Open-World Skill Discovery from Unsegmented Demonstrations
This addresses the challenge of automating skill discovery from online videos for training agents in open-world environments, offering a practical solution for instruction-following agents.
The paper tackles the problem of segmenting long, unsegmented demonstration videos into skill-consistent segments without human annotation, using a self-supervised approach based on prediction errors from an action-prediction model. The method improved conditioned policy performance by up to 63.7% on atomic skill tasks and hierarchical agents by up to 20.8% on long-horizon tasks in Minecraft.
Learning skills in open-world environments is essential for developing agents capable of handling a variety of tasks by combining basic skills. Online demonstration videos are typically long but unsegmented, making them difficult to segment and label with skill identifiers. Unlike existing methods that rely on sequence sampling or human labeling, we have developed a self-supervised learning-based approach to segment these long videos into a series of semantic-aware and skill-consistent segments. Drawing inspiration from human cognitive event segmentation theory, we introduce Skill Boundary Detection (SBD), an annotation-free temporal video segmentation algorithm. SBD detects skill boundaries in a video by leveraging prediction errors from a pretrained unconditional action-prediction model. This approach is based on the assumption that a significant increase in prediction error indicates a shift in the skill being executed. We evaluated our method in Minecraft, a rich open-world simulator with extensive gameplay videos available online. Our SBD-generated segments improved the average performance of conditioned policies by 63.7% and 52.1% on short-term atomic skill tasks, and their corresponding hierarchical agents by 11.3% and 20.8% on long-horizon tasks. Our method can leverage the diverse YouTube videos to train instruction-following agents. The project page can be found in https://craftjarvis.github.io/SkillDiscovery.