AI assist will help enable human-scale robotic systems for USAF

News

October 06, 2020

Under the terms of the contract, the project will base the platform on the upper body of the existing Guardian XO from Sarcos, an existing wearable exoskeleton robot that can be taught to perform dextrous tasks through positive reinforcement and imitation machine learning (ML) technologies, which the company calls Cybernetic Training for Autonomous Robots (CYTAR).

According to information from the company, the Sarcos AI system -- in contrast with other AI platforms that use a trial-and-error approach -- calls for human operators to teach the robotic systems to perform tasks correctly the first time. The Sarcos approach to training AI, say company officials, will significantly accelerate the speed and reduce the cost of deploying robotic systems that can perform meaningful work in unstructured environments.

“Similar to our Guardian GT robot, the upper body of the Guardian XO can be tele-operated to perform intricate tasks that require human-like dexterity such as welding, grinding, riveting, and complex assembly tasks,” said Ben Wolff, chairman and CEO, Sarcos Robotics. “By substituting the legs of the Guardian XO with other types of mobile base form factors, including wheeled, tracked, or telescoping platforms, these systems can be trained and supervised to perform dangerous and difficult tasks in places where humans can’t or shouldn’t go. The development of our CYTAR AI platform will deliver an intuitive human-machine interface that leverages human dexterity, instinct and reflexes to teach machines to perform complex tasks that will fundamentally change the way AI creates value in the real world.”