Abstract

In this paper we propose a probabilistic model for Human-Robot Spatial Interaction (HRSI) using a Qualitative Trajectory Calculus (QTC). In particular, we will build on previous work representing HRSI as a Markov chain of QTC states and evolve this to an approach using a Hidden Markov Model representation. Our model accounts for the invalidity of certain transitions within the QTC to reduce the complexity of the probabilistic model and to ensure state sequences in accordance to this representational framework. We show the appropriateness of our approach by using the probabilistic model to encode different HRSI behaviours observed in a human-robot interaction study and show how the models can be used to classify these behaviours reliably. Copyright © 2014, Association for the Advancement of Artificial Intelligence. All rights reserved.