Could a Robot Learn By Itself?

Ashutosh Saxena

Photo Source Cornell Engineering

“We now live in a world where robots are helping humans in their daily lives, and just like humans, robots need to learn new skills in order to do their jobs successfully. And we shouldn’t expect a robot to learn on its own from scratch, any more than we’d expect a human to do so—imagine a child growing up with no access to textbooks, libraries, or the Internet.” – Ashutosh Saxena

What if robots could learn to carry out tasks autonomously? In other words, when giving a robot a new task it could “figure out” on it’s own how to do it? This is exactly the question that Ashutosh Saxena is working to answer.

Professor Saxena is a roboticist at Cornell University working to develop a massive online search engine that robots could access and find the required knowledge to carry out tasks, the aptly named RoboBrain. When given a question, RoboBrain will search the internet for relevant knowledge databases and images, sidestepping the need to teach robots to do tasks through step by step instructions.

“In 2014, I started a project called RoboBrain at Cornell University along with PhD students Ashesh Jain and Ozan Sener. We now have collaborators at Stanford and Brown. What we’re working on is a way of sharing information that allows robots to gather whatever knowledge they need for a task,” writes Saxena. “If one robot learns, then the knowledge is propagated to all the robots. RoboBrain achieves this by gathering the knowledge from a variety of sources. The system stores multiple kinds of information, including symbols, natural language, visual or shape features, haptic properties, and motions.”

The implications of such a project, if successful, could be enormous. It would lead to an increase in efficiency and reduce downtime and spending spent in “training” robots how to carry out tasks. RoboBrain could also lead to robots with more capacity to carry out objectives than previously intended.

Saxena’s work has garnered him several awards and recognition, including Eight Innovators to Watch in 2015, Smithsonian Institution; World Technology Award, 2015; The 50-years of Shakey at AAAI-RSS Blue Sky Ideas award, 2015; RSS Early Career Award, 2014; NSF Career award, 2013; Microsoft Faculty Fellow, 2012; Alfred P. Sloan Fellow, 2011; Best Cognitive Robotics paper, IROS’14. Best student paper, RSS’13; CUAir at AUVSI’12: First prize, mission performance; Google Faculty Research Award, 2012.

These breakthroughs in robotic learning coupled with the unparalleled potential of a developed RoboBrain is what makes Ashush Saxena our pick for “Featured Innovator of the Month.”

Note: Ashush Saxena does not work for Willamette Valley Company nor is he affiliated with our company.


Stanford University

Rethink Robotics Youtube

MIT Technology Review