Réalité virtuelle et formation : conception d'environnements virtuels pédagogiques

Research in Virtual Environments (VE) for training has been essentially focused on technical aspects and doesn’t provideuser-centred approach of design (trainee and trainer). In order to design a powerful and useful training-oriented VE, wepropose a design methodology which handles different aspects. Indeed, in a first step, it is important to determine tutorialobjectives of trainers and sensori-motor abilities of trainees. On a next step, technological restrictions have to be taken inconsideration. Finally, the VE and useful functionalities have to be specified. Our design methodology is based on threefundamental elements in VE : sensori-motor immersion and interaction (i.e. physical interactions of users with the system),cognitive immersion and interaction (i.e. reasoning and behaviour of users in the VE) and functional immersion andinteraction (i.e. task learning and task transfer). Virtual Reality (VR) is mostly used as a method for the immersion of a user in an environment that reproduces the realitywith a high level of realism. VR can not exactly reproduce reality due to current technical limitations. Moreover, trainersoften need to use rather abstract representation to provide supportive and adaptive explanations about this reality. Thenumerous capabilities of VR allow to provide different realism levels. In order to manage this different realism levels and in order to optimise VR functionalities that support tutorial objectives andthe trainees’ learning process, we propose a Intelligent Tutorial Agent (HAL : Help Agent for Learning). HAL has been designed to provide the trainer with a powerful tool which allows to integrate his tutorial goals. In order toaddress this issue, we propose two types of adaptive tutorial strategies. The first one is responsible for the modification of thescenario (failures, climatic conditions). The second one provides help for understanding the situation (explain rules, show theconsequence of errors). These strategies can be represented by different realism levels of VE (e.g. enhancement, visualisationof hidden mechanisms, modelling of abstract concepts, etc.). HAL is based on a multi-agent system which fulfils modularityand genericity needs. An industrial application has allowed us to implement our design methodology, to exploit many VR potentialities and toshow the advantages of HAL in managing VR functionalities. The main goal of this application is to train rail drivers how tointervene on high-speed rail tracks.