Search

The rise of autonomous driving may sharply reshape ride-sourcing markets by influencing drivers' income and participation, customers' choices, ride-sourcing platforms' operations, and governments' regulations. We introduce a model that characterizes a ride-sourcing market with a mixed fleet of non-atomic autonomous vehicles (AVs) and atomic human-driven vehicles (HVs). Human drivers decide whether and where to offer ride-sourcing services independently, while AVs are fully controlled and dispatched by the platform to serve on-demand ride requests that are distributed heterogeneously. We solve for the optimal platform operating strategy and examine the implications of different government regulations on the platform's optimal decisions and the resulting market equilibrium. Although the emergence of AV might undermine human drivers' welfare, it can simultaneously increase platform profit, customer surplus, and social welfare. Moreover, by looking into the regulatory outcomes of a few prevailing regulations (i.e., price cap, minimum wage, minimum fleet size, and maximum fleet size), we find that the policymakers should determine the regulations according to the conditions of labor supply and the development stage of AV technology. • Develop an analytical model to characterize a mixed fleet of autonomous vehicles (AVs) and human-driven vehicles (HVs). • Identify the platform's operating strategies under distinct market structures. • Compare the monopoly optimum with the social optimum. • Examine multiple regulations that can potentially improve social welfare. [ABSTRACT FROM AUTHOR]

Before having a massive deployment of fully connected and autonomous vehicles (CAVs), CAVs with different automation levels and human-driven vehicles (HVs) will coexist on roads for a long time. To quantify the impacts of CAV technology on vehicle market penetration and travelers' route choices in the future from the perspective of transportation planning, this paper investigates a two-sided market equilibrium problem, which considers the vehicle market on one side and the road traffic equilibrium market on the other side. The two markets interactively affect each other through market penetration, information quality, and spatial distribution of congestion, which are all endogenously determined in our model. In the vehicle market, two-stage decision-making is considered to describe various vehicle choices users may face in the future. In the first stage, users between each origin–destination (OD) pair choose a vehicle type between HV and CAV. In the second stage, CAV users further choose a vehicle automaton level. To account for the similarity of CAVs with different automation levels, we use a nested logit (NL) model to capture the two-stage decision-making. A vehicle choice with a higher market penetration provides higher-quality information, which affects the traffic equilibrium market. For the traffic equilibrium, route choices of users with different information quality are described by a multinomial logit (MNL) model. The spatial distribution of congestion determined by the traffic equilibrium also affects vehicle choices in the vehicle market. Specifically, users with higher quality information are more likely to choose routes with the lowest travel time. Consequently, a vehicle choice with a lower expected travel time attracts more users. The two-sided market equilibrium is formulated as a combined NL-MNL model so as to solve the two interactive market equilibria simultaneously. Then, we explore the properties of the equilibrium state. Sufficient and necessary conditions for the path flow pattern and demand pattern at equilibrium are derived, respectively. Based on the properties of the equilibrium state, we derive an equivalent variational inequality (VI) for the combined NL-MNL model. A new approach is provided to deriving two sufficient conditions, either of which guarantees the uniqueness of the VI solution. To solve the proposed problem efficiently, we develop a path-based modified self-regulated averaging (PMSRA) algorithm embedded with a modified K-shortest path method. Finally, numerical experiments are conducted to analyze the effects of CAV technology and demonstrate algorithm efficiency. Sensitivity analysis of parameters in the algorithm is also performed. Our results show that the market penetration of CAVs at the early stage of introduction is low due to the high purchase cost. With the development of CAV technology and mass production, fully automated CAVs may gradually dominate the market, while partially automated CAVs tend to be squeezed out of the market. In addition, our results reveal that the travel time saving from CAV technology and high-quality information is more pronounced for long trips and congested networks. • Consider the mixed traffic flow of CAVs at various automation levels and HVs. • A combined NL-MNL model for travelers' vehicle choices and route choices. • Interdependence of market penetration, endogenous information quality, and congestion. • Derive an equivalent VI whose solution's uniqueness is proved by a new approach. • Pronounced travel time saving of CAVs for long trips and congested networks. [ABSTRACT FROM AUTHOR]