With very little addition, something like this could be used to allow people to control machinery remotely. I'm thinking of commercial janitors (vacuum cleaners could be remotely controlled) and lawn mowers. Both of these activities are mainly just walking or riding around guiding a noisy piece of equipment which could be stopped if something went wrong. Also, machines could be semi-autonomous: controllable by a human when one is available, and able to do useful work by itself when not being human controlled.
You probably wouldn't want to drive a car or fly a plane with this yet, as a network glitch could result in disaster, but there are plenty of activities where a shut-down can be done safely.
Just saying, lawn mowers are not exactly the safest things in the world. I'm working in field of robotics (more outdoors than anybots) and I wouldn't really want to see that yet. Robots are great at detecting obstacles to avoid with Lidar, they are not great at handling edge case scenarios. With our bot even in "autonomous" mode somebody is watching, not sure if I want to add a spinning blade to it.
All of the military bots are remote controlled, not autonomous. I can't think of any major autonomous vehicle being used for fluid tasks (i.e., not inventory robots). So what, you are going to have someone sitting there watching 10 robots mow lawns remotely? How much cheaper can that be than just hiring people to mow lawns.
I don't think robots are very good at replacing people in low wage jobs. A lot of uniformed people think that makes a lot of sense because robots replaced factory workers, but robots did not replace factory workers because they were cheaper. Robots replaced factory workers because they were better. They were faster and more accurate. A paint job done by a robot is better than one done by a human. I worked on a project automating the assembly of a medical device - it was just too damn small to be done by a person reliably. The system cost almost a million dollars, you could have paid the girl it displaced for years and years and years before you pay it off - except that the liability of making a medical device incorrectly is astronomical.
So I do not target low wage jobs as a prime market for robots, there is no money in it. Robots are good for replacing jobs that are the dull, dangerous, dirty, impossible for humans to do, repetitious etc.
The work done to build a robot is amortized over all the robots built. So I think it will be cheaper. Factory robots do replace workers because they are cheaper. Humans can do high quality paint jobs too. It just probably takes them longer, or maybe humans need multiple tries because they mess up more often. The point is, no factory owner will invest in new technology if it didn't make economic sense. Quality is not orthogonal to price.
A mower robot could also be safer. A person riding a mower is exposed to a lot of noise, sun, and shaking. If you are relatively disabled or have a bad back, you can't operate a mower. Just about anyone who can see, on the other hand, could operate a mower remotely. No exposure to noise, to dust, to sun, to bumps. Also, you would want to err on the side of safety, so if the robot is in any way unsure of itself (say because there are moving objects nearby) it should just shut off and wait for human guidance. Some jobs, like mowing, cleaning, planting, harvesting, etc, can easily be stopped with no problems. On the other hand, you wouldn't want to just shut down a car in traffic when something goes wrong.
If people can attend meetings in another country remotely, why couldn't they also push around a vacuum cleaner or pick strawberries remotely? This would avoid all the costs of travel and food and boarding in a foreign country. Rather than immigrants working illegally in the US, they could be operating machines in the US from their home countries.
Janitors already travel from building to building to clean the floors. If they could operate the cleaning equipment from home, why wouldn't they? Physically transporting a person is more expensive than transporting information, and the gap between the two will only widen.
It seems to me that the hard part is remote controlling the robot so that it can move around while not hurting itself and other people. The Anybot robot is already semi-autonomous in this sense. Once you have this basic feature in place, it's straightforward to add other capabilities.
To answer your question (the one without a question mark), much cheaper, since people in poorer countries are willing to do same jobs for much less. IIRC that's the master plan behind Anybots.
That was the first model...the second model has a much lower center of gravity (too bad the 45 yo. danish man couldn't wait till the region got the second version) but there should be a sensor if more than 2 wheels are off the ground for more than 1 second the blades disengage.
Yes, I think there's a huge opportunity to automate the easy 80% of many jobs without requiring general AI. You just bring the people back on line when it gets stuck.
$15K is on the pricey side, but for some industrial inspection tasks I can imagine this being a cost which can be justified. For just meeting folks in the office I'm not sure that many businesses are going to go for this. For casual use the cost is going to need to be lower, and from a hardware point of view there's no reason why the cost can't go much lower.
My biggest beef with this robot though is its mechanical design. There's no good reason that I can see why it has to be balancing on two wheels. As a technology the balancing system is fine, but it doesn't appear to be fail safe. If the robot suffers a failure, the most common of which will be battery failure, it should fail in a safe position. With a balancing robot if there is a failure it's just going to fall over and crash into stuff, generating a hazard and liability issues. Depending on the environment in which the robot is deployed, uncontrolled toppling could have very serious consequences (for example, starting a fire).
Two wheel balancing does have some advantages: We wanted the robot to be human height and have a small footprint so that it could navigate nicely through doors and tight spaces. To keep a tall, small based robot from falling over you have two choices:
1) You can put a whole bunch of weight in the base. (This is the route that Willow Garage seems to be going. They have a huge lead acid battery in the base. The advantage of this is that it is static and 'always on'. The disadvantage is that your robot has to be really heavy.
2) You can be dynamically stable. With this approach when you are bumped you use your drive system to actively deal with the disturbance. The advantage of this is you can deal with much larger disturbances (http://www.youtube.com/watch?v=mkzgglYjNvY) The disadvantage is that it only works when there is power.
I totally agree on the battery failure scenario and it is something we have thought about a lot. We have a separate board that just does the balancing and it has a separate Ultra-capacitor power supply that can power the drive system for several minutes; long enough to put our kickstand down and sit. You can crash the main CPU and pull the batteries out and the robot will sit safely.
I don't know if self balancing is for sure better than statically stable, but I think it is less clear cut than what you imply.
Having a kickstand or third wheel deployed when battery voltage gets low, or perhaps just when the robot has been static for a while to save on power, does make sense and if this is a separate piece of electronics then that's probably good enough as a safety system to prevent toppling.
From a safety point of view as long as you've identified the possible hazards and done something which mitigates against them to some reasonable extent then that's probably good enough.
I was able to play with it the other day. I couldn't believe how light it was. Trevor ran over my foot, and instead of saying "ow" I said "ohh, pardon me" like it was a real person there. That's probably because the 2-way audio makes it seem like the driver really is there. It has the presence of a person because of the height, but the inherent safety of something like a roomba because of the weight.
QA had two big problems. First, it looked too much like a sci-fi robot and people wanted to interact with it rather than the guy controlling it. Second, it was heavy (65 lbs) and fragile. After a couple of smashups we found ourselves just picking up the phone instead of using it.
We redesigned QB to look more abstract, so it doesn't seem like a 3rd party in the conversation, and it's light (35 lbs) and padded so we just drive it full tilt all over the office without worrying about the robot or furniture.
We originally thought the main use case would be avoiding trips overseas, but we find ourselves using it to ask questions of people in the next office. Email and IM went through the same trend, so I guess that shouldn't be surprising.
No. But robots are replaceable -- people aren't. We think one of the most valuable services a robot can perform is to get blown up instead of its operator.
Good question. When the sound comes out of one robot's speaker and into the other's microphone, it sounds a bit tinny though intelligible enough. One of our planned hacks is to shortcut the audio inside the cloud, but it's tricky to not end up with extra echoes.
Is that because when you're routing the audio directly from robot to robot, that it's technically difficult to cancel the extra transmission that is being sent from robot speaker to robot microphone?
$15k seems a little steep for what looks like a smaller version of a Segway that doesn't need to carry a 200lbs human. You'd think there would be more startups in this space, as there's definitely a market and this is the stuff of the future.
The design is by Scott Wiley. If you've been to the Anybots / Y Combinator offices, the oil paintings on the walls are all his. He's also a mechanical engineer who has designed space station lasers &c.
Modern 3D cad software & stereolithography are amazing. You can get a student edition of SolidWorks, upload the 3D files to stereolithography.com, and get parts back in a week. Cost for smallish parts is tens to hundreds of dollars.
It runs FreeBSD 8. We use SSL between the robot & cloud, but user logins to the web interface use passwords over https. I'd like to support something stronger. Can someone point to a website with truly secure logins I can poke at?
https is fine - just couple it with PKI enabled smart cards (these work generally by providing a client-side certificate for your TLS handshake).
This is the strategy the DoD is using for most/all web auth stuff now.
Choosing a good password and using some other basic precautions like HTTPS will deter nearly all script kiddies and most everyone else. When you don't look like an easy target, your actual security risks are much lower than a lot of other people's because almost no one will care enough to try to go any further.
Not at all. These robots, if they become popular, will be a fat fat target for all kind of hackers. All it takes is to add the web interface URL pattern to the popular key loggers. If there are many ppl who use the robots, some will have their PCs rooted already.
this thing is adorable. i smile every time i tab past the IEEE article in my browser. now you've just gotta get Beyonce to sing about it (a la "video phone") and you'll be in business.
You probably wouldn't want to drive a car or fly a plane with this yet, as a network glitch could result in disaster, but there are plenty of activities where a shut-down can be done safely.