Learning spatio-temporal characteristics of human motions through observation

The current work addresses the problem of learning the spatio-temporal characteristics of human motions through observation. Learned actions can be subsequently invoked in the context of complex Human-Robot Interaction scenarios. Unlike previous Learning from Demonstration (LfD) methods that cope only with the spatial features of an action, the formulated approach effectively encompasses spatial and temporal aspects. The latter are compactly depicted in a latent space representation of human motions. Learned actions are reproduced in the studied scenarios under the high-level control of a time-informed task planner. During the implementation of a given scenario, temporal and physical constraints may impose speed adaptations in the reproduced actions. The employed latent space representation readily supports such variations, giving rise to novel actions in the temporal domain. Experimental results demonstrate the effectiveness of the proposed formulation, as well as the proper execution of more involved scenarios.