There's not much new in home appliances these days, even if you count Android-enhanced fridges. Where there is innovation, it's in automation. What we're seeing is, quite simply, — this time, however, they're not here to destroy all humans: instead, they're here to do the housework.
We've spent much of this year's CES looking at home robots, and talking to thinkers at various robotics companies. The result has been some very interesting conversations, and a meditation on the future of the home robot.
Complex or simple?
There are two different types of home robot: some use complex software to map out environments before handling tasks, while others use emergent behaviours to turn simple actions into what seem like complex — almost intelligent — operations. It's a fundamental philosophical issue: do we make our robots complex and general purpose, or do we produce simpler, single-purpose devices? The former are often large and expensive, but can cope with a wide range of problems; the latter are simpler, often single purpose, and a lot cheaper.
So far the winners seem to be simpler, with iRobot's Roomba a clear market leader. It's a simple device at heart — a robotic vacuum cleaner that uses a mix of paths to clean an entire room. Set it up, let it go, and all you need to do is empty its captured dust. If you want to do more you can set a schedule, and maybe clear large items from the floor.
Although Roomba is simple, we as end users aren't. We tend to demand more functionality from our devices, and that's leading to a new set of philosophical questions for home robot manufacturers. Can I monitor my robot remotely, for example? Can I trigger it to operate when I suddenly realise I have visitors coming? Just how much connectivity do we need, and does it affect how an emergent behaviour is employed?
One customer demand is to see where Roomba is in a room. With the way Roomba works, using random paths to cover the floor surface of a room, there's no map — and so no way to show just where the robot is. Long exposures of a Roomba exploring a room show a decorative and highly random pattern of movements, with no two sessions the same.
What's more likely is simple monitoring, and simple triggering. That way Roomba's complex simplicity isn't changed, and what gets additional automation are the few physical interactions we have with a robot — pressing its start button and reading the warning lights. We get more information on our smartphones, and we're able to extend our reach remotely.
Roomba is a single-purpose device, as are the rest of iRobot's products. But they're only stepping stones on the road to a robotic future, where the end state is a flexible, general-purpose robot (much like Flexible Frank in Robert Heinlein's novel The Door Into Summer).
That road leads us into more questions about our robot companions, and how we develop them. Do we put more computational power into the machines, or do we take advantage of the growing power of cloud computing services?
If a robot is to be flexible and adaptive, then it needs considerable compute resources. It also needs to be able to share information with other robots: this is how to hold this tool, this is the action needed to do this task, for example.
Connectivity and the cloud make it possible to use cloud services to host a distributed robot intelligence — one that's shared across multiple chassis, and can use knowledge gleaned from one robot to improve the operation of another.
Today's robots are basic tools, with minimal intelligence. Tomorrow's will be able to take advantage of recent developments in natural user interfaces (like) and in cloud computing, to deliver a distributed flexibility based on shared context. It's a brave new world out there for our robot pals.