On the use of reservation-based autonomous vehicles for demand management

•Allocating road capacity to bookable autonomous vehicles can reduce congestion costs.•Users with low cooperation cost choose autonomous vehicles and experience no congestion.•User Equilibrium (UE) and socially optimal (SO) demand splits are found analytically.•The UE with optimal capacity split Pareto-dominates the case without booking and achieves a social cost reduction similar to that of the SO.•Profit-maximizing strategies are hardly compatible with social cost reductions.

Automated mobility on demand is foreseen as the future of urban passenger mobility. While mixed-traffic for autonomous and conventional vehicles could be considered, separation amplifies the benefits of automation. Combined with mobility as a service, separation also opens new possibilities in terms of demand management. We consider in this paper a single bottleneck dynamic framework, in which the capacity of a freeway is dedicated either to conventional or to autonomous vehicles. Users of conventional vehicles freely choose their departure time from home and compete for the best departure times from the bottleneck. Users of autonomous vehicles need to book their trip in advance. As the number of time slots available for booking does not exceed the capacity, booking users are guaranteed no delay at the bottleneck. An individual-specific cooperation cost is introduced in the modeling framework. We then investigate how a central planner should allocate the capacity to these two vehicle types depending on the regime (laissez-faire, welfare- or profit-maximizing). Two major findings are that the equilibrium demand split Pareto-dominates the case with only conventional cars and that the social cost difference between equilibrium and socially optimal demand splits is small compared to their benefits. Although the Pareto-improvement result may not hold for every single user in the case of richer heterogeneity, it remains a key advantage of our booking scheme. Profit-maximizing strategies however turn out to be hardly compatible with welfare maximization.