SKID RAW: Skill Discovery from Raw Trajectories

TL;DR

This paper introduces SKID RAW, a Bayesian method for robots to learn and segment skills from raw, unlabelled trajectory data, enabling natural task teaching and adaptive behavior in complex environments.

Contribution

It presents a novel Bayesian approach that simultaneously segments trajectories and learns skill representations from unlabelled demonstrations, enhancing robot learning capabilities.

Findings

Successful segmentation of trajectories into skills

Effective learning from unlabelled data

Applicable to synthetic and real demonstrations

Abstract

Integrating robots in complex everyday environments requires a multitude of problems to be solved. One crucial feature among those is to equip robots with a mechanism for teaching them a new task in an easy and natural way. When teaching tasks that involve sequences of different skills, with varying order and number of these skills, it is desirable to only demonstrate full task executions instead of all individual skills. For this purpose, we propose a novel approach that simultaneously learns to segment trajectories into reoccurring patterns and the skills to reconstruct these patterns from unlabelled demonstrations without further supervision. Moreover, the approach learns a skill conditioning that can be used to understand possible sequences of skills, a practical mechanism to be used in, for example, human-robot-interactions for a more intelligent and adaptive robot behaviour. The Bayesian and variational inference based approach is evaluated on synthetic and real human demonstrations with varying complexities and dimensionality, showing the successful learning of segmentations and skill libraries from unlabelled data.