Hacker News Reader

Comment by ghshephard | Hacker News Reader

We're about 10-15 years (at least) from the point at which room cleaning can be done robotically. It's actually an incredibly complex task, which requires a lot more in terms of executive skills/dexterity that just aren't on the near-term roadmap for robotics or their control systems.

Even doing something simple, and highly constrained like changing the linens on the bed, is still 5 years out (at least).

Now, what they might have done, is constrain the environment, such that the tasks can all be automated. I.E. designing the bed/linens, restricting what/where things can be left in the room, or alternatively, restricting what types of cleanup work will be done.

That would be an interesting approach, but far from the much more challenging task of creating a generalized robotic solution to hotel automation.

View on HN

Oh dear. I now have images, thanks to numerous police/murder TV series, of a robot finding a dead person on the bed, carefully folding them up to stick in the waste chute, replacing the bloodied linen with clean linen, and going on it's merry way with no-one the wiser...

One would have to think that simple weight sensors may prevent such automated mishaps.

To align with the current guest service model, the robot would need to have a use case for "guest sleeping in bed during room-cleaning hours without a 'do not disturb' indicator".

Depending on how smart that solution is, the story might just turn into "mummified hotel guest evicted after 2 year stay; was discovered only when credit card expired."

"Constraining the environment" is a perfectly valid approach, IMO. At least it is if your question is 'how do I automate a hotel.'

The idea of human-like robots performing tasks in human-like ways is (again, IMO) an analogy that makes it easy for people in the 1950s to imagine the 2050s. How will check in work? Robots working the front desk. Cleaning? Robot maids. It's a placeholder for technology that doesn't exist yet but is ultimately expected to to some job.

To update the analogy now that we're closer to achieving the reality, lets imagine a hotel that is a machine Instead of a hotel staffed by machines. In that context, specialized beds, sheets, buildings, and other adaptions to the practicalities of machine abilities are fine. After all, our regular hotels were built specifically to be convenient for human workers changing sheets or delivering room service.

I think that's ultimately how a lot of technology will progress, as part of a complex of people and machines. Self driving cars need to work on "human roads" because these are the roads that exist. But, I think adaption will go all ways if the tech succeeds. Roads will be built or updated to be more suitable for self driving cars. Humans will adapt too.

That's a great analogy - the robot maid is a placeholder for expected automation, but the form is not known yet. I hadn't thought about it like that.

I often tell people I have a house full of robots. I have a robotic clothes washer and a robotic dish washer. I also have a robotic vacuum cleaner. All are specialized tasks, automated, but for whatever reason only the vacuum is usually called a robot.

The reference point is factories, I think.

Some of the things they have look like robots as most people imagine them. An arm spray painting cars. But in most cases its a gradual process of using less human resources and more machinery. Futurists call it robots. Engineers call it automation. Managers call it efficiency. Economists call it capital. Everyone chooses the abstraction that works for whatever it is they are trying to solve.

The point is, sheets must be freshened. If factories are the reference point, I think it shows that marginal automation is harder which makes sense considering that we start from the easiest. Today's factories use far fewer people per unit of output, but few are 100% machine. People still work in factories.

Fully automated hotels sounds like a really cool tech project or PR project, maybe an idea lab. But for actual wide adoption, I would guess the best chance is the kind of engineering that turned 1000 person factories into 100 person factories. Some of it may be pretty low tech.

Anyway.. how about a sheet dispenser in the headboard so that sheets can be pulled on to the bed in one smooth motion? I might get one for home if they come in cheap enough. I'll be ready if they invent robot girlfriends.

Heck your wall thermostat is a robot. Its the textbook example of the definition of a robot - an automated machine that makes decisions for you and takes action on your behalf.

I meam, it's already been done in the industrial area. Factories are mostly packaged around machines. There is still a human component but that decreases per iteration. The real challenge is to make the automation as safe and non-obstrusive as possible. Since humans tend to err on the wrong side of the average intelligence scale. This calls for an approach that requires changing everything that has been done up until now. A self cleaning bed is no longer a frame with a mattress and some sheets. It is now a sophisticated machine that is able to move itself in an algorithmic manner to achieve a defined goal. Think about it Beds are simple technology. Thisn would turn them into something as complex as an elevator.

If we move away from engineering for humans to machines, all of this will be more feasible. Problem is that it will take decades before that happens. Even if the singularity even happens because smart machines cannot overcome human stupidity and greed.

Saying that robots are about to overtake people in most jobs is like saying that humanity has evolved to a point where it doesn't mind losing money or changing it's ways.

Seems to me the future has always been 15 years away, yet usually fails to materalize.

These problems that are being solved are non-trivial. We don't really have the computer vision, the robot dexterity, the high capacity batteries, the fuzzy logic, etc to make this work. We dont have it today and we probably won't have it anytime soon. Heck, we don't even have a useful home robot yet! Where's the Apple //e or Commodore 64 of robots?

If anything makes sense, its to build an automated hotel from scratch, not try to shoehorn in solutions. Self-cleaning toilets, self-making beds, roomba-like devices, etc so that a majority of the housekeeping work is done. In other words make your problem fit the limits of technology and don't be afraid to break from tradition. We didn't build robotic horses in the early 20th century. We built cars with four wheels and we paved down expensive roads to drive them on. Dirt roads passable for horse and buggy were a non-starter for cars. We all just ate up that expense for the convenience of owning a car.

The problem here is that housekeeping labor is dirt cheap and robots are crazy expensive. What's the incentive here to do these things? If there's an industry that's dying for cost savings, it won't be the hotel industry. Now look at, say, unionized auto manufacturers or unionized public sector workers. Lots of trivial easy to automate jobs there. Why aren't city buses driving themselves? Or drones writing parking tickets? Or robots laying down asphalt? Or robots walking a beat? Or teaching a class? That's where I could see some advancement. A lot of municipalities are looking at their retiring baby boomer workforces who negotiated very generous pensions via a corrupt political process and now don't want to fall into that trap again.

Let us first focus on building printers that cannot get "paper-jammed" :)

Have you got any insights to share on your suggested timeframes, e.g. changing linens "still 5 years"? How did you come up with those figures?

You can guesstimate these things a bit by the Kurzweil type technique of projecting Moore's laws type improvements in computation and comparing them to animal neural systems as in one of his graphs:

http://cdn.singularityhub.com/wp-content/uploads/2015/01/exp...

I would take anything that Kurzweil says with a mountain of salt.

My timeframes are all minimum numbers - I.E. the dates by which I would give you 10 to 1 odds that they won't be beat. I don't have any insight into the maximum time frames - but anybody who spends a lot of time watching the growth curve in robotics would agree that we aren't advancing very quickly in optical/visual feedback dexterity See: http://newsoffice.mit.edu/2014/fingertip-sensor-gives-robot-... for what is considered to be state of the art.

The dates regarding generalized cleaning, come from having done a n analysis of the higher-level thinking that takes place when cleaning a room - its very, very impressive the level of analysis that needs to take place for anything but the most straightforward cleanup job.

What's interesting of course, is that anybody can make pretty good predictions as to what dates won't be beat, but it's hard to guess when they will be. Whats awesome to think about, is once we have technology to do things like clean up a room, the vast majority of other robotic tasks, are going to fall very quickly.

The first step would be to build a remotely-operated robot and have maids from cheaper places "telecommute".

Then I'm imagining the robots would slowly develop more semi-autonomy.

The challenge is to have the right VR on the other end so that the robot can be controlled intuitively.

I would extend that 10-15 years.

We can't even get toilet seat cleaning right (filth warning):

http://i.imgur.com/USTbZRl.gif

Are you really making 10 year predictions? Do you believe in yourself?

Yes - I am 95%+ confident that we will not have a generalized robotic cleaning system capable of doing what your average human hotel room cleaner can do today. I've been closely watching advances in robotics for 10+ years, as well as AGI advances. Neither of them have demonstrated any trends which would suggest that hotel room cleaning is even remotely possible in 10, or even 15 years.

In comparison, the predictions that I would have made about computing systems in 1995 about 2015, would have been almost bang on, with the exception that online teaching never really took off the way I thought it might, and libraries are still buying lots and lots of books.

Reply on news.ycombinator.com