Traffic-aware optimization of heterogeneous access management

This paper studies the problem of admission control and air interface (AI) selection in heterogeneous network environments. The statistics of random user arrivals, channel conditions and service durations are considered for the optimization of heterogeneous access management strategies with respect to minimizing the expected mean cost for blocking events. Based on state aggregation in a semi-Markov decision process formulation, an efficient approximation algorithm using state aggregation for policy optimization is proposed. Though this solution is suboptimal, it still offers considerable performance gains in comparison to simpler heuristic strategies, which is demonstrated by simulations in a heterogeneous GSM-EDGE (Global System for Mobile Communications Enhanced Data Rate)/UMTS (Universal Mobile Telecommunications System) scenario. Furthermore, structural properties of optimal user assignment policies are studied, proving certain monotonicity properties for a specific type of systems.