Refined car-following model incorporating the behavior of the vehicles from the adjacent lanes

Abstract

This thesis approaches the road traffic modeling process at the microscopic level, focusing on the car-following models. A review of the recent trends in this field is provided by the state of the art chapter. This research emphasizes the disadvantage of the standard car-following model consisting of neglection of the behavior of vehicles moving on the adjacent traffic lanes. This study proposes a refinement process of this standard car-following model considering the behavior of the vehicles moving on the adjacent road traffic lanes in the follower vehicle acceleration control. Traffic lanes are modeled as nodes in a Markov chain, and the lane choice probabilities computation employ the Bayesian reasoning concept. A fault detection based on parity equations is performed to detect computational faults introduced by the refinement process. This research also brings novelties to the calibration process by proposing a hybrid online calibration method that combines the concept of Kalman filtering with the Takagi-Sugeno fuzzy inference system. This calibration method proved its utility for both standard and refined car-following models. All these new approaches have been validated through simulation experiments done in Simulink, part of MATLAB R2020a (MathWorks, Natick, MA, USA) based on real traffic data received from Timișoara City Hall - General Directorate of Roads, Bridges, Parking, and Utility Networks - Traffic Monitoring Office, Timișoara, Romania.