Virtual Reality (VR) has been around for many years and most of us have, in some way or the other, been confronted with it, at least once. Since the introduction of this innovative form of human-computer interaction in the mid nineteen eighties, VR has gained a lot of interest and will continue to do so for many more years. 

Basically, VR is about creating Virtual Environments (VEs), which are believable, realistic, or at least visually attractive. However, this was not the case for the very rst VR applications. Over the last ten years, the technology has evolved considerably and fairly realistic systems can be built nowadays. Most of the research eorts have been put into improving the modelling of the static part of a VE, i.e. the visual appearance of objects and how to compose the overall scene. As a result, a number of software tools and application programming interfaces have been developed. These packages have eased the task of the designer mainly on the modelling of the static part of a VE. However, in essence VEs are really dynamic, i.e. objects have behaviours and users can interact with these objects. Unfortunately, the modelling of the dynamic part of a VE did not receive the same attention and therefore has not undergone the same progress as the static part. The specication of behaviour is mostly being done at a low level by manually programming it using either dedicated scripting languages or general purpose programming languages. As a consequence, the process of building behaviours is still not accessible to designers without programming skills. In addition, current VR applications are becoming more realistic and therefore require more complex behaviours. For this reason, there is a strong need to have new ways to make the modelling of behaviour easier and more intuitive. In other words, there is still no adequate answer to the question "How can the development process of VR applications, and more precisely the aspect of modelling the behaviour, be improved further?"

This dissertation presents an approach that is addressing the above question. The work is done in the context of VR-WISE, an existing modelling approach for Virtual Reality. The main goal of the VR-WISE approach is to ease and shorten the development of VEs by including an explicit conceptual modelling phase into the overall design process. It enables the designer to specify a VE through intuitive high-level specications without the need of considering any implementation details. Until now, VR-WISE only focused on modelling the static part (i.e. scene) of a VE. The purpose of this dissertation has been to extend VR-WISE in order to facilitate the modelling of the dynamic part (i.e. behaviour) of a VE. Firstly, this dissertation will introduce a behaviour modelling approach and an associated graphical Behaviour Modelling Language. This language allows designers to visually specify behaviour at a high level and without requiring advanced programming skills. Secondly, the dissertation will describe how the behaviour modelling approach has been extended to overcome some of the limitations associated with graphical notations when it comes to modelling complex behaviours. A Behavioural Design Patterns Framework has been developed to facilitate the specication of more complex behaviours by incorporating behavioural patterns into the graphical modelling language. This framework allows less experienced designers to use predened behavioural patterns while more experienced designers can create their own behavioural patterns. Thirdly, the dissertation will explain how this approach has been validated by implementing a proof-of-concept software tool supporting VE development using the VR-WISE approach. It will also report on an experiment performed showing how laymen have used this tool to model behaviours for real case scenarios. Furthermore, a number of other case scenarios have been elaborated and are presented in this dissertation.