An intelligent road traffic management system using NVIDIA GPU

Abstract

Road traffic congestion remains a global phenomenon that causes great problems in the cities of the world; especially developing countries, resulting in massive delay, unpredictable travel times, increased fuel consumption, manhour and monetary loss. In order to get a better solution, one of the proposition is to divert traffic to less congested route. One of the solution of collecting road traffic condition is crowd sourcing. We proposed that the crowd sourcing information will change road network graph of the city according to sourcing result. Based on this updated graph, the driver in jam needs information that which path is suitable for him based on current updated position of vehicle to intended destination almost in real time. To get a near real time performance for large graph, we try to investigate the use of parallel point to point graph search algorithm. For this purpose we use CUDA enable GPU for parallel implementation of Dijsktra's algorithm. We tested our modified algorithm for New York, Rome and Dhaka city (partial). It is found that parallel implementation gives a better result for all pair search.