Comparison of probabilistic cruising-for-parking time estimation models

Models of parking search time predominantly rely on the binomial approximation, the calculation of which is dependent solely on the occupancy of the parking network. This paper sheds light on the importance of accounting for competition between parking searchers and the effects of dynamically changing occupancy. To this end, we investigate the accuracy of two probabilistic models, namely the binomial approximation and a model proposed by Cao and Menéndez (CaM). We utilize an agent-based microscopic parking simulation model as a benchmark for comparison purposes. Our findings highlight the significance of considering the competitive dynamics among drivers in the quest for parking spaces and the fluctuations in parking occupancy rates over short time intervals. We highlight the ranges of occupancies for which the binomial approximation and the CaM model provide more reliable estimations of parking search times. The CaM model provides, in general, more accurate estimations of search times since it accounts for various parking network characteristics such as competition and varying occupancy. The binomial approximation underestimates parking search times more substantially at high occupancies with a large number of searching vehicles, whereas the CaM model performs worse for lower spacing between parking spots and lower parking occupancies. Insights from this research are then used to discuss policy implications of accounting for competition and short-term occupancy variations when designing either parking models or parking-related control strategies.