Sizing Finite-Population Vehicle Pools

We refer to a vehicle pool as a number of vehicles at a single location used for the same purpose. We focus on the problem of sizing vehicle pools for a finite set of subscribers who can use the pool. Our goal is to minimize the number of vehicles in the pool while still meeting nearly all subscriber requests. Formally, we propose three analytical techniques to size a vehicle pool for a finite population of subscribers, according to the pools´ busy period demand to guarantee all requests are served with probability 1 - ε, i.e., a quality-of-service (QOS) guarantee. Moreover, we propose an additional heuristic sizing method, which requires no prior data about pool demand. Although this method does not provide probabilistic bounds on QOS, we show in practice that it still achieves a high QOS. We evaluate our sizing methodologies using seven years of data from a local car share, using three performance metrics: availability (percentage of requests served), utilization (the percentage of time that vehicles in the pool are used) and member-to-vehicle ratio (the size of the pool relative to the size of its user population). We show that our methods perform well with respect to these metrics.