Traffic estimation and real time prediction using adhoc networks

This paper presents the process of developing a multilayer feed forward neural network combined with a backpropagation algorithm for forecasting travel time and traffic congestion. Prediction of travel time and traffic congestion based on past and current traffic information is not straightforward due to among others, the high complexity and ill predictability of traffic process, incorrect observations and different data sources. However it appears that neural networks can be exhaustively used to solve these problems. The system is designed on top of a mesh based communication infrastructure for the mobile nodes to communicate. Communication network comprises of multiple networks, i.e. VHF, UHF. The mesh based communication approach enables easy deployment of the system in real world. OLSR routing protocol is used for establishing an ad hoc network for peer-to-peer-communication