Vinsamlegast notið þetta auðkenni þegar þið vitnið til verksins eða tengið í það: http://hdl.handle.net/1946/32234
Optimizing the level of detail of a simulation involves avoiding the computation of unnecessary features, if doing so is transparent to one or more observers. This concept dates back to graphic level of detail which has already been thoroughly studied, and there are various techniques that adjust the rendering quality. Non trivial simulations, in addition to the graphic component, also have a model that is responsible for the behavior of intelligent agents in the environment. Modern computing allows simulations to be always larger and more complex, thus requiring an ever growing amount of resources to function. In this dissertation we propose a system that is able to decompose a simulation in different layers of abstraction, adjust the level of detail of the simulation according to the observer's perception of the world, and present the resulting visual representation to the user. In particular, my contribution to the system revolves around devising ways of understanding when to switch between different levels of detail, and how to perform the adjustment without altering the consistency of the simulation with regards to the user's perception of the environment.