A behaviour sequencing and composition architecture based on ontologies for entertainment humanoid robots

Research in Human-Robot interaction stresses the importance of non verbal behaviours in order to enforce credibility and believability in robots. When using small and cheap entertainment robots, verbal cues are very limited, gaze and facial gestures can be absent, whereas body gestures comprise – as a matter of fact – only stereotyped behaviours. The main contribution of this article is an evaluation study about the effects of pure body gestures when entertainment robots are observed by humans in an interaction process. Robots are required to execute a sequence of behaviours, as if they were playing on a stage. To this aim, a model for composing behaviour primitives for entertainment humanoid robots using AND/OR graphs is presented. Both single and multi robot scenarios are considered. The framework can sequence robot behaviours using complex scripts, which can be easily programmed using standard ontology-based knowledge representation mechanisms. Procedures are used to operate on the ontology representation in order to execute behaviours in a fully hierarchical and recursive fashion. Experimental validation with an interview campaign has been carried out using toy Kondo robots.