Abstract
Wireless networks are typically characterized by user mobility and unreliable channels due to shadowing and multipath fading. As a result, a set of admitted users by connection admission control at arrival epochs may become inadmissible. In this paper we propose to classify users into clusters based on their geographical locations. Intuitively, users that are geographically close to each other are likely to produce strong interference to each other. We propose to form user groups based on users' cluster information and their signal-to-interference values. User groups are then scheduled in a round robin manner. We conduct simulations for both uniformly and non-uniformly distributed user locations. We show that our method obtain a feasible transmission scenario with much less iterations. In addition, for nonuniformly distributed users, our method produces solutions that are nearly optimal.