Four years ago, researchers at the University of California, Berkeley, uploaded a video to YouTube. It featured a demonstration they’d done using a powerful new robot called PR2, a dishwasher-size machine with two hefty arms and six camera eyes on its face. In the demo, PR2 stands before a disorderly pile of small towels. Then, slowly but surely, it stretches its arms, picks up a towel, and neatly folds it, even patting it gently to smooth out the wrinkles. The robot repeats the routine until no more towels are left in the heap.
The researchers were pleased with their work, but they didn’t quite expect what came next: Their video went viral. Within days, hundreds of thousands of people watched it as news of the robot spread through social media and the blogosphere. Reports popped up on newscasts and publications around the world. One Twitter user humorously summed up what the achievement might portend: “I, for one, welcome our towel-folding robot overlords.”
This robotic laundry experiment had obviously struck a nerve. The idea of robots doing chores around the house has long captured people’s imaginations. For some, robots would mean freedom from tasks they don’t have time for or don’t want to do. For others, robots would mean even more: They would help them live independently longer, providing care and perhaps even some degree of companionship.
It’s disappointing, then, that other than robotic toys and vacuum cleaners, robots are a rare sight in our homes today. And yet, here we are, still eagerly waiting for this technology to blossom. So, where are the robot servants?
Some recent developments suggest that they might not be too far away. Processors, sensors, and other components that robots need have gotten much better and cheaper, propelled by advances in smartphone technology. And open-source software is now available for robot simulation, control, vision, and many other functions. Also, rapid-prototyping tools like 3-D printers and laser cutters are lowering the barriers for designing and building new robots.
The result is that interest in robotics is exploding among researchers, hobbyists, and entrepreneurs. Robot-related start-ups and crowdfunded projects are proliferating. Venture capitalists have taken notice of this trend—and so have tech giants like Amazon.com and Google, both of which seem to have big plans for robots.
Now, where this is all leading is hard to tell. Some skeptics predict that in 20 or 30 years, personal domestic robots will still be just laboratory playthings. Others believe that in that same time frame intelligent machines will routinely be providing household help, advice, and company. Then, of course, there are the countless possibilities in between.
We editors at IEEE Spectrum don’t pretend to know how things will unfold, but of all the scenarios we’re able to envision, there’s one in particular we find not only possible but desirable. It’s one in which personal robots won’t be exclusive, expensive offerings few can afford. Rather, they’ll be as commonplace as computers or mobile phones are today, and their development will provide a global engine for innovation and prosperity.
So allow us to expand on our upbeat vision.
The domestic robots of tomorrow will be a varied bunch. Many will be small, single-task machines that’ll clean floors, water plants, feed pets, and such. They’ll hide most of the time, and we won’t pay much attention to them. Think of them as smart, mobile versions of today’s household appliances.
The robots that will really change things will be bigger and more powerful, and they will perform multiple tasks for us. They’ll have wheels or tracks, and some may sport legs. They’ll certainly have manipulator arms to handle objects. As with any piece of technology, you’ll be able to choose among many models with different styles, capabilities, and prices. But the key thing they’ll have in common is how they learn to do what they do.
These robots won’t do everything right out of the box. Instead, these general-purpose programmable platforms will start by doing just a few simple tasks, and they’ll continually acquire new capabilities. They’ll come equipped with attachment points for new accessories and, more important, standard interfaces that allow new third-party software to add functions to the robots’ repertoire that their manufacturers never imagined, much like the apps we install on our phones.
It’s a scenario that, in one form or another, some roboticists already have in sight. “While a computer program is pure software, some robot apps will be a combination of software and hardware, ” says Colin Angle, chief executive and cofounder of iRobot Corp., an industry leader based in Bedford, Mass. “These apps will include not only software but maybe also files that people can use in their 3-D printers to make parts at home. ”
What will the apps and accessories allow this new generation of home robots to do? Anything and everything, we hope. Some possibilities that immediately come to mind: Robots will learn to fetch things from the fridge, operate the microwave, clear the table after meals, load and unload the dishwasher, answer the door, pick up toys from the floor, clean toilets, and of course, do laundry. A free app will let your robot find all the socks in the bin; a paid version will find, pair, and place them in your drawer. The app may require that you 3-D print a new accessory—a sock-friendly gripper, say—or get it sent to your home. (Yes, people still receive packages in the future, perhaps delivered by Amazon drones.)
S.K. Gupta, a roboticist at the University of Maryland College, notes that cellphones were invented for people to talk, but we have found many new uses for them. “I believe that the same thing is likely to happen for home robots. Initially people will be interested in getting robots at home to help with basic household chores, but soon they will find new uses for these robots.”
For this to come about, companies making domestic robots will have to give up control in the name of openness. How soon that will happen is unclear, though. A big impediment, argues Ryan Calo, a professor of law at the University of Washington, is the possible legal liability such openness would engender. People have come to expect their personal computers to sometimes act a bit buggy with third-party software, and as a consequence lawsuits are rare. But if personal robots ever went haywire, it’s likely that their owners would sue the manufacturer for damages.
Calo thinks that some tweaks to the law could balance the danger of making robotic platforms open against the great promise of this approach. One reason to want to see things evolve this way is that creating a capable domestic robot will probably be too hard for a single company to figure out by itself. The robot makers that will prevail will be the ones that allow others to expand and enhance their creations. The same was true for personal-computer manufacturers decades ago, and this paradigm reigns today in the mobile world, with Apple’s iOS and Google’s Android running apps from all over.
So expect standard platforms to take hold in robotics, too. New apps and accessories for them will then gush from a vast number of companies and individuals—be it a kid in a garage in Silicon Valley, a group of programmers in Finland, or skilled engineers in India or Brazil. And as personal robots are opened up to the world, they’ll become the next transformative technology after the PC, the Internet, and the smartphone.
Our desired scenario will come to pass only if technologists solve some key challenges. To appreciate what they’re facing, it might help to understand what a state-of-the-art, general-purpose robotic platform looks like today. For that, consider our towel-folding robot, PR2.
Built by a team of robot wizards at a company called Willow Garage, in Menlo Park, Calif., PR2 is a mechanical marvel, possibly the most capable research robot ever built. Willow engineers have programmed the robot to grab drinks from a fridge, play pool, and plug itself in for recharging. Other research groups have taught it to bake cookies, prepare a traditional Bavarian breakfast, and even venture out of the lab to go buy sandwiches.
As impressive as these demonstrations are, PR2 also underscores the current limitations of robotics. The robot is awkwardly big. It can’t maneuver in tight spaces, and, weighing 220 kilograms, it could do some real damage if it crashed into something or someone. Another problem is that it does things slowly. In the Berkeley towel-folding demo, PR2 took 25 minutes to fold each towel (the researchers had to speed up their video dramatically to make it watchable). And then there’s the price tag: Each PR2 costs a whopping US $400 000.
But again, look at what happened in the computer industry: While mainframes sold for hundreds of thousands of dollars in the 1960s and ’70s, the first IBM PC, introduced in 1981, cost about $4000 in today’s dollars. So if we peer into the future, we can easily imagine a smaller, smarter, and cheaper successor to PR2 becoming a universal robotic platform that costs about as much as a typical car, perhaps less. What would it take to build such a machine?
Thanks to Moore’s Law, the computers and sensors used in robots are advancing at a formidable pace. High-resolution cameras, inertial measurement units, laser range finders, and 3-D depth sensors like Microsoft’s Kinect have profoundly changed robotics in the past half decade. One hardware component, however, has been holding robots back: their actuators. These are essentially electric motors connected to gears or other mechanisms. In a robot like PR2, they power its wheels and arms. The problem is that the actuators in robots today tend to be bulky and costly, and as a result those robots are big and expensive. (PR2’s arms alone accounted for about half of the robot’s hardware costs.)
Researchers are experimenting with alternatives, including bioinspired artificial muscles, cable-driven manipulators, and even soft, inflatable arms. Another possibility is building actuators that replace most metal parts with carbon fiber, plastics, and other lightweight materials.
“For me a big unknown is the ways in which new materials will change what we can put in robots, ” says Rodney Brooks, the celebrated roboticist and founder of Rethink Robotics in Boston. He’s hopeful, though, that materials scientists will indeed have a lot to offer robotics.
Reducing size and weight will be critical in making tomorrow’s personal robots nimbler, safer, and cheaper. But some say an even greater challenge will be software. “Like the early PCs, you need hardware of basic capabilities, but for the most part the industry did not really accelerate until you had great software, ” says Tandy Trower, founder of Hoaloha Robotics, a Seattle-based start-up developing robots to assist the elderly. He adds that the things that will pave the way to consumer personal robots are the same as they were for computers: software and applications that offer a “strong value proposition” and a great user experience.
The problem with the software today is that roboticists spend too much time reinventing the wheel, each creating different modules to handle control, communication, and other basic tasks in the robots. This, in part, is necessary because robots are still highly customized machines, and so reusing code is difficult or impossible.
But it’s hard to see personal robots flourishing until large numbers of different manufacturers’ machines are equipped with standardized components and interfaces. This, in turn, will make programming them—and sharing code—easier. And indeed, some robotics software is already moving in this direction.
Consider again Willow Garage’s PR2. Its software foundation is called the Robot Operating System, or ROS, which takes care of all the basic functionalities of the robot, such as making sure sensors and computers are talking to one another and much more. ROS is also integrated with a set of libraries that programmers can use to make the robot perform tasks that involve simulation, 3-D data processing, computer vision, and other things. That means they can develop complex applications—such as folding towels—without worrying about low-level details like how to acquire data from a sensor or keep track of where different parts of the robot are with respect to one another: ROS does all that for you.
The greatest virtue of ROS, which is free and open source, is that it can run not only on PR2 but also on pretty much any other robot with a computer. Indeed, the ROS community has grown to thousands of developers, and ROS is now running in everything from small humanoids to big industrial robots [watch the video below]. ROS, now maintained by the Open Source Robotics Foundation, is bringing us closer to a world where people don’t have to re-create basic software modules every time they build a new robot.
If all goes well, ROS or software like it will allow developers to focus on high-level applications rather than basic systems operations as they build ever more sophisticated capabilities into their robots. Those developers will then package their creations into apps, which users will download from robot app stores. Perhaps robots will make these purchases themselves. Or maybe newly created functionalities will be stored in a cloud-based knowledge repository, which robots would access to learn new skills. At the same time, robots would contribute to this repository on their own: If a robot learned to grasp a certain teacup, it would upload information describing how to do that in a way other robots would understand. That is, tomorrow’s robots would learn not just from people but also from one another.
And if we glimpse perhaps a few decades ahead, we see robots becoming more than just labor savers in our homes. As they become more common—and as we become more accepting of them—certain models will also serve as companions to many of us. You may find this idea uncomfortable, even repulsive. But the next generations, growing up alongside robots, will embrace them—or at least demand that they have more “personality ” than a disk-shaped vacuum bot.
Already filmmakers are beginning to speculate about our relationship with machines in the near future. In Robot & Frank (2012), a grumpy retiree is initially reluctant to accept the robot servant his son buys him, but he soon finds its presence indispensable. In Her (2013), a lonely writer played by Joaquin Phoenix becomes romantically attached to a talking operating system. “Like when I talk to her, ” he says, referring to the OS, “I feel like she’s with me.”
Ken Goldberg, a roboticist at UC Berkeley, doesn’t believe that ever smarter machines will replace human contact. “But people aren’t always available, ” he says. “So having a robot to tell me jokes, play videos and pictures of my past, and keep me company, I think that’s something we’re going to see. ”
To be sure, nobody knows whether robotics will evolve exactly this way. But we think it should. Roboticists will have to overcome many obstacles to get there, but a future in which robots become true helpers and at least surrogate friends is awfully appealing. Of course, they’ll fold our towels too.