Learning and generalization of complex tasks from unstructured demonstrations

摘要: We present a novel method for segmenting demonstrations, recognizing repeated skills, and generalizing complex tasks from unstructured demonstrations. This combines many of the advantages recent automatic segmentation methods learning demonstration into single principled, integrated framework. Specifically, we use Beta Process Autoregressive Hidden Markov Model Dynamic Movement Primitives to learn generalize multi-step task on PR2 mobile manipulator demonstrate potential our framework large library skills over time.