Automated vehicles are rapidly maturing; AVs will necessarily have different capabilities than human drivers, yet there is a major gap in understanding their likely drive cycles (the profile of speed versus time). Any changes in patterns of speed with respect to time will have structural consequences for the main outcomes from the transportation sector (e.g. mobility/accessibility, energy consumption, pollutant emissions, crash risk exposure, induced travel, etc.)

This research has two objectives:

•      1)  To develop algorithms for plausible and legally-justifiable freeway carfollowing and arterial-street gap
•           acceptance driving behavior for AVs
•      2)  To implement these algorithms on a representative road network, in order to generate representative
•           drive cycles for AVs that are both theoretically-grounded and based on empirical driving conditions.

The Main Proposal Narrative describes the current state of knowledge in this area; the research gap is that studies to date have considered neither the theoretical conditions under which faster travel speeds may be possible (or slower speeds may be required) on the basis of both vehicle kinematics and legal/liability considerations, nor the empirical distribution of the likely changes to speed profiles arising from them.

The theory underpinning the colloquial concept of defensive driving is known as Assured Clear Distance Ahead. ACDA-compliant driving strategies were initially implemented for AVs (in the specific context of queue discharge at signalized intersections) in research recently undertaken by the study team.

Addressing this gap in knowledge regarding AVs’ speed profiles will require extending the application of the ACDA concept to cover a broader set of representative contexts. This will lead to the first major contribution of this research, in the form of novel algorithms for AVs’ car-following and gapacceptance behavior that are grounded on the standard ACDA criterion which human drivers are also required to observe.

The algorithms will then be implemented in a representative empirical context, in order to generate the empirical drive cycles which are the second major product of this research. This requires a representative road network (we will use the enhanced version of the standard Sioux Falls network abstraction), building footprints (available in GIS format), digital elevation map (DEM), and traffic demands. The Sioux Falls network contains a diversity of roadway Page 2 of 2 environments (arterial and freeway), and has been widely-applied by researchers on network-analysis problems beginning in the mid-1970s. ArcGIS software will be used to generate sight lines, and the road network will be coded in VISSIM traffic microsimulation software.

AV driving behavior algorithms developed previously by the study team will be combined with those developed in the earlier phase of the present study, and will be applied in VISSIM using VISSIM’s “COM” developer’s toolkit. Simulations will be undertaken subject to a detailed experimental design. Results (drive cycles of AVs under a range of varying assumptions/inputs) will then be generated via statistical analysis of the vehicle trajectory data.

The targeted outlets for disseminating findings are the Summer 2017 edition of the Automated Vehicles Symposium Mid-Year TRB Conference, and the journal Transportation Research Part C: Emerging Technologies.