Helping Robots Learn By Letting Them Fail
Originally published on hackaday.com by Richard Baguley on 17 September 2023
The MIT Technology Review recently unveiled its annual list of top innovators under the age of 35, featuring individuals who have achieved remarkable accomplishments at a young age. One standout innovator on this list is Lerrel Pinto, an associate professor of computer science at NY University. His research focuses on a novel approach to teaching robots how to perform household tasks through experimentation and learning from their failures.
Pinto’s innovative methodology involves creating a simulated home environment in his lab, complete with childlike toy versions of household appliances such as ovens. Further he then tasks a robot with performing everyday chores like warming up lunch. The robot is equipped with a machine learning model that learns from 30-second videos of Pinto’s students performing the same tasks using similar tools. The robot continuously attempts these tasks, learning through a technique known as reinforcement learning, where it adjusts its approach based on how closely it matches the desired outcome.
This approach has been documented in a paper titled “Teach a Robot to FISH,” emphasizing the idea that teaching a robot to learn tasks is akin to teaching someone to fish for a lifetime of knowledge. The paper provides valuable insights into breaking down and training robots for various tasks and offers an informative video presentation by Pinto to the CMU Robotics Initiative.
We at PiShop have always been interested in robotics so we collect kits for hobbyists or for educational use:
Lerrel Pinto’s groundbreaking work not only showcases his innovative thinking but also provides a promising avenue for the advancement of robotics in domestic settings. His approach to teaching robots to learn from their mistakes represents a significant contribution to the field of artificial intelligence and robotics, with the potential to revolutionize how robots interact with and assist humans in everyday life.
Read more: Helping Robots Learn By Letting Them Fail
Disclaimer: The content in this post includes excerpts from other blogs and websites. Full credit for the original work goes to the respective authors and publications. This content is shared here for informational purposes only, to give our customers direct insight on where to purchase the mentioned items, and all rights remain with the original creators.