Abstract:
In Bangladesh and especially in Dhaka City, traffic congestion has been a persistent problem for over a decade. Several attempts have been made, particularly in recent years, to mitigate the congestion problem. However, the congestion situation has continued to deteriorate. In order to combat such a persisting setback a robust solution is required which will be effective, economically feasible and sustainable. Microscopic traffic simulation tools, which model individual driver maneuvers (e.g. car-following/acceleration, lateral movement, etc.) and deduce network condition from those, can be used as laboratories for testing effectiveness of candidate traffic improvement initiatives before their actual field implementation. These tools are therefore increasingly being popular worldwide for selecting the most effective transport planning scheme and evaluate its economic feasibility.
Acceleration is an important driving maneuver which has been significantly modeled to work as a component of micro-simulation tools for several decades. The models are mainly developed for homogeneous motorized traffic where strict lane discipline is maintained. But when traffic stream also comprises non-motorized traffic and lane discipline is barely maintained, the conventional acceleration models developed for homogeneous traffic can no longer be effective. In the previous researches involving acceleration behavior in mixed traffic with „weak‟ lane discipline, only one front or lead vehicle has been used to develop stimulus-response concept based acceleration models. This thesis proposes an updated acceleration model that captures the effect of more than one front vehicle (which is frequently found in the traffic streams with weak lane discipline) and aims to better replicate real traffic situations.
The acceleration model proposed in this research uses an econometrics based approach. The data used for developing these models have been collected from two locations of Dhaka City; Kalabagan and Shukrabad using video cameras mounted on over-bridges. GPS equipped vehicles were run as well in a pre-specified route including these locations in order to supplement the video data. Image processing software „TRAZER‟ (KritiKal Solutions Pvt. Ltd.) was used to perform vehicle count, measure average speed and flow and provide trajectory data. The data was then fed into „MATLAB‟ code and the input variables for the model were generated. A number of models were run using statistical software „STATA SE 11‟ to obtain the best model in terms of variables involved and model result. The effect of reaction time on acceleration/deceleration maneuver of drivers has also been analyzed in this regard. At the end, two distinct models, one for acceleration and one for deceleration maneuvers have been developed. The models, when plugged in a microscopic traffic simulator, can be used to predict the possible actions for a set of factors (independent variables) which directly affect the acceleration and deceleration decisions of drivers in traffic streams with weak lane discipline.
