Drives and Controls Magazine
Home
Menu
AI software will allow anyone to teach robots complex tasks
Published:  21 November, 2017

A Californian start-up company has raised $7m in seed funding to develop AI (artificial intelligence) software that will make it easy for anyone to teach robots new, complex skills. Embodied Intelligence – a spin-out from University of California Berkeley – has been set up by a team with 30 combined years of experience in artificial intelligence, deep learning and robotics.

Traditional programming of robots requires code to be written – a time-consuming task, even for experts. Embodied Intelligence’s software will allow anyone to program a robot simply by wearing a VR (virtual reality) headset and guiding the robot through the task. This demonstration will train deep neural nets, which will be further tuned by reinforcement learning, resulting in robots that can be easily taught a wide range of skills.

These skills could include complicated tasks, such as:

•  picking and ordering parts from cluttered, unstructured bins;

•  manipulating deformable objects such as wires, fabrics and food; and

•  completing assemblies where hard automation struggles because of the variability in parts, configurations, and individualisation of orders.

“Traditional robot programming requires substantial time and expertise,” explains Pieter Abbeel, president and chief scientist at Embodied Intelligence. “What we will provide is an AI layer that can be added to any existing robot, enabling robots to learn new skills rather than requiring explicit programming.”

Embodied Intelligence’s AI software will allow anyone to program a robot by wearing a VR headset and guiding the robot through the task

Abbeel’s lab at Berkeley has pioneered many breakthroughs in robot learning, including robots that learn vision-based manipulation through trial and error, robots that learn (simulated) locomotion, and a robot that organises laundry. He and two of Embodied Intelligence’s co-founders, Peter Chen and Rocky Duan, recently spent more than a year at the non-profit AI research company, OpenAI, working on learning technologies. They founded Embodied Intelligence, together with Tianhao Zhang, to bring modern deep learning and AI technology to robotics.

“Recent breakthroughs in AI have enabled robots to learn locomotion, develop manipulation skills from trial and error, and to learn from VR demonstrations,” says Sunil Dhaliwal, a general partner at Amplify Partners, which led the fundraising. “However, all of these advances have been in simulation or laboratory environments. The Embodied Intelligence team that led much of this work will now bring these cutting-edge AI and robotics advances into the real world.”

“Advanced robotic capabilities have been confined to established players that can afford costly R&D efforts,” adds Rocky Duan, Embodied Intelligence’s chief technology officer. “Our teachable robots will empower any size businesses to incorporate robotics into their manufacturing processes and keep up with the competition.”

The company’s CEO, Peter Chen, points out that VR devices provide an easy way to control and teach robots. “Since the robot simply mimics the hand motion that’s tracked by VR, a person without any special training can make the robot do the right thing right from the beginning. The robot will keep learning and, after a while, the robot says, ‘I got this, I can do this task on my own now.’”