Microscopic, Macroscopic and Stability Analysis of Reactive Agent-based Car Following Models

This paper presents the development and evaluation of a reactive agent-based car following model. The formulation of the model is similar to the desired spacing models which do not consider reaction time or attempt to explain the behavioural aspects of car following. The models were developed using field car following data, comprising relative speed and headway between vehicles. Quantitative statistical tests were used to compare the performance of the agent-based models against established car following models using the field data. The results from the model development phase showed that a simple back-propagation neural network model outperformed the Gipps and Psychophysical family of car following models. Qualitative microscopic comparisons of distance trajectories, speed, headway, and drift behaviour provided very close agreement with the field data. Macroscopic speed-flow-density relationship also confirmed the findings. Car following stability analysis in mild and severe disturbances showed that the reactive agent-based models performed reasonably under the two disturbances. The model was then implemented in a microscopic traffic simulator, AIMSUN, to test the computation performance which was also satisfactory. (a) For the covering entry of this conference, please see ITRD abstract no. E214133.