Many approaches for designing robotic applications have been proposed in recent years. If they are relevant to the problem they solve, most of them are platform- or language-dependent. To solve this problem, the use of formal models for scenario or activity generation is essential to ensure a high-quality experience. In this context, the use of Petri nets or automata is widespread. Based on this observation, we propose the formal CIT (content interaction time) model [37], which is dedicated to the development of interactive robotic applications and based on networks of input/output timed automata. In this paper, we propose the use of the CELTIC (Common Editor for Location Time Interaction and Content) and EDAIN (Execution Driver based on Artificial INtelligence) software platforms to allow for a simplified and generic modelling of robotic applications and their supervision based on the CIT model. We used this approach to design a serious game with the humanoid robot Nao. This game allows young people to playfully discover the ethnography exhibition at the Natural History Museum of La Rochelle, France, and the archaeology exhibition at the Sainte Croix Museum of Poitiers, France. After these experiments, we identified problems regarding dynamic adaptations of the scenario (robot listening time, speech recognition threshold...). The end of this article presents the beginning of our reflections on the integration of a reinforcement learning algorithm that optimises the parameters of a scenario’s execution.
