Analyzing the Effects of Pick-Up and Drop-Off Duration and its Stochasticity on Mobility-On-Demand Services

Mobility-on-demand (MoD) services have gathered increasing attention, and with automation reducing operational costs, their expansion is anticipated. The complex operation and user interactions inherent in MoD services present numerous challenges, highlighting the importance of thoroughly investigating their operational aspects. This paper focuses on a largely overlooked facet: the duration of pick-up and drop-off (PUDO) processes. From operator and user perspectives, we investigate how the magnitude of the average PUDO duration, its heterogeneity, and its stochasticity affect the service. Besides, we explore the potential benefits of predicting PUDO duration for each request compared to using an informed constant value, when inserting new requests into the fleet’s schedule. For this purpose, we simulated a fleet of shared autonomous vehicles operating in Munich, Germany, considering different demand levels, underlying PUDO duration distributions, and information levels by the operator. Our results reveal that modest changes in the mean duration significantly affect the performance of the service and that duration stochasticity substantially degrades the level of service (LoS), particularly for higher demand levels. For example, a duration increase from 30 to 60 s reduces served requests by 2.7%, while the presence of stochasticity increases late arrivals by 1.7%. Furthermore, we demonstrate that predicting each request’s PUDO duration or using the average duration from historical data improves the LoS. Our study highlights the need for more realistic approaches to account for the presence of stochasticity, its impact on LoS, and user behavior. We also suggest a more thorough consideration of PUDO duration in future MoD studies.