Parking infrastructure design for repositioning autonomous vehicles

Abstract

Fully automated vehicles (AVs) have the potential to drive empty (without a passenger). For privately-owned AVs, such empty repositioning has the potential benefit of avoiding parking costs at their destination. AV owners could have their vehicle drop them off at their destination, then drive elsewhere to park. Although previous studies have considered the congestion effects of AVs repositioning to park at their owner’s residence, this study models the choice of parking location when AVs reposition away from the traveler’s destination. We model this behavior through a modified static traffic assignment with a logit model for destination choice, in which AV passenger-carrying trips can create a second empty repositioning trip to an alternate parking zone. The traffic assignment is formulated as a variational inequality. Numerical results on the Chicago sketch network show the effects of AV market penetration, fuel costs, and parking fees on the number of repositioning trips, as well as the impacts of repositioning trips on network congestion. Next, we study the problem of adjusting zone-specific parking costs to influence the repositioning behavior. In particular, when zones have asymmetric parking infrastructure costs, optimized parking fees combined with empty repositioning can encourage AVs to park at cheaper locations, thus reducing the land used for parking at zones with high land value. This network design problem is formulated as a bi-level program. Since it is bi-level and non-convex, a genetic algorithm is used to find a good solution. Results on the Sioux Falls test network show that the adjusted parking costs are effective at reducing the congestion caused by empty repositioning and encouraging more optimal parking choices for repositioning AVs.