Differentiating and modeling the installation and the usage of autonomous vehicle technologies: A system dynamics approach for policy impact studies

Existing research that forecasts market penetration of installed connected and autonomous vehicle (AV) technologies is often confused with the traffic composition in roadway networks. Users may override AV mode due to arrival time pressure, facility constraint (e.g., “I will have to make a U-turn a mile away if I do not cross the solid double-yellow lines here”), drug and alcohol influence, pleasure, envy (e.g., “why the front car can surpass that slow truck but I can't?”), insufficient law enforcement, driving culture, media and public sentiment, etc. Therefore, the installation and the usage of AV technologies should not be instantaneously assumed ignorable in planning and policy studies. This paper is dedicated to clarifying this confusion by demonstrating that ignoring the difference between the installation and the usage of AV technologies might lead to systematic bias in evaluating policy and investment decisions. Through a system dynamics (SD) model, the complex interactions of relevant factors are captured so that the mixed traffic condition influences traffic law enforcement adjustment effort and system investment decisions, which, in turn, influence the AV technology usage share and the system performance. The case study applies to the greater Washington, D.C. area for demonstrating the feasibility and advantages of the proposed model and for studying policy implications. This paper does not attempt to forecast; instead, it proposes a modeling framework for studying the conditions under which differentiating the installation and the usage of AV technologies might be critical in forecasting the traffic composition trend and system performance for public policy and investment decisions.