Abstract
This thesis deals with the collisions detection problem within virtual environments, especially in the framework of interactive simulations handling complex scenes. Our research has been particularly directed towards the implementation of optimization techniques in this context. Our main contribution is a novel algorithm exploiting an evolutionary approach allowing to manage the high computational complexity related to the collision detection problematic. Baptized ESPIONS (Evolution Strategy on PIONS), this algorithm may be considered as a process aiming at identifying fields of local minimum distances between virtual objects of various natures. Moreover, it may be used jointly with classical determinist approaches either in serial or in a parallelized way. An experimental study has been developed to characterize the effect of ESPIONS parameters on its general behavior, in order to identify a set of relevant parameters allowing to improve the obtained results. Lastly, several implementations of the algorithm have been presented in order to illustrate the introduced concepts and procedures, and to discuss the contributions and limits of the proposed approach.