A Comfort-Based Vehicle Routing Methodology for Autonomous Vehicles

Ensuring passenger comfort is a way to enhance market penetration rates and realize the advantages of autonomous vehicles (AVs). Unfortunately, only a few studies explored this domain. While several factors affect passenger comfort, this study focuses on route choice as an important attribute of passengers' comfort. Accordingly, this study introduces a methodology to generate the most comfortable path for AVs (instead of the shortest travel time). The first step towards realizing this goal is defining the relationship between passenger/user comfort and AVs' speed using a Support Vector Machine (SVM) based model, which is developed based on a dataset collected by the authors from passenger comfort in real-world driving scenarios. The SVM's output is called comfort costs, ranging from 0 to 2 with 0 being comfort, 1 indicating neutrality, and 2 representing discomfort. These comfort costs are then input into Dijsktra's algorithm to find the most comfortable path. The proposed methodology is tested in the University of Illinois at Urbana-Champaign street network. The findings show that the most comfortable path may increase travel time and induce completely different network-level traffic flow dynamics. The findings also suggest the importance of considering a trade-off between the comfort costs and the travel time.