Cog-MAC Protocol: Channel Allocation in Cognitive Ad Hoc Networks Based on the Game of Learning Automata

Abstract

Spectrum allocation is a major challenge in wireless ad hoc networks due to restrictions on the number of available communication channels and also passive approaches to channel allocation in dealing with changes in the network environment, and it is impossible to achieve a perfect operational status, which includes a comprehensive communication. Adding cognition to current networks enables them to plan, learn, and reason, and also enables network nodes to self-adapt themselves to their surroundings and to adjust the required bandwidth to perform their operations. This paper introduces a cognitive ad hoc network with two cognitive elements with the aim of improving the use of wireless spectrum. These elements are placed at each node of the network. One element predicts the traffic at each node using neural networks, and the other solves the problem of channel allocation according to this predicted traffic using a non-cooperative game of learning automata and the techniques for allocating variable-width channels. Then, an analysis is done on the convergence of the algorithm and achieving the Nash equilibrium in the game. The purposes of the proposed algorithm are efficient spectrum use, reducing interference, and achieving the ultimate goals of the network, which are quality of service parameters such as high throughput and low latency. Simulation results showed the superior performance of the proposed algorithm compared to alternative approaches.