Planification and graphical model for dynamic scenario generation in virtual environments.

Summary Our work is part of crisis management training in virtual environments. Scripting plays an essential role in human learning in virtual environments. It allows both to propose and orchestrate personalized learning situations and also to lead the learner towards relevant and formative scenarios. The work presented in this thesis focuses on the dynamic generation of scenarios and their execution in virtual environments. For this scripting, we aim at a set of objectives that are often contradictory: the user's freedom of action, the generation of varied scenarios that are faithful to the author's intention, the scripting control and the resilience of the scripting system. The different approaches to interactive storytelling favor more or less some of these objectives, but it is difficult to reconcile them all, and that is the challenge of our work. In addition to these objectives, we also seek to facilitate the modeling of the scenaristic content, which is still a real issue when it comes to scripting complex environments such as crisis management. We propose an emergent approach in which the scenario experienced by the learner will emerge from the interactions between the learner, the virtual characters and our scripting system MENTA. MENTA is in charge of scenario control by proposing a set of adjustments (on the simulation) that respond to scenario objectives chosen by the trainer (e.g., to work on certain skills in particular). These adjustments take the form of a prescribed scenario that is generated by MENTA via a planning engine that we have coupled with fuzzy cognitive maps through a FRAG macro-operator. A FRAG allows to model fragments of scenarios as a sequence of scripted actions/events. The originality of our approach relies on a strong coupling between planning and graphical models which allows to keep the exploration properties and generative power of a planning engine (which favors the variability and resilience of the system), while facilitating the modeling of the scenario content as well as the author's intention through scenario fragments which will be scripted by the author and reused in the planning. We have worked on a concrete example of scenarios concerning the management of a massive influx of injured people, then we have implemented MENTA and generated scenarios related to this example. Finally, we tested and analyzed the performance of our system.