Interactive Virtual Cinematography
- Publisher: IT-Universitetet i København
A virtual camera represents the point-of-view of the player through which she<br/>perceives the game world and gets feedback on her actions. Thus, the virtual<br/>camera plays a vital role in 3D computer games and aects player experience<br/>and enjoyability in games. Interactive virtual cinematography is the process of<br/>visualising the content of a virtual environment by positioning and animating<br/>the virtual camera in the context of interactive applications such as a computer<br/>game.<br/><br/>Camera placement and animation in games are usually directly controlled by<br/>the player or statically predened by designers. Direct control of the camera by<br/>the player increases the complexity of the interaction and reduces the designer's<br/>control on game storytelling. A completely designer-driven camera releases the<br/>player from the burden of controlling the point of view, but might generate<br/>undesired camera behaviours. Furthermore, if the content of the game is procedurally<br/>generated, the designer might not have the necessary information to<br/>dene a priori the camera positions and movements.<br/><br/>Automatic camera control aims to dene an abstraction layer that permits to<br/>control the camera using high-level and environment-independent rules. The<br/>camera controller should dynamically and eciently translate these rules into<br/>camera positions and movements before (or while) the player plays the game.<br/>Automatically controlling the camera in virtual 3D dynamic environments is an<br/>open research problem and a challenging task. From an optimisation perspective<br/>it is a relatively low dimensional problem (i.e. it has a minimum of 5 dimensions)<br/>but the complexity of the objective function evaluation combined with the<br/>strict time constraints make the problem computationally complex. Moreover,<br/>the multi-objective nature of the typical camera objective function, introduces<br/>problems such as constraints conflicts, over-constraining or under-constraining.<br/><br/>An hypothetical optimal automatic camera control system should provide the<br/>right tool to allow designers to place cameras eectively in dynamic and unpredictable<br/>environments. However, there is still a limit in this approach: to bridge the gap between automatic and manual cameras the camera objective should<br/>be influenced by the player. In our view, the camera control system should be<br/>able to learn camera preferences from the user and adapt the camera setting to<br/>improve the player experience. Therefore, we propose a new approach to automatic<br/>camera control that indirectly includes the player in the camera control<br/>loop.<br/><br/>To achieve this goal we have analysed the automatic camera control problem<br/>from a numerical optimization perspective and we have introduced a new optimization algorithm and camera control architecture able to generate real-time,<br/>smooth and well composed camera animations. Moreover, we have designed<br/>and tested an approach to model the player's camera preferences using machine<br/>learning techniques and to tailor the automatic camera behaviour to the player<br/>and her game-play style.<br/><br/>Experiments show that, the novel optimisation algorithm introduced successfully<br/>handles highly dynamic and multi-modal tness functions such as the ones<br/>typically involved in dynamic camera control. Moreover, when applied in a<br/>commercial-standard game, the proposed automatic camera control architecture<br/>shows to be able to accurately and smoothly control the camera. Finally,<br/>the results of a user survey, conducted to evaluate the suggested methodology for<br/>camera behaviour modelling and adaptation, shows that the resulting adaptive<br/>cinematographic experience is largely favoured by the players and it generates<br/>a positive impact on the game performance.