INTEGRATING DRIVING SIMULATOR EXPERIMENT DATA WITH A MULTI-AGENT CONNECTED AUTOMATED VEHICLES SIMULATION (MA-CAVS) PLATFORM TO QUANTIFY IMPROVED CAPACITY

Abstract

Autonomous vehicles (AVs) at varying market penetration rates will change traffic flow and highway performance. At AV market penetration rates of between 0 percent and 100 percent, human-driven vehicles (HVs) will be interacting with AVs. However, little is known about how HVs interact with AVs. Using the Oregon State University Driving Simulator, this study measured HV headways when drivers followed an AV and integrated those data into a multi-agent simulation to quantify new highway travel time and flow predictions at varying AV market penetration levels. This study also collected galvanic skin response data to quantify drivers’ levels of stress when presented with a hard-braking AV and HV. The driving simulator experiment was successfully completed by 36 participants. The results of this study showed that drivers’ levels of stress were 70 percent higher in hard braking scenarios involving HVs versus AVs. Additionally, drivers over the age of 34.5 were found to give AVs 2 percent more headway than HVs, while younger drivers gave AVs 18 percent less headway than HVs. Thirty-six scenarios were tested in the multi-agent simulation using results from the driving simulator. Given the driving simulator results, average travel times were found to increase at most by 2.3 percent, while flow was found to decrease at most by 1.3 percent