Analysis of car sharing operation area performance: An idle time prediction approach

Abstract

Free-floating car sharing (FFCS) extends traditional station-based services by providing a more flexible car sharing alternative for users. However, the increased user flexibility introduces challenges from an operator perspective. To make services profitable, the total idle time of vehicles needs to be minimised and available vehicles should be located where demand exists. To increase profitability, it is important to carefully choose the operational area based on the expected idle time that different locations may offer, and only strategically expand into areas where the sustainability of the service can be maintained. In this paper, we present a hazard-based duration model for the idle times of a car sharing vehicle service. It is argued that modelling of idle time as opposed to bookings, which is the common approach, allows to circumvent the problem of latent demand and thereby presents itself as a simpler modelling strategy. In the paper, the model is applied to the city of Copenhagen, where we study the operational performance on the basis of 327,610 electric free-floating car trips in the period 2017-2018. We study the performance over 92 existing zones and predict the expected performance for an additional 28 zones by considering geographical and socio-economic drivers of demand. This enables the prediction of which areas to include as part of an expansion of the operational area, and thus serves the purpose of a strategic planning tool for growing such services. It is found that the additional zones differ substantially in their performance, which is a consequence of zones being more or less aligned with the local FFCS drivers of demand. This leads to a prioritisation of zones for further expansion based on performance, where the idle time of the best performing zones is seen to be as much as one hour less than the worst performing zones.