Stories

Where robot-makers go wrong

A quarter century ago, Colin Angle set out with two MIT colleagues with ambitions of putting walking robots in space. Today, Angle, CEO of iRobot, has settled for the world's best-selling robotic vacuum cleaner, the Roomba. It's not space, but, with about 16 million sold over the last 15 years, the Roomba is one of the best-selling consumer robots of any type ever.

Angle's message for commercial robot-makers: go short cool and complex; go long rugged, reliable and simple.

It's germane right now: Last Friday, Alphabet sold off Boston Dynamics, one of the world's most advanced robot-makers. The reasons: Boston Dynamics' walking, jumping robots were agile and cool, but contained no machine learning software, and, analysts say, had difficulty finding a commercial market.

That may or may not be true: Softbank, Boston Dynamics' purchaser, clearly thinks Alphabet is wrong about the company's commerciality. But, for all the hoopla over robots, not many have penetrated the consumer market yet. "Here we are, in 2017, and we're just about none of the way to achieving the potential that robots are going to play ultimately in our world," Angle told Axios at iRobot's headquarters in Bedford, Massachusetts.

Getting to vacuums was a journey: "My early days," Angle said, "was all building legged robots, and, you know, I think part of what — the reason why I ended up being so focused on practical robotics — is the transition I made from wanting to build Commander Data from Star Trek, or if not that at least legged, walking robots that could go explore other planets." He went on:

"For my master's thesis at MIT I built a robot so complicated that it just collapsed under its own crushing, like, complexity. You know, it had six legs. Each leg had four degrees of freedom. Each leg had its own microprocessor. There was hundreds of sensors on this robot, and it just _ I think the mean time between failures, was about seven minutes, and the promise of the legs was never realized."

That's where Angle learned his lesson: "Don't add anything to a robot design that doesn't need to be there. And so what we can we take out of the robot to make it more rugged, more reliable, even if it would be cool to add something."