Predictive call admission control for all-IP wireless and mobile networks

This paper proposes a novel call admission control (CAC) scheme for wireless and mobile networks. Our proposal avoids per-user reservation signaling overhead and takes into account the expected bandwidth to be used by calls handed off from neighboring cells based only on local information stored into the current cell where user is seeking admission. To this end, we propose the use of two time series-based models for predicting handoff load: the Trigg and Leach (TL), which is an adaptive exponential smoothing technique, and ARIMA (Autoregressive Integrated Moving Average) that uses the Box & Jenkins methodology. These methods are executed locally by each base station or access router and forecast how much bandwidth should be reserved on a periodic time window basis. The two prediction methods are compared through simulations in terms of new call blocking probability and handoff dropping probability. Despite the TL method simplicity, it can achieve similar levels of call blocking probability and handoff dropping probability than those of the computational demanding ARIMA models. In addition, depending on the schemes settings, the prediction methods can grant an upper bound on handoff dropping probability even under very high load scenarios.