Study Of Routing Protocols In Cognitive Networks

Traditional communication networks can not adapt to dynamic network environment well. This limitation results in obstacles to efficient optimization of end-to-end performance. In order to solve these problems, the concept of Cognitive Networks was proposed. Cognitive Networks can dynamically adapt to varying network conditions through learning and reasoning to optimize end-to-end performance. In the study of Cognitive Networks, the design of routing protocol with cognitive ability is an important issue.At first, a cognitive routing protocol called DQR (Dual-path Q-learning Routing) protocol is proposed in this paper. Using Q-learning algorithm, DQR can learn and reason the operating state of the network, and configure the live time of routing accordingly. In this way, DQR can efficiently reduce the redundant routing overhead. Then, to further enhance the performance, the newly proposed DQR protocol is integrated with a Wiener processes Prediction based routing algorithm to form a PDQR (Prediction-based Dual-path Q-Learning Routing) protocol. The PDQR predicts the traffic load of the two paths and dynamically selects the path with light traffic load to send data packets. Finally, OPNET simulation studies are presented to validate the performance of the DQR and PDQR protocol.