Recently, multimedia applications such as Internet TV, peer-to-peer (P2P) multimedia streaming/broadcasting, video conferencing, and on-line gaming are proliferating over resource constrained network infrastructures such as the Internet, overlay networks, peer-to-peer networks, and wireless networks. In order to enable these various services to become truly ubiquitous and operate transparently, multimedia users need to simultaneously compete for the scarce resources of such networks.Existing resource management solutions have been designed and developed traditionally using centralized approaches, where a central controller (or a resource manager) decides the resource allocation among the participating users in such a way that the overall system utility (i.e. the system welfare) is maximized. Moreover, they also implicitly or explicitly assume that users are static, altruistic, and most often, homogeneous. However, in general, this centralized approach becomes infeasible as the number of users increases, because the amount of information that needs to be exchanged as well as the computational complexity required for finding the optimal allocations increase significantly. These solutions can also be undesirable from the perspective of self-interested and autonomous users, who may not comply with the specified rules in order to maximize their own utilities. Moreover, the solutions become inefficient for heterogeneous users, as the solutions do not consider users with different utilities, requirements or characteristics, including the users' bounded rationality. Also, since they are designed for static users, the solutions cannot efficiently adapt to dynamic changes in users' interactions, requirements, resource availability, and information availability.This dissertation addresses the abovementioned challenges by developing a distributed multi-user resource management framework: resource reciprocation strategies for cooperative users, as required for instance in P2P networks, and resource division strategies for multiple non-cooperative users competing for the same network resources. This enables multimedia users that dynamically and repeatedly interact with each other in a dynamically varying network environment to strategically maximize their utilities, or fairly negotiate their resource divisions, based on their heterogeneous processing abilities. The interactions of the users in the coalition for resource management are modeled as games based on the users' characteristics and the availability of resources. In the case where each user can obtain resources only by cooperatively associating with the other users and the associated users' reciprocal behaviors can only be statistically estimated by learning based on locally available information, we model the interactions of the users as a stochastic game. Based on this approach, each user can identify its foresighted strategy which can lead to a maximum expected long-term utility in its dynamic and repeated interactions. Note that the foresighted strategy of each user may result in different performances depending on each user's bounded rationality, i.e., information acquisition and processing capability. However, if available resources are limited and users are competing for the resources, then the participating users need to agree on a particular resource division. The resource negotiations among the participating users are modeled as bargaining problems. A solution to the bargaining problems enables the users to fairly and optimally determine their resource division, based on utilities. We extend and generalize the existing bargaining solutions by successfully deploying bargaining powers of each user, which are determined based on each user's multimedia characteristics, channel conditions, delay constraints, etc.The proposed framework enables individual users to achieve higher efficiency outcomes from both a user's perspective and a system's perspective, when users are strategically maximizing their own performance. This can be done by enabling devices to proactively and strategically interact with each other in order to maximize their own performance based on their asymmetric information, and ability to form beliefs and heterogeneous knowledge, rather than obliging them to passively comply with rigid, pre-determined, protocols as in current networks and communication systems.
展开▼
