Prediction of short-term average vehicular velocity considering weather factors in urban VANET environments

Recently, accurate prediction of short-term traffic flow is crucial to proactive traffic management systems in ITS; however, the drivers need the average vehicular velocity more than traffic flow while driving. The drivers could change the path immediately according to the average vehicular if the average velocity of the next road segment is predicable. In this paper a neural network is used for prediction of average velocity, besides vehicles can collect the average velocity of current road segment to adjust the predicted average velocity of the next road segment. The collected average velocity is acquired from neighbor vehicles through VANET. There is no research considering the impact of weather factors on the average vehicular velocity previously. An example of weather condition affects the velocity, it is always low vehicular velocity on rainy day or in fog. In this paper, the proposed prediction considers the weather factors that include temperature, humidity and rainfall. This research is focus on urban VANET environments of Taipei in Taiwan, and the results show that the prediction of average velocity considering weather factors is more accurate than that without considering weather factors.