Traffic prediction tool for enhancing wireless network management
Chiman Kwan, Roger Xu, P. Clark, Charlie Hammond · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 2001
Wireless networks have several special characteristics. First, the bandwidth is extremely limited and hence a careful sharing is needed in order to fully utilize this scarce resource. Second, due to the unreliable nature of wireless channels, quality of service becomes very important to ensure good communications among mobile users. Third, the network structure changes frequently as users may move around from one location to another. This dynamic change in network structure causes some problems in network resource management. The key idea of this paper is to manage bandwidth based on predicted future traffics in the system. Conventional approaches to network management are reactionary in nature. That is, the decision on bandwidth allocation is based on past measurement of traffic and hence is slow and inefficient in response. In contrast, our traffic prediction based bandwidth management tool is a look-ahead approach and will be more efficient and quick in bandwidth management. Moreover, much less collisions will occur among users and less bandwidth will be wasted. Here we summarize some preliminary results on traffic prediction by using neural networks. Future work will include the integration of traffic prediction tool with network bandwidth management.