Admission control in UMTS networks based on approximate dynamic programming

This paper presents a Connection Admission Control (CAC) algorithm for Universal Mobile Teleconimunications System (UMTS) networks based on an Approximate Dynamic Programming (ADP) approach. To deal with the non-stationary environment to the time-varying statistical characteristics of the offered traffic, the admission policy, has to be computed periodically based on on-line measurements. If standard algorithms are used, the optimal policy computation is excessively time-consuming to be performed on-line. Thus, an ADP approach for the computation of a sub-optimal admission policy is proposed. The ADP approach is based (i) on the reduction of the policy space, and (ii) on an approximated state-space aggregation. Theoretical results and numerical simulations show the effectiveness of the proposed approach, which is currently being implemented in a real UMTS testbed.