学术文献库

Parking plays a central role in transport policies and has wide-ranging consequences: While the average time spent searching for parking exceeds dozens of hours per driver every year in many Western cities, the associated cruising traffic generates major externalities, by emitting pollutants and contributing to congestion. However, the laws governing the parking search time remain opaque in many regards, which hinders any general understanding of the problem and its determinants. Here, we frame the problem of parking search in a very generic, but mathematically compact formulation which puts the focus on the role of the street network and the unequal attractiveness of parking spaces. This problem is solved in two independent ways, valid in any street network and for a wide range of drivers' behaviours. Numerically, this is done by means of a computationally efficient and versatile agent-based model. Analytically, we leverage the machinery of Statistical Physics and Graph Theory to derive a generic mean-field relation giving the parking search time as a function of the occupancy of parking spaces; an expression for the latter is obtained in the stationary regime. We show that these theoretical results are applicable in toy networks as well as in complex, realistic cases such as the large-scale street network of the city of Lyon, France. Taken as a whole, these findings clarify the parameters that directly control the search time and provide transport engineers with a quantitative grasp of the parking problem. Besides, they establish formal connections between the parking issue in realistic settings and physical problems.