Developing an Automated Drivability Index for Swedish Roadways

Abstract: The masters thesis aims to determine a method to understand which road segments could pose safety risks or where Level 4 autonomous vehicles would operate poorly in the Swedish road context. This vehicle’s ability to operate adequately can be defined as ‘drivable’. Drivability for Level 4 autonomous vehicles is defined as an autonomous vehicle’s ability to operate safely, efficiently, and comfortably. This includes the autonomous vehicle’s capacity to perceive and navigate its environment, interact with other road users, ensure a smooth and comfortable ride, minimize abrupt movements or harsh maneuvers, and optimize driving behaviors for energy efficiency. A real time drivability index is created as it allows for easy comparison of road segments spatially as well as temporally. A general framework is proposed to calculate this real time index that includes elements from various sources including aspects of the physical and digital infrastructure, the presence of other road users, and environmental conditions. The framework structure is broken into the following 6 components, each represented by several static and real-time indicators. The components are: the number and diversity of objects in the street space, the condition and configuration of the road infrastructure, the speed limit, the stability of the operational design domain, the digital connectivity, and localization ability. Two weighted sum models are proposed dependent on the availability of data to calculate the drivability index for a given road segment. To calculate the relative importance of each component, experts in the field of AVs in the Swedish road context were contacted to complete a survey. The results of the survey were used as part of an analytic hierarchy process analysis to determine the relative importance of each indicator in each component and the relative importance of the components. The survey and analysis yielded the following relative weights: the number and diversity of objects in the street space: 0.35, the condition and configuration of the road infrastructure: 0.23, the speed limit: .09, the stability of the operational design domain: 0.15, the digital connectivity: 0.10, and localization ability: 0.08. A proof of concept was conducted for the City of Uppsala to better understand the requirements of implementing the framework to calculate the drivability index. Various indicators were combined as well as omitted from the framework calculation due to technical, financial, and time constraints. Data was gathered over two days from May 15th and May 16th. 4 generalized scenarios were created to understand the different time periods throughout the day. The proof of concept showed that the centre of the city had road segments that exhibited the poorest drivability index in the whole road network of Uppsala. Roads farther away from the city centre exhibited higher drivability indexes, the best classified road segments included motorways and primary roads. During the off-peak hours when the level of pedestrians was low, and schools were out of session, residential roads and roads closer to the centre of the city exhibited higher drivability than during the peak hours. Many residential roads that exhibited low drivability during the peak hours exhibited the same driveability as the best preforming road segments. The study shows that a dynamic real time index provides valuable insight into the potential for Level 4 autonomous vehicles in Uppsala. The city centre needs be avoided by autonomous vehicles during the day but during the off-peak hours, the centre is more suitable for autonomous vehicles.