Optimizing Curbside Parking Resources Subject to Congestion Constraints

Abstract

To gain theoretical insight into the relationship between parking scarcity and congestion, we describe block-faces of curbside parking as a network of queues. Due to the nature of this network, canonical queueing network results are not available to us. We present a new kind of queueing network subject to customer rejection due to the lack of available servers. We provide conditions for such networks to be stable, a computationally tractable "single node" view of such a network, and show that maximizing the occupancy through price control of such queues, and subject to constraints on the allowable congestion between queues searching for an available server, is a convex optimization problem. We demonstrate an application of this method in the Mission District of San Francisco; our results suggest congestion due to drivers searching for parking stems from an inefficient spatial utilization of parking resources.