A robonaut performs chores for astronauts aboard the International Space Station. Google speeds ahead with self-driving cars. A robot has learned to cook from watching YouTube videos. Some are telling jokes and dancing for our entertainment; others are consoling us.
Robots are already moving out of the research lab, off the factory floor, and into our everyday lives. But how will those robots evolve—and how will we see them—in the next five years? I asked some of the humans who build them, leaders at the robotics companies Fast Company selected as the year's most innovative.
Robotics developer Hanns Tappeiner is frustrated by the evil killer robots trope that still dominates much of the discussion about robotics. No big surprise there: He's the president and cofounder of Anki Drive, whose dinky autonomous racing cars were Amazon’s second best-selling toy of the 2014 holiday season. Nevertheless, his tiny cars require much of the same sophisticated robotics technology as a full-size autonomous vehicle, from positioning and path planning to computer vision.
Much to the disappointment of science fiction writers everywhere—and contrary to the anxieties of Elon Musk and Stephen Hawking—Tappeiner insists that robots will not be taking over the world anytime soon. "Definitely not in the next five years," Tappeiner says. "Probably not in the next 50 years."
Even though robotics and automation in military research has helped to animate the specter of killer robots, Tappeiner argues that robots will continue to serve humans for the near future, largely because current AI techniques still fall far short of the capabilities of the human brain. While machine learning excels at specific tasks like translation (Google Translate, for example, uses a technique called statistical machine translation), that intelligence is not easy to generalize. (This hasn't quelled fears that robots and other forms of automation are coming for the jobs of routine decision-makers, like X-ray technicians, legal clerks, and news writers.)
"If you have a machine learning system that is good at translating, say, English to Spanish, you can’t use that same system to look up camera images and find scenery in an image," he says. The human brain, in contrast, performs all kinds of complex tasks, from understanding a language to driving a car. "There is absolutely nothing at all in technology which is able to do anything even remotely close," Tappeiner concludes.
None of that skepticism has stilled Google’s dream of orderly lanes of driverless cars. Still, that vision remains off in the distance due to safety and regulatory issues, as well as some technical ones. It’s hard enough to make a robot that performs a complex task like driving. It’s much harder to ensure that the robot never fails or makes mistakes.
Tappeiner calls this problem "the last 1 percent." "It might actually be the last 0.1%," he adds. "Making sure that that thing never ever makes mistakes, not even in 0.1% of the cases."
So some car manufacturers are approaching the issue of autonomy from a different angle. "The progress made by companies such as Google and Tesla, along with the efforts of the major auto manufacturers, will lead to a more 'hands-off' approach to driving," says Parag Batavia, founder and president of Neya Systems, which develops advanced unmanned-systems technologies.
Instead of being fully automated, the cars that actually hit the road in the next five years will help people drive more safely, efficiently, and with less effort. "Tesla’s approach to a slow rollout of increasingly capable auto-steering using over-the-air software updates," says Batavia, "will push state-of-the-art forward at a pace that allows regulators to keep up, while also continually improving performance."
Defense aerospace industry giant Lockheed Martin's most famous "robot" may be one of its most secretive: the RQ-170 Sentinel, a stealth drone that's been used by the U.S. over places like Pakistan and Iran since 2007. But the company also makes a range of other unmanned vehicles for less secretive missions in the air, at sea, and on land. Its Squad Mission Support System road vehicles tag along with troops in Afghanistan, while the K-MAX unmanned cargo helicopter supplies bases. Some unmanned systems are operated remotely. Others are more autonomous.
Jay McConville, Lockheed’s director of business development for Unmanned Integrated Solutions, expects different types of unmanned vehicles to start working together in missions well beyond the military, such as disaster relief, law enforcement, and media. "These different types of systems will start to proliferate to become more capable in concert with air, so that the use of unmanned systems together can accomplish things we really couldn’t do before," he says.
Lockheed recently demonstrated the use of two different UAVs (unmanned aerial vehicles) for fighting a fire in Australia. A five-pound Indago quadcopter spotted the fire, while the unmanned K-Max helicopter dropped water on the flames.
In an emergency evacuation demo scenario, K-Max delivered an unmanned ground vehicle to a remote location where someone had been injured. "All they had to do was climb on to the vehicle and make their way back to a pickup point," says McConville. "Or in the future, [the victim] might even be airlifted by the unmanned asset."
Autonomous robots like Boston Dynamics' mechanical menagerie and humanoid robots may also come to play more prominent roles in emergency situations as they become better at carrying heavy loads, conducting surveillance and fighting fires.
Bay Area startup Counsyl has delegated the tedious lab work of DNA-testing blood and saliva samples to custom-built robots, cutting the cost of genetic early-warning screenings in half. Kyle Lapham, Counsyl’s Director of Laboratory Automation, told me he expects a new generation of smarter, gentler robots that can work with humans on a variety of new tasks in the next five years.
"Traditionally, allowing humans to operate in close proximity to industrial robots has been too dangerous," he says. "With the advent of smart sensors and control software that can dynamically slow down robotic arms, avoid collisions, and sense contact with objects, humans and robots will be able to safely coexist in a shared workspace, opening up a number of new manufacturing environments."
Lapham predicts that, even when humans and robots don’t work together directly, autonomous indoor robots will start to share our environment. "Robots excel at the tedious chore of moving items from one point to another," he says. "Whether they are delivering prescriptions and meals to patients in hospitals, performing mobile security monitoring tasks, or moving silicon wafers between robotic workcells in billion-dollar semiconductor plants, autonomous robots will become a more familiar sight in our day-to-day lives."
Simple robots have already been used to teach programming concepts to children as young as five, but Lapham expects that everyone will be able to hack their own more serious robot in the next five years.
Traditionally, robotics hardware and control systems have been expensive, complex, and proprietary. Open-source software like the Robot Operating System coupled with cheap sensors, faster and more power efficient processors, and simpler design tools will make it much easier to build and hack robotic platforms. "A driven individual with an aptitude for engineering or science armed with the right tools will be able to do amazing things in robotics without the need for industry backing or millions in capital," he says. "Smart creatives will be able to bring their robot ideas to life easier than ever before."
Lockheed Martin’s Jay McConville sees many cases where it makes more sense to rent than to buy a robot. A farmer, for example, might use an unmanned aircraft to apply fertilizer or water only where it is needed, or to get real-time information on the status of his crops. "The farmer doesn’t need the aircraft," he says. "What they need is a service when they need it. Companies will step in to provide that service."
These robot service providers will need a range of different robots to serve various purposes. McConville thinks that will also drive demand for what he called common control: a single hardware and software architecture that can be used to control many types of unmanned vehicles. In the past, each vehicle had its own ground control system.
Lockheed Martin already supports the U.S. government's initiative to develop a common control system for multiple vehicles for military use. "We use STANAG 4586 for our aircraft," says McConville. "There are more standards developing for ground robotics and surface vehicles."
The development of a service provider market and lowering barriers to entry into the robotics market may lead to a robot gold rush in the next few years, says Parag Batavia of Neya Systems.
"Today we are in the Wild West days of VC investment in robotics. A lot of money is being put into companies that either do not have viable business plans or do not have defensible technology," he says. "That will change." Batavia believes that the large amount of money flowing into the robotics space will result in the creation of large companies that do not yet exist or are at a very early stage today.
Bill Culley, the president of Empire Robotics, agrees. "The next five years are going to see a continual stream of new companies popping up to solve the hard problems of industrial automation in new and unique ways," he says. But he urges caution: For new companies, developing bold, blue-sky ideas is the easy part. "What’s hard is gaining the knowledge and insights into what the industry really needs when you don’t have decades of knowledge and experience," he says. "And what’s even harder is to survive long enough to deliver a complex physical product with critically high demands for reliability, repeatability, and life cycle."
As a result, Culley predicts a flurry of partnerships between young and established companies in the next five years. The future of robotics isn't just about robots cooperating with each other and with humans—it will depend on old-fashioned human collaboration too.