Action prediction based on physically grounded object affordances in human-object interactions

摘要

Nowadays in human-robot interactions, robots must do reasoning beyond the present with predicting the future actions. This task requires the subtle details inherent in human movements that may imply a future action. In this paper, we employ a probabilistic method for action prediction in human-object interactions. The key idea of our approach is the description of the so-called object affordance, the concept which allows us to deliver a trajectory visualizing a possible future action. We experimentally validated the proposed method using two datasets.