Traffic Delay Estimation Using Artificial Neural Network (ANN) at Unsignalized Intersections

Sahraei, Mohammad Ali; Puan, Othman Bin Che · 2018 · OpenAlex-citations

DOI: 10.11159/iccste18.106

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This study addresses the inadequacy of existing theoretical models, specifically the Malaysian Highway Capacity Manual (MHCM), in accurately estimating control delays at unsignalized intersections in Malaysia. The research was motivated by findings that MHCM formulas significantly overestimated delays compared to observed field data, particularly under high conflicting traffic volumes. The authors aimed to develop a more accurate prediction model using Artificial Neural Networks (ANN) to estimate control delays for vehicles turning left and right from minor roads. The methodology involved collecting traffic data from three unsignalized intersections in Johor Bahru, Malaysia, using video camera recordings over 27 hours. Data collection included measuring control delays via stopwatch, counting traffic volumes, and recording gap acceptance behaviors. The study compared observed delays against MHCM estimates, revealing statistically significant differences and poor fit ($R^2$ values of 0.16 and 0.45 for left and right turns, respectively). To address this, the researchers developed an ANN model using the Levenberg-Marquardt algorithm with two hidden layers. Input parameters included critical gap, follow-up time, conflicting flow rate, movement flow rate, and proportions of motorcycles, lorries, and heavy vehicles. The network was trained on data from two sites, and explicit mathematical formulas were derived from the trained weights and biases. These formulas were validated using independent data from a third site. The results demonstrated that the ANN-derived formulas provided a significantly better fit to observed data than the MHCM model. Validation using the third site yielded high correlation coefficients ($R^2$ of 0.81 for left turns and 0.80 for right turns) and low error metrics, with t-tests indicating no statistically significant difference between predicted and observed delays. Sensitivity analysis of the derived formulas revealed that movement flow rate and conflicting flow rate had the highest impact on delay, potentially increasing control delay by up to 39% when these factors increased from 10% to 50%. In contrast, the proportion of heavy vehicles had the lowest effect, causing only a 1% to 3% increase in delay under similar conditions. The significance of this research lies in providing a more accurate, context-specific tool for estimating control delays at unsignalized intersections in Malaysia. By demonstrating the limitations of the standard MHCM model and offering validated ANN-based formulas, the study contributes to improved traffic engineering practices. The findings suggest that neural networks can effectively capture the complex, non-linear relationships between traffic variables and delay, offering a robust alternative to traditional theoretical models that may not account for local driving behaviors and intersection characteristics.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• traffic density