Contribution à la formation en réalité virtuelle : scénarios collaboratifs et intégration d'humains virtuels collaborant avec des utilisateurs réels

Industrial training is an emerging applicative domain for virtual reality. GVT (Generic Virtual Trai-ning) is a platform to create virtual environments for individual training on industrial procedures suchas maintenance procedures. Nevertheless, industrial needs keep on evolving and a new request concernstraining on collaborative procedures. In this thesis, we propose models to extend GVT possibilities totraining on collaborative procedures where real users and virtual humans collaborate.We present an activity model for the actors which allows the dynamic substitution of a real userby a virtual human. This model makes an actor perform actions depending on his characteristics, onthe scenario, on the environment and also on his partners’ activity. We also propose an extension tothe scenario language LORA in order to describe collaborative scenarios. Such a scenario describes theassignation of people to actions and integrates collaborative actions. The scenario has been simplifiedwhile making implicit some basic actions such as to take or to put back a tool. Finally, we presentthe action selection mechanism we have set up. Its aim is dual : to enable a virtual human to selectan action to perform and to give a trainee some pedagogical advice about the best action to choose.This mechanism has two parts : a global module of action distribution in charge of determining thebest candidate for each scenario action and one decisional module per actor. Indeed, an actor uses hiscollaborative profile (a set of behavioral rules) and the suggestion made by the action distribution moduleto choose the next action to perform.These models have been integrated to GVT in a prototype. In this thesis we present an applicativescenario which consists in collaboratively assembling a piece of furniture delivered in a kit.