Active Learning Of Probabilistic Movement Primitives