La gestion des ressources dans les reseaux multimedias sans fil mobiles (French and English text)

This thesis lies within the scope of resource management in wireless multimedia mobile networks. It starts with a critical survey of resource management issues in these networks. In particular, it reviews channel allocation, mobility management, and bandwidth distribution mechanisms.; Then, it proposes solutions to some resource management problems in wireless multimedia networks. More specifically, this thesis addresses the complementary issues of traffic characterization, bandwidth allocation, congestion control and call admission control.; The first addressed issue in this thesis deals with congestion control for applications requiring an end-to-end delay guarantee in wireless ATM networks. In this context, we propose a multi-agent system that allows reducing ATM cell loss ratio by dynamically redistributing the resources among ATM switches according to the network state. The agents manage the resources at the switch level and are able to reconfigure themselves in case of handoff.; The second investigated issue is the support of variable bit rate MPEG video streams at the base stations level. In this context, we analyze the statistical data sets of several MPEG encoded videos and propose a model of the elements of a scene. Based on our novel model, we propose a dynamic bandwidth allocation scheme that allocates bandwidth on the scene basis. Our algorithm is designed to operate with both pre-recorded and live streams. The proposed algorithm allows a high resource utilization while guaranteeing the required quality of service.; The third addressed issue relates to call admission control. To allow mobile multimedia networks continue providing the QoS required by multimedia mobile users, we designed a novel distributed call admission control algorithm. The proposed algorithm works in a distributed way by involving not only the cell that receives the call admission request but also a cluster of neighboring cells. Thanks to this algorithm, the network offers a low call dropping probability while maintaining a high network resource utilization.