nbfigueroa / auto-ds-learning Goto Github PK
View Code? Open in Web Editor NEWThis package includes code to segment trajectories of a multi-step demonstrated task. The tasks considered are those that can be defined as a sequence of goal-oriented sub-tasks. In this work, each sub-task is characterized by a a Dynamical System (DS) Motion Policy and an attractor (representing the policy goal) found inside each Action Proposition (AP) region. Segmentation points are inferred tracking state changes in the pre-defined AP regions of the task.