A Learning Automata Based Dynamic Guard Channel Scheme

Dropping probability of handoff calls and blocking probability of new calls are two important QoS measures for cellular networks. Call admission policies, such as fractional guard channel and uniform fractional guard channel policies are used to maintain the pre-specified level of QoS. Since the parameters of network traffics are unknown and time varying, the optimal number of guard channels is not known and varies with time. In this paper, we introduce a new dynamic guard channel policy, which adapts the number of guard channels in a cell based on the current estimate of dropping probability of handoff calls. The proposed algorithm minimizes blocking probability of new calls subject to the constraint on the dropping probability of handoff calls. In the proposed policy, a learning automaton is used to find the optimal number of guard channels. The proposed algorithm doesn’t need any a priori information about input traffic. The simulation results show that performance of this algorithm is close to the performance of guard channel policy for which we need to know all traffic parameters in advance. Two advantages of the proposed policy are that it is fully autonomous and adaptive. The first advantage implies that, the proposed policy does not require any exchange of information between the neighboring cells and hence the network overheads due to the information exchange will be zero. The second one implies that, the proposed policy does not need any priori information about input traffic and the traffic may vary.