Dynamic user optimal model for shared autonomous vehicles system: Development and systematic comparison with social optimal model

Abstract

Shared autonomous vehicle (SAV) system is a new type of public transportation in which autonomous vehicles shared by the society transport travelers using optimal routes and ridesharing matching. In the literature, the dynamic system optimum (DSO), in which travelers' behavior is completely controlled by some transportation authority to maximize the entire society's benefit, is mainly considered for SAV system analysis. However, it is known that DSO could cause unfairness among travelers and may not be acceptable to some travelers due to the lack of freedom. On the other hand, the dynamic user optimal (DUO) assignment computes transportation system's state based on travelers' free and selfish behavior, and is useful to simulate traffic congestion with a realistic traveler behavior. This study develops a novel DUO assignment model for SAV systems. It is formulated as multi-step linear optimization problems to describe the multi-layer structure of SAV systems (i.e., travelers and SAVs follow different behavioral principles) while keeping the computational efficiency. The model is evaluated using actual network and demand data in Japanese urban area, and systematic comparison with a DSO model for SAV systems is conducted. As results, we confirm that overall performance of the SAV system was degraded in the DUO model, because the congestion level of some popular links were significantly increased due to the selfish behavior of travelers. Qualitatively, this is an expected result; the contribution of this study is to develop a method for quantitative analysis of this phenomenon. Such quantitative results would be useful to develop appropriate management schemes for SAV systems such as dynamic pricing, which is the most important future work.