Modeling Interactions Between Human Factors and Traffic Flow Characteristics

Abstract

To serve research needs for traffic flow model development and highway safety enhancement, we model interactions between human factors and traffic flow character- istics, this topic includes methods on collecting data, modeling impacts of parameters on flow, and calibrating parameters on observed data. An example of successful traf- fic data collection is NGSIM data, which contains location, speed, and acceleration information of vehicles. An algorithm was designed to match and extract vehicles’ trajectory records, and utilize the extracted information for pattern recognition of lane changing maneuvers. This algorithm reads records from an NGSIM data set, pick out vehicles executing lane changing maneuvers, and note the corresponding time stamps. Also through matching these records by vehicle ID and time stamp, we obtain a map of vehicles when a lane changing is happening, thus calculating gaps and relative speeds becomes possible. An example of utilizing these information is pattern recognition on lane changing maneuvers. We analyze lane changing maneu- vers with speed data and gap data. The approach with speed data shows convincing results, as most lane changing vehicles have a descending and then ascending pattern on their speed profiles before executing the maneuver. On the other hand we can use collected data for calibrating parameters in traffic flow models. A heuristic method- ology is implemented to provide results with high accuracy, high efficiency and high robustness. Techniques include data aggregation and bisection analysis are applied in this approach to ensure achieving these goals and further requirements. Two traf- fic flow simulation models, Longitudinal Control Model (LCM) and Newell’s Model are calibrated by applying this approach using traffic data collected at Georgia 400 highway in July, 2003, with satisfying accuracy and robustness produced in a running time of less than 2 seconds. Meanwhile we can enhance human factors by applying new technologies, and connected vehicle is a good example which is rapidly devel- oping. Future vehicles will be able to communicate with each other which greatly improves drivers’ situational awareness. Consequently, drivers may be able to re- spond earlier to safety hazards before they manifest themselves in forms of imminent danger. Therefore, the overall effect of this technology can be attributed to drivers’ enhanced perception-reaction (P-R) capability which, in turn, translates to improved flow and capacity. However, it is critical to quantify such benefits before large-scale investment is made. In our research, a statistical transformation model is formulated to predict the probability distribution function of flow. By entering distributions of P-R time and enhanced P-R time, this model helps compare before and after distri- butions of traffic flow, based on which benefits of connected vehicles on traffic flow can be analyzed.