National and Regional Road Network Optimization for Senegal Using Mobile Phone Data



National and Regional Road Network Optimization for Senegal Using Mobile Phone Data

Authors

Yihong Wang, Delft University of Technology, Goncalo Correia, Delft University of Technology, Erik De Romph, Delft University of Technology

Description

In our research we used mobile phone data provided by Orange for the D4D Challenge. We developed a methodology and created a tool for optimization of the national and regional road network in Senegal.

Abstract

Due to the scarcity of mobility data in Senegal, mobile phone data provided by Orange for the D4D Challenge is used for optimization of the national and regional road network in Senegal.
We first applied a filtering algorithm to estimate inter-departmental origin-destination trip matrices (OD matrices) of sampled users in 2013. We name these matrices relative OD matrices, since we believe that they can reflect the mobility patterns in Senegal in a relative way.
Secondly, based on a literature study on the relations between travel and telecommunication, we explored such relations empirically by comparing the estimated relative OD matrices and the cell phone interaction matrices. The cell phone interaction matrices contain the number of calls and text messages of all Orange’s users, between pairs of departments. We found that the number of trips made by sampled users between each two departments is almost proportional to the number of cell phone interactions and inversely proportional to the travel cost between departments.
Thirdly, based on this observation, we constructed a new type of gravity model, based on the number of cell phone interactions instead of population where the traditional gravity model is usually based on. We estimated the parameters of this new model which gave us a model to predict elastic travel demand pattern for potential road network changes.
In the final step, we used this model to optimize the national and regional network for Senegal. We used an optimization model with the objectives of efficiency and equity. In the model two kinds of action can be performed: the construction of a new road of a given level; and the upgrading of an existing road to a higher level. A local search algorithm is used to find the solutions to this road network design problem. We found that the created tool gained good insight into where and how to expand the Senegal network.

Publisher

Association for European Transport