PROBABILISTIC SHANNEL RESERVATION FOR MOBILE CELLULAR NETWORKS

Abstract

A Probabilistic Channel Reservation (PCR) scheme in which channel reservation is based on the probability of handoff from one cell to another has been proposed as a more flexible and dynamic means of providing handoff priority in mobile micro/picocellular networks in which frequent handoff events are to be expected. The handoff probability is derived from the movement of mobiles observed from earlier time intervals. The reserved channels are de-allocated when a call is either ended or handed off to another cell. In a ring network architecture environment the PCR scheme has been shown to outperform two benchmark channel reservation schemes, namely, the Trunk Reservation model (TR) and the Complete Sharing (CS) model. PCR is shown to have significantly improved the new call blocking probability and channel utilization over TR. This is at the expense of poorer handoff call dropping probability which, however, is still within reasonable levels. PCR also outperforms CS for both dropping probability and channel utilization, while the difference in blocking probability between the two is not significant. The PCR scheme has also been investigated under a different network architecture, that of a grid network to simulate a real-life urban environment. Three possible variations are compared. The first, FALL, involves reserving channels in all of a mobile's neighbouring cells according to the handoff probability (as for PCR above). This is compared with FBEST, where channels are only reserved in a mobile's most likely neighbour (derived from the handoff probability) with a probability of one. A third variation, FBESTPROB, where channels are reserved in a mobile's most likely neighbour according to the handoff probability has also been examined. It is seen that FBEST shows the best results for blocking probability and channel utilization. A further variation of PCR, the Dynamic Channel Reservation Scheme (DCR), has also been examined. DCR de-allocates reserved channels after a pre-assigned time and is shown to be well approximated by an analytical model. An optimization study has been carried out using the analytical model to obtain optimal values of N (the cell capacity for radio channels), Nr (the maximum number of reservation requests) and µ, (the service rate of the reservation request). The objective function is defined in terms of the revenue generated by the network. It is seen that, to accommodate higher network loads and tighter dropping probability constraints, Nr has to be adjusted to higher values without N having to be increased too greatly. This optimization study thus shows the advantage of having a flexible, adaptive value of N. It also allows us to obtain a thorough knowledge of the values of (N, Nr, µ,) to aid in the dimensioning of network parameters at different network loads.