When I saw iRobot's Ava at CES this January, I was pretty insanely excited. Just a month or so prior I had been prodding iRobot CEO Colin Angle about building a "real" robot, something that was multipurpose (not a vacuum cleaner), had true vision and mapping (didn't just bump into things to find out where it's going), and maybe even a few conversation skills. Colin played it pretty close to the chest: "do people really want that?" Well, Ava is his answer.
Ava is almost a "real" robot, according to the Standard American Paul Miller Definition (SAPMD) of the term. It has a speedy mobility base, with plenty of sensors (like sonar) to make sure it doesn't bump into anything, it has a PrimeSense 3D camera (the exact same one the Kinect uses) for getting a solid read on its surroundings, and it has a "head" in the form of your tablet computer of choice. Ava can extend its body from a three foot crawl to a eye-level-ish five feet high, making it a robot truly ready to interact with the people and animals in your home -- not just avoid them as it goes about its business. It's all good stuff, but there were still a few lingering doubts for me, keeping it from getting that full-on SAPMD stamp:
I think a single, somewhat articulated arm is an absolute minimum for a vital household robot. It's what elevates the bot beyond toy or novelty or Skypebot status. It's also potentially one of the most expensive parts of a robot, and difficult to do well. Still, I don't imagine a bunch of teleconferencing robots taking Best Buy by storm -- people hardly buy those $100 webcams to improve their chat experience, why would they buy a $3,000 robot?
The arm allows for all sorts of great little lifesavers, especially when you're away from home, that would make a robot and actually worthwhile addition to the household:
- Turn the lights off.
- Get the frozen chicken out of the freezer.
- Unlock your door if you forgot your keys.
- Lock your door if you forgot.
- Buzz you back in to your apartment (noticing a theme here for me?)
- Pinch out candles.
- Turn off the oven.
- Turn off a hair iron.
- Feed your pet.
- Bring you the remote.
- Bring you an adult beverage.
- Water the plants.
- Unload the dishwasher.
- Clean your room.
- Make your bed.
- Switch laundry loads.
Of course, no robot is going to be able to do many of those things autonomously, but just being physically capable is half the battle. The other half is...
The best and most important robots in the world, in my opinion, are bomb diffusement robots. They're doing something that no human would want to do, and they're protecting human life. But there's no "they," because bomb diffusement robots are completely remote controlled. Nobody pushes a button and says "go diffuse that bomb." Maybe one day, but for now these bots are painstakingly operated by a human.
On the flip side, the most adorable and useless robots in the world are completely autonomous. Furby, Pleo, Tickle Me Elmo. They're incredibly loveable, but they're less likely to do anything useful for the world than your cat.
So, on this continuum, I think a useful multipurpose household bot should be mostly remote controlled, but, since it's in the home, it needs to be a little lovable. Also, the remote control should be less like an Xbox 360 controller and more like an app, where you tap stuff to pick it up, or issue specific commands. A little bit of voice capability would go a long way as well. Robot interaction should be built around the concept of a computer addressing a human on the human's own terms, instead of the typical computer equation of a human addressing a computer on a computer's terms.
This is where it gets really tough. I think the price range of $2k to $4k is totally doable for a truly useful robot. But there's always going to be a delicate decision making process in robots: how much hardware do you have to add to make the bot capable, and how much hardware can you afford to make the bot still buyable? For instance, RoboDynamics just announced a bot called Luna, which will retail for $3,000 in its initial incarnation, but eventually go as low as $1,000. Only problem? Its arms are just metal rods. Great look, great price, great size even. But what can it actually do outside of carry a tray?
See, I've set up all these strawman problems up merely to say this: iRobot is on track to solve them all, and might actually succeed.
iRobot brought Ava to Google I/O last week, and on the surface it didn't look like much. Ava was rolling around and not bumping into people at CES with an iPad perched atop its tablet stand, and now Ava is rolling around and not bumping into people at Google I/O with a Xoom perched atop its tablet stand (video is above). But behind the scenes, there's some serious work being done that could truly cement the Ava as the first of a new generation of home robots:
First off, an iRobot rep has confirmed to me that the company is totally looking at robot arms, and sees them as absolutely vital. Right now they're trying to figure out if a home bot should have one arm or two, and how articulated that arm or those arms might be. We might even see something of an asymmetrical setup, where one arm does the complicated manipulation, and the other arm is available for bracing stuff up against and looking pretty. It still seems likely to me that we might see a stripped down, arm-free version of Ava sold as a teleconferencing solution, but it's a large comfort to know that a company capable of building a bomb diffusement robot is planning on bringing robo arms into the home.
As for personality, it looks like Google could lend a hand. It turns out Google is very interested in robotics, and has tasked some of its devs with building applications for Ava thanks to a new partnership between the companies. The great thing about this partnership is that Google's strengths are perfect for filling in iRobot's weaknesses -- and vice versa. iRobot is accustomed to delivering mission critical hardware and flawless (if feature limited) software on top of that. Google is used to throwing all sorts of beta functionality into its software projects and seeing what sticks, while it leaves hardware mostly up to partners.
Specifically, Google's voice recognition, speech capabilities, image recognition, and facial recognition features should provide some wonderful raw materials for devs to get started with. Those core technologies can be utilizied by Android devs -- both in house and out -- to create truly personable robots. One app that's already being tested on Ava allows you to remotely ask the bot to hunt down someone in a home and bug them until they answer the phone. Sure, it's a little creepy, but it's something completely impossible without this specific combination of iRobot hardware (wheels, navigation sensors), iRobot software (home mapping, path finding), Google software (facial recognition), and the actual Android tablet app that calls on all these disparate elements and allows the annoyed person you've tracked down a chance to interact with the proceedings.
Now, iRobot plans to have Ava be platform agnostic -- you'll still be able to perch an iPad or any other sort of tablet on top of the bot and start talking to iRobot's Aware Robot Intelligence Software middleware. Still, it's great to see Google get excited about robots and specifically invest in them -- and it might just ensure that Android becomes the platform of choice for folks looking to do interesting new things with Ava.
Finally, there's price. This was the scariest one with me, but iRobot seems to understand it as well as anybody. They see the price range being around $1,000 to $3,000, or "a good high end HDTV." Naturally, they'll have to strip the robot pretty bare to hit that price, so they'll be spending a lot of time figuring what's absolutely essential and what can be done away with in a home incarnation. iRobot is also looking at applications outside the home, such as hospitals, where the balance can potentially swing further in favor of capabilities and quality.
iRobot is also sure to emphasize the fact that it's not a "garage" company, and that it actually has experience building and shipping consumer hardware -- sorry RoboDynamics, but they've got a point. Still, a part of me is rooting for another Steve Wozniak to emerge in this not-yet-born market and to make a cheap, capable robot from off the shelf components, slap a copy of ROS (Willow Garage's open source Robot Operating System) on there, and let developers go wild. Likely? Maybe not. Romantic? Hell yes.
There are still plenty of unanswered questions in robotics, and I don't expect any answers for a while. How much of this stuff can be done better and cheaper by home automation? When will Ava be a shipping product instead of just a fancy demo? Do people really have room in their lives, hearts, or foorplans for a bulky and mostly incompetent robot? Is Ava really capable of the human emotion love?
I think one thing's for sure: it's going to be a pretty wild decade.