“…Some approaches, e.g. [12], [13], [16], focus on the demonstrated trajectory that is analysed and decomposed, employing polynomial decomposition [12], Hidden Markov Models [16], Non-Negative Matrix Factorisation [13] and detection of critical points [17]. Other methods instead, particularly those based on reinforcement learning, assume the presence of an agent and a process of learning-by-doing [14], [10], [15].…”