Modeling Of Car Following Behavior Considering Visual Attention Performance

Car following behavior which is the most fundamental and most important driving behavior describes interaction between the leader and follower on the single lane with no passing. Car following model is the study of the dynamics of follower behavior which induced by the motion of leader. Through the analysis of car following approach one by one, the traffic flow characteristics on the single lane can be understood and the bridge between microcosmic driver behavior and the macroscopic traffic phenomena can be set up. The study of car following behavior has a wide range of applications such as microscopic traffic simulation, driving behavior analysis, road capacity analysis, intelligent vehicles, and traffic safety. Car following theory has been one of the core content of the traffic flow theory.Conventional modeling methods of car following behavior use driver-vehicle unit as a follower and the general method of mechanical motion is used in the car-following model. However, driver is the core of the car following process and his/her performance will greatly affect the car following behavior. Modeling of car following should have more consideration on human factors. Therefore, the dissertation is based on driver visual attention performance and the visual information will be introduced to the car following behavior modeling. Then the conceptual model will be modeled by taking the driver visual attention performance into account. Visual information in car following behavior is introduced into the GM models and the OV models and has the important impact of car-following behavior, so that the car following models are more in line with the performance of the human driver. In addition, the car following models considering the visual information can explain the macroscopic traffic flow characteristics, thus the relation between the micro-behavior and the macro-phenomenon is established. This dissertation mainly divided into the following four parts.1) Review of existing research results This dissertation review the car following model development process from a system point of view. Based on the modeling ideas, the study of car following models are divided into the traffic engineering point of view and the statistical physics point of view. Then the various types of modeling ideas, models structure, and parameter calibration process are overviewed. In the traffic engineering point of view car following models are classified into stimulus- response models, safety distance models, psycho-psysical models, and artificial intelligence models. In the statistical physics point of view car following models are classified into optimal velocity models, intelligent driver models, and cellular automata models. In addition, other types of driving behavior models, such as changing-lane models and the integrated driving behavior models were summarized. At last, the development of car following models are: the cross-integration of a variety of methods, the unification of micro-behavior and macro- phenomenon, the specialization and integration of car following models, more consideration to human factors, and a unified calibration standard.2) Car following abstract modeling taking into account drivers'visual attention performance Firstly, through the various stages of car-following behavior analysis, car following process is divided into four parts including the driver's perception, processing, decision making and manipulation. Such the driver visual attention performance has importance on car following behavior. Secondly, the abstract model of car following considering visual attention performance is proposed. The abstract model mainly includes three aspects, namely, modeling of the drivers'visual perception characteristics, the drivers'visual selective attention mechanism and vehicle kinematic and dynamic models. The effects of the characteristics of the object, shape, position and state of motion are considered in the modeling of the drivers'visual perception performance. Visual selective attention mechanism is presented to explain the behavior of the drivers'selective attention process of each object in the car following situation. Based on two degrees freedom vehicle model, the kinematic and dynamic models of the vehicle are established.3) GM model considering visual information Based on the drivers'visual angle information, the expressions of time to collision of the leading vehicle are set up under the car-following situation. And then the time to collision of any vehicle is extended. Let the time to collision as input variable of car-following model to improve the GM model, and GM model considering drivers'visual information is proposed. Through the integration of the proposed models, the macroscopic flow-density relation model is found. The effects of vehicle lateral deviation to the macro-relation models and the lane capacities were analyzed. Through simulation, the local stability, the asymptotic stability, lateral movement and the effect of the neighboring lane vehicles were analyzed. The simulation results show the validity of the proposed models.4) Optimal velocity model considering visual information Put the drivers'visual angle information into optimal velocity model. It means that the driver is based on visual angle and its rate of change to optimize speed. Through the theoretical analysis and numerical simulation, the linear stability conditions of the presented model are obtained. Compared to full speed difference model, the stability region in low-density is smaller and the stability region in high-density is greater, which coincides the actual conditions. Simulation results show that the introduction of visual angle information can be well explained the actual asymmetry acceleration and deceleration behavior and the wide scatter data in the macroscopic traffic flow. Finally, through the two-dimensional expansion of the car following model, a pedestrian-vehicle interference model was established. The model was used in the simulation of pedestrian crossing systems. The effects of pedestrian volume, pedestrian critical gap and vehicle density to the delay of the pedestrian and vehicle and their capacity were analyzed. The pedestrian crossing signal warrant is presented based on simulation data.In the end, the dissertation summarizes the findings and achievements, and brings forward the issues for further research.