Hom Bot seems to negate your furniture issue by mapping the ceiling instead.
Though frankly, I think you could let the thing roam around your house and map it... then import the map into you computer... and explain to the software what is and is not furniture.
And regardless, just because it maps it, it doesn't mean that it won't still be keeping its eyes open. it will try to get from point A to point B first assuming things haven't moved. But if they have it will orient around the obstacle and note that something moved. This could permit dynamic flagging of furniture in that the 3d map of the area will be compared to successive maps of the same area. Then you do object recognition on those things and if you see the same object somewhere else at a different orientation then it gets flagged furniture.
Then you can either basically ignore it for purposes of general navigation knowing that you'll just orient around it when you encounter it where ever it might be this time. Or you might have probability clouds because given pieces of furniture tend to be in predictable places.
2. As to "you need to do all of it regardless"... lets not pretend that you want it to do an equally heavy cleaning everywhere. The corner of the room that nothing ever happens in doesn't need a serious a cleaning as the front door area or something. You want to argue shag or mop... I'm not saying you don't clean parts of the room... unless you actually don't want to... in which case that's the owners decision. But what I am saying is that to be effective you need to have areas where you seriously scrub the shit out of them on a regular basis. There are a few spots in my house that are dirty pretty much after four days. There are other parts of my house that take weeks to get dirty... and the dirt they get is dust that just falls out of the air.
3. Most vacuuming doesn't actually bother with the literal middle of couches and stuff. At least not every time. Generally what people do is they angle the vacuum to get everything you might see under the couch and then leave the proper under the couch cleaning to some quarterly cleaning... the once every 3 month thing that involves actually cleaning the stuff that you were too lazy to clean before.
A robot in this case could simply extend a cleaning head under the couch or whatever that it detected as being something that had to be cleaned this way... and just get everything it could reach. Is this ideal for everyone? Its fine for pretty much everyone and the benefit remember is having a much more robust and powerful machine that can be bigger and thus doesn't suffer from the problems that all ultra compact devices suffer from.
4. As to 2d vs 3d... I get what you're talking about but the sensors need to ultimately be 3d. The map might not need to be 3d but sensors are going to need to be 3d.
And really the 2d map is going to need to be 2d+ because you're going to want a map of the carpet, overhang for couches, and any ledges or stairs. That isn't really a 3d map. You're not quantifying how high or low anything is in the same way you are with the 2d map. But you are specifying specifically where notable features in the 2d map that are nonetheless traversable... or seemingly so until you go crashing down the stairs into a shattered mess of fucking electronics.
So if you want to quibble about the 2D versus 3D... Fine... Quibble accepted *puts on quibbling war paint* (its hot pink and has glitter in it).
I retort... 2D+ *RAWR*
*Flexes to be intimidating* :-D
5. As to barriers that might be there not... just because shit moves around that does not mean there shouldn't be a map.
And before I continue, I was talking about the doors to the exterior... which the stupid thing shouldn't go through anyway unless you want to vacuum the side walk or something.
So inside, if we're talking
I've decided to stop wasting my time responding to AC trolls/sockpuppets... so if you want a response from me... login.
Stuff like laundry? Not that big a deal when the machine does all of the work and I just have to load the wash, move the washed clothes to the dryer, and put them away when they're dry....
As you say, it's different things for different people. If you have an extended family, or have several small children, the laundry is simply absurd. It is something that you have to do basically every day. And it is a pain in the ass to have to check every pocket first (because otherwise you get pens, candy, or other stuff in the wash), un-ball them (since kids are amazingly good as making their clothes into tightly wrapped origami when taking them off), and then at the end folding a zillion shirts that have one tiny sleeve out the wrong way. A robot that could spend an hour or two a day doing laundry would be a magical device for me and many other people, on the order of the changes of washing machines or dish washers in the first place.
The more people I meet, the better I like my dog.
Hmm, my Roomba allows for different scheduling each day of the week, so the Saturday thing isn't a big deal. The 'virtual lighthouse" and "virtual wall" (one unit, with a switch to select mode) allows you to do the "don't go there" or "avoid this area". There's even a widget (virtual wall variant) that can be set up near the cat/dog dish/bed to keep the Roomba away from there if you'd like.
IOW, I'm failing to see the problem with the Roomba.
"I do not agree with what you say, but I will defend to the death your right to say it"
I want a robot to do household cleaning chores. Clean the bathroom, or at least the flat surfaces. Mop the tiled floors. Vacuum the carpets. Clean the windows. Possibly clean the patio and the sidewalks. Possibly deal with some of the kitchen cleaning. If the robot did this kind of cleaning daily then it never would be all that bad and consequently the robot wouldn't have to work very hard on any given day so long as it started out relatively clean.
HEY - robots exist for at least a few of the items you listed here. You really need to buy one if you haven't already! They're certainly not perfect but the vacuum and mopping robots I have work good enough for me. They keep the floors clean for me during the day, I just have to empty their canisters and throw the mop cloths in the laundry. I wish they'd make robots that can dust and vacuum stairs though :\
Cooking dinner? I'd rather do that my self for the moment as I like variety, which I don't think a robot would be as good at compared to extremely repetitive cleaning tasks.
Invert it. Robots prepare the food and you cook it. It's pretty damn near what I do now for my dinner. When I lived in DC, I got used to a place called "Let's Dish" it's one of those places where you prepare your meals from a list of recipes each month. You then stick them in your freezer for when you need them. Later the 'cooking' is basically just adding heat/sautéing/baking and you have a full meal in 30 minutes or so. It ends up being cheaper for me because even with the overhead of "Let's Dish" they buy their ingredients in bulk and my meals end up being about $3-5 per serving. I'm not careful enough in my meal planning to beat that.
So what do I mean by invert it? Robots in the preparation, humans do the final cooking. The early work such as chopping, slicing, packing, marinating, sorting are the kinds of mindless timeconsuming portions which seem to be right in line with the type of work a robot could do. The final cooking and cleanup requires much more subjective actions which are suited to humans. Thus invert it and you could have robots prepare your meals except for the final steps.
(Personally I love the make and take places because I always know that a meal has 6 servings, will take 30 minutes from freezer to plate, and keeps me from impulse purchases when in the grocery store)
Out of modpoints but really liked a post? 1BDkF6TtmmeZ3yqXbz9yhdYVqRYnwFoXDj