Prediction of operating speed on horizontal curves of two-lane rural highways using artificial intelligence

Horizontal geometric characteristics have significant impact on vehicle operating speed of vehicles on two-lane rural highways. Majority of the studies have used the conventional approach of modelling, found to be location specific and provide false judgement of determining operating speed. Thus, it becomes important to apply methods based on artificial intelligent for predicting operating speed of vehicles on two-lane roads under mixed traffic conditions. Field data was collected on 40 different locations (curves and tangent sections) that includes free speed of vehicles and geometric parameters. Geometric parameters such as curve radius, curve length, deflection angle, degree of curvature and preceding tangent length were measured in the field with total station, whereas free-flow speed data was collected using radar gun at the mid of the horizontal curves. The statistical analysis concluded that the curve radius, curve length, degree of curvature and preceding tangent length are found significant on the operating speed of vehicle type Car, Two-wheeler, Three-wheeler, Light commercial vehicle and Heavy commercial vehicle and developed MLR model. Further, the data driven soft computing methods such as Artificial Neural Network (AN), Adaptive Neuro-fuzzy Inference System (ANFIS) and Support Vector Regression (SVR) are applied to predict the operating speed of these vehicles and results were compared with MLR. The performance of the models evaluated using various goodness-of-fit measures indicates that the SVR model gives better results in prediction of operating speed in compared to other models. As for future research, further investigation could be conducted to explore uncertainties, and the model could be enhanced by utilizing other geometric and traffic parameters, and techniques like random forest, XGBoost etc.,