US Patent Application 18233251. SYSTEM(S) AND METHOD(S) OF USING IMITATION LEARNING IN TRAINING AND REFINING ROBOTIC CONTROL POLICIES simplified abstract

SYSTEM(S) AND METHOD(S) OF USING IMITATION LEARNING IN TRAINING AND REFINING ROBOTIC CONTROL POLICIES

Organization Name

GOOGLE LLC

Inventor(s)

Seyed Mohammad Khansari Zadeh of San Carlos CA (US)

Eric Jang of Cupertino CA (US)

Daniel Lam of Mountain View CA (US)

Daniel Kappler of San Francisco CA (US)

Matthew Bennice of San Jose CA (US)

Brent Austin of Munich (DE)

Yunfei Bai of Fremont CA (US)

Sergey Levine of Berkeley CA (US)

Alexander Irpan of Palo Alto CA (US)

Nicolas Sievers of El Cerrito CA (US)

Chelsea Finn of Berkeley CA (US)

SYSTEM(S) AND METHOD(S) OF USING IMITATION LEARNING IN TRAINING AND REFINING ROBOTIC CONTROL POLICIES - A simplified explanation of the abstract

This abstract first appeared for US patent application 18233251 titled 'SYSTEM(S) AND METHOD(S) OF USING IMITATION LEARNING IN TRAINING AND REFINING ROBOTIC CONTROL POLICIES

Simplified Explanation

- This patent application is about using imitation learning techniques to train and refine robotic control policies. - Initially, a robotic control policy is trained based on human demonstrations of various robotic tasks. - The robotic control policy is then refined based on human interventions while the robot is performing a task. - The robotic control policy can determine if the robot is likely to fail in performing a task and prompt a human to intervene. - A representation of the sequence of actions can be visually rendered to the human, allowing them to proactively intervene in the robot's task performance.

Original Abstract Submitted

Implementations described herein relate to training and refining robotic control policies using imitation learning techniques. A robotic control policy can be initially trained based on human demonstrations of various robotic tasks. Further, the robotic control policy can be refined based on human interventions while a robot is performing a robotic task. In some implementations, the robotic control policy may determine whether the robot will fail in performance of the robotic task, and prompt a human to intervene in performance of the robotic task. In additional or alternative implementations, a representation of the sequence of actions can be visually rendered for presentation to the human can proactively intervene in performance of the robotic task.