Artificial Intelligence Techniques Based Modeling of Bicycle Level of Service for Urban Road Segments

Abstract

Bicycle is one of the essential modes of transportation in a developing country like India. There is no safety to bicyclists in the mixed traffic and high speed moving vehicles. Hence we should emphasize on them by considering factors affecting bicyclists comfort in planning and design stage itself. In India there is heterogeneous traffic where we find interactions between bicycles and vehicles. Since no methodologies are available there is a need to develop a methodology that gives the perceive comfort level of bicyclists on road segments in mid-sized cities. In this study BLOS Model is developed using three techniques namely Artificial Neural Networks (ANN) and Multi Gene Genetic Programming Methods (MGGP) and Multi linear Regression (MLR). Overall 59 segment data is used for analysis which is collected from Rourkela, Bhubaneswar and Rajahmundry. Eight significant input parameters are considered in the models namely, width of outside through lane (WOTL), peak hour volume for a single lane (PHV/L), pavement condition index (PCI), land use pattern (LU), average traffic speed (S), interruption by public transits (IBPT), on-street parking activities (P) and presence of commercial driveways (D).BLOS model equations are developed for all three techniques. Sensitivity Analysis is carried out to determine the important parameters highly affecting the BLOS. Performances of these models have been tested in terms of several statistical parameters such as: correlation coefficient (R), maximum absolute error (MAE), absolute average error (AAE) and root mean square error (RMSE) and the best model is obtained. In present study MGGP based BLOS model has good performance compared to ANN and MLR techniques.