In a telecommunication system, the denial of service requests according to customer type or systems state, even when excess capacity exists, can improve system performance. Finding good call admission policies is therefore an important part of system design. The principal investigators (PIs) propose a probabilistic algorithm to search for call admission policies in telecommunications systems. Such a procedure decouples policy evaluation from policy search and in this way allows consideration of problems which seem intractable using analytic methods. The policies found are expected to yield insight which will allow useful analytic approximations to be employed. As a case in point, the PIs considered the M/M/c/c queue with two customer classes. The search procedure was able to find nearly optimal solutions whose structure strongly suggested the known optimal policy structure. It is proposed that this study be extended first to other simple systems which appear frequently at communications systems models and then to specific problems in the fields of mobile communications an high speed networks.
