Communication of spatial expressions on multi-agent systems using the qualitative Ego-Sphere

Abstract

© 2016 IEEE.The need for spatial representations and spatial reasoning is omnipresent in various real world applications of autonomous systems. The task of the qualitative spatial reasoning sub-field of Artificial Intelligence is to provide formalisms allowing a machine to represent and make inferences about spatial entities. In this work we make use of one such formalism for representing qualitative location, named Qualitative Ego-Sphere (QES), that discretises the world around a visual agent into sectors, as well as with respect to the relative distance of objects from the observer's point of view. QES was used in this paper as a means for communicating spatial expressions between pairs of agents. Four situations were proposed and implemented in order to address interactions between pairs of artificial agents and between an artificial agent and a human. Tests with human volunteers suggested that the human description of space in sectors agrees with the qualitative discretisation provided by QES. However, no similar agreement rates were obtained regarding the description of space related to the distance between objects. This was arguably due to the fact that qualitative distance judgements imply the existence of some relative external reference frame (not taken into account in the Qualitative Ego-Sphere formalism).