Automated Route Selection; Short Term Traffic Decision Support For Nairobi

Printer-friendly version

The city of Nairobi is currently grappling with the problem of rapidly increasing traffic, and its management.  We have developed a prototype decision 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 prediction 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 optional 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 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 optional 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 Dijkstra) from a naive perspective.  A one week survey of existing road traffic speeds was conducted using a probe car fitted with a GPS.  The average speed recorded for Nairobi wa approximately 25km/hr.

A.I. techniques can be deployed within the framework of GIS based decision support systems to fundamentally predict short term traffic congestion, simulate scenarios to enhance traffic management and help in creating policy for long term sustainability of infrastructure.

The A* search is effective for small networks as seen in Nairobi However, care needs to be taken in developing the heuristic component.  If it is small, the A* decomposes to a greedy search and perfoms similarly to the Dijksta’s algorithm.  Other factors need to be considered as identified in this research in fine tuning the A* search in terms of road characteristics and traffic influence for instance surface condition, location, width and gradient..

A critical generic component of a DSS is a visualization system or graphical user interface.  As demonstrated in this report, GIS is critical in traffic management as a visual data mining tool.  By visualizing the results of the search module, a user is ableto assess the maturity of our road network and identify suitable routes to expand or build mechanisms to control traffic.  The speed survey carried out identifies roundabouts as most critical bottle necks.

In conclusion, the city of Nairobi needs to deploy a traffic and route management system as proposed by this research.  This will cut down the response time of emergency services and also warn people on identified routes of oncoming emergency vehicles and personnel thus create space.  It goes without saying that data is not ready accessible in Kenya as experienced by the researchers.  It is important for the government and academic institutions to partiner in research and surveys to ensure that data collected is readily available for future research and analysis.

We have reported our encouraging findings here.

UoN Website | UoN Repository | ICTC Website


Copyright © 2018. ICT WebTeam, University of Nairobi