Learning Automata Based Dynamic Guard Channel Algorithms Hamid Beigy ⋆ Computer Engineering Department, Sharif University of Technology, Tehran, Iran Institute for Studies in Theoretical Physics and Mathematics (IPM), School of Computer Science, Tehran, Iran beigy@ce.sharif.edu M. R. Meybodi Computer Engineering Department, Amirkabir University of Technology, Tehran, Iran Institute for Studies in Theoretical Physics and Mathematics (IPM), School of Computer Science, Tehran, Iran mmeybodi@aut.ac.ir Abstract. In this paper, we first propose two learning automata based decentralized dynamic guard channel algorithms for cellular mobile networks. These algorithms use learning automata to adjust the number of guard channels to be assigned to the cells in the network. Then, we introduce a new model for nonstationary environments under which the proposed algorithms work and study the behavior of algorithms when they use LR-I learning algorithm. It is also shown that a learning automaton operating under the proposed nonstationary environment equalizes its penalty strengths. Computer simulations have been conducted to show the effectiveness of the proposed algorithms. The simulation results show that the performances of the proposed algorithms are close to the performance of guard channel algorithm that knows all traffic parameters. 1 Introduction In the last decade, there is an increase in the popularity of mobile computing systems, which results an increase for channel (bandwidth) demands. Since the number of channels allocated for this purpose is limited, cellular networks are introduced, in which the service area is partitioned into regions called cells and evry cell is serviced by a server called base station. When a mobile station moves across the cell boundary while using channels, handoff is required. If an idle channel is available in the destination cell, then the call is resumed; otherwise the call is dropped. The dropping probability of handoff calls (B h ) and the blocking probability of new calls (B n ) are important quality of service (QoS) measures of the cellular networks. Since the disconnection in the middle of a call is highly undesirable, dropping of handoff calls is more serious than blocking of new calls. Blocking more new calls generally improves dropping probability of handoff calls and admitting more new calls generally improves blocking probability of new calls. In order to control these QoS measures, call admission control algorithms are introduced, which determine whether a new call should be admitted or blocked. Both blocking probability of new calls and dropping probability of handoff calls are affected by call admission control algorithms. Several call admission algorithms have been proposed in the literature. Fractional channel algorithm accepts new calls with a certain probability that depends on the current channel occupancy and accepts handoff calls as long as channels are available [1]. A restricted version of this algorithm is uniform fractional channel algorithm (UFC), which accepts new calls with probability of π independent of channel occupancy [2]. It is shown that there is an optimal π ∗ , which minimizes the blocking probability of new calls with the constraint on the dropping probability of handoff calls. An algorithm for finding π ∗ and conditions for which the uniform fractional guard channel performs better than guard channel is given in [2]. Another restricted version of fractional channel algorithm is called guard channel algorithm, which reserves a subset of channels allocated to a cell, called guard channels, for handoff calls (say C − T channels) [3–5], where 0 ≤ T ≤ C is a threshold and C is the number of channels allocated to the cell. Whenever the channel occupancy exceeds the threshold T , the algorithm rejects new calls until the ⋆ This research was in part supported by a grant from Institute for Studies in Theoretical Physics and Mathe- matics (IPM), Tehran, Iran.