Application of Artificial Intelligence Techniques for Developing Operating Speed Models for Rural Multilane Highways in Egypt

Abstract

For many years, roads in Egypt suffered from inconsistency between posted speed and operating speed as well as high speed differentials between operating speed of successive elements. This may be due to inconsistency between road’s geometric elements and design speed. Numerous studies developed operating speed models based on regression analysis to study design consistency. However, little is known for multilane rural highways in Egypt. The main objective of the current research is to develop operating speed models for rural multilane roads in Egypt using artificial neural network modeling technique and to compare it with regression analysis technique. Free flow speeds were collected at the middle of the horizontal curve and at the middle of the straight section in six roads (Cairo Alexandria desert Road, Katamia-Ain Sokhna Desert Road, and Cairo-Alexandria Agricultural Road) for curve sections and straight sections. A set of 65 curved sections and 66 straight sections were studied from these roads. Operating speed prediction models for curved sections and straight sections were developed using neural network technique and traditional regression analysis. It was found that operating speed exceeded posted speed at almost 60% of studied sections. Therefore, more speed enforcement by mean of traffic control devices is recommended. In addition, it was found that most significant parameters were deflection angle, posted speed Vexceeding, and curve radius for curved sections. While in straight sections, most significant parameters were posted speed, Vexceeding, and pavement width. In addition, it was revealed that neural network explains the effect of parameters on operating speed more clearly than regression models with higher predictive power. However, it needs more data to improve accuracy of the model.