An application of Learning Automata Based ARL to Subchannel Allocation in Cellular OFDMA System

In this paper, a new subchannel allocation schemes for cellular OFDMA networks employing an adaptive frequency reuse factor (FRF) strategy is considered. The allocation algorithm is semi-distributed solution comprising two phases. In the first phase, the Radio Network Controller (RNC) adaptively determines the FRF of each subchannel in a centralized manner. In the second phase, each base station autonomously allocates subchannels to the users using a simple algorithm (i.e. MaxC/I). To solve the first phase, we introduce a hybrid associative reinforcement learning (ARL) model combining self organizing map (SOM) and Learning Automata (LA) to deal with large size and continuous nature of the problem space. The simulation results illustrate that the proposed model achieves a better throughput gain in comparison with other allocation algorithms. It is noteworthy that the proposed algorithm has a low computational cost and achieving this throughput gain is only due to proper assignment of FRF to subchannels.