Brain Corporation Builds BrainOS to Train and Democratize Robots
If you took apart a smartphone, you’d have many of the components needed to build a consumer robot: a low-power processor, software to run it, and an array of sensors. Qualcomm-backed Brain Corporation has turned the electronic guts of a phone into a foundation for making consumer robots, adding brain-inspired software to quickly program them.
The San Diego-based company plans to release its first product early next year—a suite of software tools, 3-D printable designs, and a hardware board with a Qualcomm processor. The development kit, called BrainOS, is designed to make it easier for third-party robot developers quickly build custom service robots for homes and businesses.
The problem with today’s robotics industry is that designers are forced to build a lot of the underlying systems that all robots need, says Todd Hylton, senior vice president of strategy at Brain Corporation. “Our feeling is that that’s not a good or scalable business situation. This industry really needs a horizontal technology provider, much the same way that the computing of mobile phone industry has companies and provide the core components,” he says.
Central to its plans is software that lets people program robots by showing them tasks. At the RoboBusiness conference last month in Boston, Brain Corporation showed off a demo of its “learn from demonstration” programming. A person with a remote control instructs a small, two-wheeled robot to, for example, run a figure 8 pattern around chairs. Once it was shown the pattern, the robot was able to perform the task alone and avoid obstacles.
Before joining Brain Corporation two years ago, Hylton worked at DARPA on a neuromorphic computation project to design chips that mimic the brain’s processing. And he says that the company has more algorithms, both neural-inspired and classic machine learning, under development designed to speed up robot programming.
Programming robots through demonstrations is not new. Rethink Robotics uses this technique for its manufacturing robot and academics have demonstrated the technique for many years. But Brain Corporation, which was founded by computational neuroscientist Eugene Izhikevich, thinks this technology is key to growing the robotics industry. A consumer could, for example, teach a robot to put things away or do specific tasks, such as vacuuming, Hylton says.
“We think there could be lots more apps, more sophisticated apps once the technologies we’re working on are broadly available. It shouldn’t be that hard,” he says.
By providing what it hopes will be a common hardware platform for consumer robots, Brain Corporation is essentially attempting to do what IBM did for personal computers in the early 1980s. Once a common platform was created, the market for software and standards-based PCs flourished.
In the case of robots, Hylton hopes that a couple of hardware developers could rapidly build prototypes without having to spend millions of dollars. Much of the lower-level computing and software infrastructure, which runs on Linux rather than Android, will be taken care of.
Historically, common platforms and standards have helped fuel innovation in digital technologies. With Brain Corporation, the key will be how easy it can make programming service robots and whether a population of robot developers will form around its developer kit.
In parallel, Qualcomm is developing a neuromorphic chip, which in theory could greatly improve the performance of service robots, such as their ability to navigate indoors. If that technology succeeds as hoped, Brain Corporation’s bet on neural-inspired software could pay off in a bigger way.
Trending on Xconomy
By posting a comment, you agree to our terms and conditions.