Improving call admission policies in wireless networks

It is well known that the call admission policy can have a big impact on the performance of a wireless network. However, the nonlinear dependence of new calls and handoff calls makes the search for a better call admission policy - in terms of effective utilization - a difficult task. Many studies on optimal policies have not taken the correct dependence into consideration. As a result, the reported gains in those studies cannot be confirmed in a real network. In this paper we develop a solution to the problem of finding better call admission policies. The technique consists of three components. First, we search for the policy in an approximate reduced-complexity model. Second, we modify the Linear Programming technique for the inherently nonlinear policy-search problem. Third, we verify the performance of the found policy in the exact, high-complexity, analytical model. The results shown in the paper clearly demonstrate the effectiveness of the proposed technique.