Automated route selection: Short term traffic decision support for Nairobi

Abstract

The city of Nairobi is currently grappling with the problem of rapidly increasing traffic, and its management. We have developed a prototype decision support system for short term traffic prediction and subsequent shortest path analysis for this City. We investigated on the use of artificial neural networks in time series predication and the application of the optimal A* search algorithm for the shortest path between two points. A geographical information system was used to visualize both the road network and optimal paths. Topographical maps of Nairobi were digitised and a GIS topology build to support the A* search routine. For purposes of simulation, historical traffic data collected from Kenya Institute of Public Policy Research and Analysis was formatted, analysed and pre-processed using a sliding window time series and modelled using a feed forward back propagation artificial neural network. The resulting network was used to predict one step-ahead traffic speeds. With the traffic speed and other road network parameters such as lane width and surface type just to mention, these values were then used to calculate the time taken to traverse a node or a link. In essence the actual length of the road was modified to a virtual length, while the speed determined from the ANN. The resulting time value was used to process the A* search routine resulting to an optimal path visualised on a GIS interface. For purposes of objectivity, the Dijkstra search routine was deployed to compare and contrast the two search routines (A* and Dijikstra) from a naive perspective. A one week survey of existing road traffic speeds was conducted using a probe car fitted with a GPS.