Markov decision process approach to optimal resource management in multiservice cellular networks

The field of cellular networks is undergoing a significant transformation in several dimensions. The main driving forces behind these changes are the demand for new services and Quality of Service (QOS) guarantees, which require sophisticated resource management schemes. The major design objective of resource management schemes is to build control structures that can guarantee the requisite QOS to each class while making efficient use of network resources. This dissertation addresses this optimal resource management issue arising in multiservice wireless networks. Given that each service may be characterized by different call arrival statistics and have a different QOS requirement, we consider Markov decision processes and related algorithms as a framework for optimal resource management problems. We demonstrate the effectiveness of that approach by applying it to a typical radio resource management problem arising in a multiservice cellular network. In particular, we consider an optimal asymmetric channel borrowing problem in two-zone reuse partitioning supporting two classes of services: narrowband and broadband. We show via simulation that there is a trade-off between increasing overall channel utilization, by allowing inner zone users to borrow free outer zone channels, and decreasing the blocking probability of outer zone calls. We formulate this problem as a constrained Markov decision problem, and show that the solution can be found using linear programming.