Book Description

Simulating a vehicle's motion within a road network has many applications, including autonomous vehicle development and transportation planning. These simulations are expected to include not only predictions of an ego vehicle's behavior, but also interactions with other vehicles, changing rules-of-road, and environmental factors. Some of these interactions are best predicted by vehicle chassis models and simulations, such as the friction utilization of a vehicle on a roadway in changing weather. However, other interactions, such as between vehicles and the rules-of-road or between the ego vehicle and other large groups of vehicles, are best predicted by traffic simulations. As well, the management of data between traffic and chassis simulations can be difficult due to the large number of vehicles and geographic scale in which they operate. There is a need for methods to integrate these toolsets with each other to where the synergistic function of the tools achieves research needs. One research need is to map areas of likely friction utilization within a road network. In a road network, differences in road geometry, traffic patterns, and traffic laws require a range of typical maneuvers, each of which strongly affects the vehicle's expected friction utilization. Therefore, maps of this likely utilization are extremely valuable because they would reveal geolocations within a road network that require little relative friction utilization, and locations of large friction utilization. The purpose of this thesis is to predict areas of large friction utilization within a traffic network by using recorded vehicle trajectories. Because traffic simulations predict infeasible vehicle trajectories from the standpoint of vehicle chassis behavior, a solution based on the adapted hybrid simulation method was proposed and shown to provide physically realistic chassis data for individual vehicles operating in larger road networks with network-scale interactions between vehicles. The resulting feasible vehicle trajectories were then used as reference paths within a simulation of chassis dynamics along with a steering algorithm to predict the friction utilization as a function of road location. Finally, the friction utilization values are mapped to geolocations to identify zones where friction utilization is largest. Knowing these locations allows the planning of maneuvers by both drivers and driving algorithms such that friction margins are maintained. The results show that, within a typical traffic network, there are significant and very highly localized areas where large friction utilization is typical. This finding is very useful to identify risk-prone operational areas, location-specific areas that challenge driver/autonomous driving performance, and/or areas where additional data collection could be helpful. Such areas may gain additional benefit from targeted network management solutions such as reduced speed limits, specialized and/or weather-related light-timing sequences, proactive pavement management methods such as pre-storm road salting, or even advanced Vehicle-to-Infrastructure (V2I) area deployments.