Application of dynamic lane grouping and artificial intelligence techniques in predicting the optimum lane groups at isolated signalized intersections

Abstract

Signalized intersection is an important element of any road network. Its operations impact adversely the environment and safety and further affect significantly the performance of the whole road system. A considerable variability in traffic demand is expected at most signalized intersections in urban areas. Most of such intersections nowadays are prone to the phenomenon of tide traffic where different traffic movements at each approach (left, through and right) are fluctuating significantly with time. This phenomenon has a significant role in degrading intersections performance and results in congestion along with excessive emissions of harmful gases. This study was conducted to investigate the effectiveness of applying dynamic lane assignment strategy, which is also known as dynamic lane grouping, to optimize signal timing plans. The concept of Dynamic Lane Grouping (DLG) has been introduced to mitigate such operation problems. MATLAB environment was used to build an optimization model to find the optimal lane groups at all intersection approaches for hypothetical massive traffic demand combinations using an objective function of minimizing intersection delay. A comparison was conducted between the average intersection delay for DLG and Fixed Lane Grouping (FLG) at different demand combinations. It is observed that applying DLG yields a significant reduction in average intersection delay compared to FLG. This study also introduced a plausible quick method to predict the optimum lane group in the field instantaneously using the percentage of turning movements at the approach without conducting massive calculations. On the other hand, interviews were conducted to explore the drivers’ response to the information about the existing configuration when disseminated via Variable Message Signs (VMS). The effect of drivers’ characteristics, such as age, occupation, driving experience and education level on their response to VMS, was statistically tested using contingency analysis. It was found that the most significant variable that will affect the drivers’ understanding of VMS is the level of education. Moreover, the Artificial Neural Networks (ANN) model was developed to predict the optimal lane group combinations for any turning movement combinations with an average accuracy of 92%.