A new process has been developed to train robots using “how to” videos, as opposed to laborious data entry.
Researchers at Cornell University said their new robotic framework, powered by artificial intelligence, allows robots to learn tasks by watching a single how-to video.
Doctoral student Kushal Kedia said RHyME (Retrieval for Hybrid Imitation under Mismatched Execution) could fast-track the development and deployment of robotic systems by significantly reducing the time, energy and money needed to train them.
Mr Kedia said robots could be finicky learners and often required precise, step-by-step directions to complete basic tasks.
“One of the annoying things about working with robots is collecting so much data on the robot doing different tasks,” he said.
“That’s not how humans do tasks. We look at other people as inspiration.”
Mr Kedia said home robot assistants were still a long way off because they lacked the wits to navigate the physical world and its countless contingencies.
He said to get robots up to speed, researchers were training them with videos of human demonstrations of various tasks in a lab setting.
“The hope with this approach is that robots will learn a sequence of tasks faster and be able to adapt to real-world environments.”
Assistant Professor Sanjiban Choudhury said their work was like translating French to English.
“We’re translating any given task from human to robot,” Assistant Professor Choudhury said.
He said RHyME paved the way for robots to learn multiple-step sequences while significantly lowering the amount of robot data needed for training.
“This work is a departure from how robots are programmed today.”
Read the full paper: RHyME enables one-shot imitation from human videos with mismatches in embodiment and execution.
