Intelligent simulation for the estimation of the uplink outage probabilities in CDMA networks

In CDMA mobile networks, callers that are transmitting through a power station may cause interference at other power stations. When many users are already connected in the network, a new call may cause the signal to noise ratio to drop below a tolerance threshold. This phenomenon is called 'outage' and it provides an important measure of performance, useful in the design and control of the system. Evaluating this probability analytically has proven unsuccessful and only approximations exist today. Direct simulation of such networks is at present very slow because outage occurs infrequently - it may take hours to simulate directly a realistic model if a reasonable precision is desired. Thus this approach is not useful for design problems where one wishes to evaluate and compare performance of many different network models. In this work we implement a change of measure to estimate the outage probability using importance sampling. We present a functional estimator and a stochastic approximation method that are capable of learning the best parameters for the change of measure.