A Geometric Perspective on Visual Imitation Learning
This addresses the problem of enabling robots to learn from visual demonstrations in a more autonomous and generalizable way, though it appears incremental as it builds on geometric vision and imitation learning concepts.
The paper tackles visual imitation learning without human supervision or interactive RL environments by proposing VGS-IL, a method that infers geometry-parameterized task concepts from demonstration videos, resulting in explainable and invariant representations that map to robot actions.
We consider the problem of visual imitation learning without human supervision (e.g. kinesthetic teaching or teleoperation), nor access to an interactive reinforcement learning (RL) training environment. We present a geometric perspective to derive solutions to this problem. Specifically, we propose VGS-IL (Visual Geometric Skill Imitation Learning), an end-to-end geometry-parameterized task concept inference method, to infer globally consistent geometric feature association rules from human demonstration video frames. We show that, instead of learning actions from image pixels, learning a geometry-parameterized task concept provides an explainable and invariant representation across demonstrator to imitator under various environmental settings. Moreover, such a task concept representation provides a direct link with geometric vision based controllers (e.g. visual servoing), allowing for efficient mapping of high-level task concepts to low-level robot actions.