Efficient shortest path routing in computer communication networks using neural network and heuristic algorithm
Loading...
Date
2011
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
University of Management and Technology
Abstract
Computer Networks are a breakthrough in current communication technologies providing an always-on and ubiquitous connectivity to virtually millions of users with the score still mounting up. This increased usage strongly requires enhancement in network infrastructure and improvements in management to facilitate users with better quality services. As the rate of users, who are transmitting data through a network link, increases the quality of service has to be compromised if not improved. Routing, which is a major aspect of this domain, has significant impact on the effectiveness of Communication Systems. Enhanced technologies should be introduced to let the system assist problems in a better way, preventing data loss and degradation in quality.
Efficient routing algorithm should strive for the most appropriate and shortest path to route data through a network. Prior information of routers and data links, which helps to create paths, is required in almost every technique. A search algorithm, exploiting the properties of mathematical routing algorithm as well as of heuristics, can compute shortest path between given pair of routers more efficiently.
On the other hand artificial intelligence can be very helpful in this domain. Artificial Neural networks, a domain of artificial intelligence, have found their way in engineering as well as Medical Sciences. The particular adaptive ability of neural networks for dynamic situations becomes a strongest feature which renders it best suitable for dynamic systems. John Hopfield used this feature and found their application in NP-complete optimization problems. Therefore, neural networks similar to Hopfield's can also help to enhance shortest path computational techniques in routing.
A* search algorithm and neural network are used to find shortest routes for unicast routing problems. Computer simulations are used to analyze and compare the results for different network sizes. A* search algorithm outperforms the neural network so far; however, limitations of neural networks are discussed.
Description
Project Advisor: Dr. Abdul Aziz Bhatti
Keywords
MS Thesis, Communication Technologies, Computer Networks