Case-Based Skill Transfer in Robotic Agents

Abstract

Imitation learning is an effective method for interactively teaching a robot learner to complete a task.  We address the problem of transfer for robotic agents that learn tasks from demonstrations, where a robot is asked to adapt a learned task to be repeated in a related, but unfamiliar, environment.  We take a case-based approach to transfer, where the robot learner stores individual task demonstrations in a case memory such that they can be used at a later time for adaptation and reuse in new, target environments. We describe our ongoing work to enable transfer for robots that imitate task demonstrations.

Case-Based Skill Transfer in Robotic Agents

Posted in .