Dynamic pricing method to maximize utilization of one-way car sharing service

Abstract

A one-way car sharing service, which allows a car to be dropped off at an arbitrary station, is highly convenient for users. However, the uneven distribution of users’ departures or destinations causes the situation where user cannot access to available car. This situation incurs a heavy loss for both users and the operational side of service. For this problem, we introduce a dynamic pricing scheme using reinforcement learning to set the charge for each station and propose a method to maximize the utilization rate by suppressing the uneven distribution of cars. The experimental results show that dynamic pricing improves the uneven distribution of cars compared with flat rates.