Q-Learning Based Optimized Link State Routing Protocol for Different Mobility Pattern

Now days wireless networks have become popular as the mobile applications are increasing day by day and mobility of nodes has become an important feature. The desirable property which separates mobile network from wireless networks is the mobility of communication devices. Therefore, there is a need to design routing mechanism in such a way that they can easily adopt to the frequent changes in the mobility pattern of the network. In this paper, Optimized Link State Routing protocol has been modified by implementing Q-Learning concept, a reinforcement learning algorithm which guides network to select next node to which it should forward packets by first calculating the reward R and then calculation of Q-value with neighbors. Performance of this modified routing protocol has been evaluated for parameters like delay, throughput and delivery ratio. Two mobility models have been used, Random Waypoint and Walk. It is observed that performance in terms of above parameters improve considerably in both mobility patterns when intelligent Q-Learning algorithm is implemented in Optimized Link State Routing.