Vehicle sharing systems: Does demand forecasting yield a better service?

Although the idea of vehicle sharing systems (VSSs) emerges back in 1940s, sustaining such a system became simpler with the improvements in technology in the past decade. In addition to that, people have become more concerned about environmental effects and try to find solutions on reducing emission and energy consumption. On the other hand, VSSs require effort to make them profitable. In this paper, we focus on the two of the operational level challenges, which are the demand forecasting and routing for the rebalancing operations, in a one-way station-based bike sharing system (BSS). Since the data collection is exhaustive and costly, we would like to find the answer to whether it is worth to collect data and develop demand prediction models. In order to do that, we create a simulation of a city BSS in operation during the day. Then, using a mathematical model from the literature, we assess the rebalancing costs under two scenarios: one where we assume the perfect demand forecast, and the other where the future demand is unknown. By this way, we determine the trade-off between the lost demand and the rebalancing cost under the mentioned scenarios, and assess the benefit of forecasting the demand. Lastly, we present a case study on the Swiss BSS named PubliBike.