Citation

Patnaik AK, Ranjan AR, Bhuyan PK. Transp. Res. Rec. 2018; 2672(15): 35-43.

Affiliation

Department of Civil Engineering, NIT Rourkela, India Corresponding Author: Address correspondence to Ashish Kumar Patnaik: ashishkumar.patnaik@yahoo.in

Copyright

(Copyright © 2018, Transportation Research Board, National Research Council, National Academy of Sciences USA, Publisher SAGE Publishing)

DOI

10.1177/0361198118777603

PMID

unavailable

Abstract

The primary objectives of this study are to develop two roundabout entry capacity models using a regression-based multiple non-linear regression model (MNLR) and artificial intelligence (AI)-based ANFIS (adaptive neuro-fuzzy inference system) model under heterogeneous traffic conditions. ANFIS is the latest technique in the field of AI that integrates both neural networks and fuzzy logic principles in a single framework. Required data have been collected from 27 roundabouts in eight states of India. To assess the significance of these models and select the best model among them, modified rank index is applied in this study. The coefficient of determination (R2) and Nash-Sutcliffe model efficiency coefficient 'E' values are found to be 0.92, 0.91 and 0.98, 0.98 for the MNLR and ANFIS model, respectively. The ANFIS model is found to be the best model in this study. However, from a practical point of view, the MNLR model is recommended for determining roundabout entry capacity under heterogeneous traffic conditions. Sensitivity analysis reports that critical gap is the prime variable and shares 18.43% for the development of roundabout entry capacity. As compared with the Girabase formula (France), Brilon wu formula (Germany), and HCM 2010 models, the proposed MNLR model is quite reliable under low to medium ranges of traffic volumes.

Language: en