Flexible and Scalable Routing Approach for Mobile Ad Hoc Networks by Function Approximation of Q-Learning

Date
2016
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Wireless mobile devices are rapidly spreading to the extent that it is hard to find a person not exposed to such technology. These devices could be connected directly or indirectly by wireless channels to form a mobile ad hoc network (MANET). Finding a route for flow from a source to a destination in a network is known as routing. Dynamic topology and unstable link states are the main problems facing routing in MANETs. This thesis employs reinforcement learning, namely Q-learning to develop a routing mechanism. Features inspired from the network are used in approximating the Q-function to form a new intelligent routing metric. This way, the routing process concentrates on specific routes instead of network-wide broadcasting. Accordingly, it is possible to achieve flexibility and scalability in routing. Advantages of the proposed routing technique have been highlighted by conducting experiments in two MANETs environments, namely hand-held devices based MANETs and VANETs.
Description
Keywords
Artificial Intelligence, Computer Science
Citation
Elzohbi, M. (2016). Flexible and Scalable Routing Approach for Mobile Ad Hoc Networks by Function Approximation of Q-Learning (Master's thesis, University of Calgary, Calgary, Canada). Retrieved from https://prism.ucalgary.ca. doi:10.11575/PRISM/26186