A Stochastic Optimization Model for Managing Energy Storage Using a Driverless Fleet of Electric Vehicles

Abstract

As driverless technology becomes more effective, affordable, and accessible, the transportation industry is due for a major change. In the near future, it is quite possible that ridesharing and taxi companies will operate driverless fleets of electric vehicles in order to meet their customers' demands. In this paper, we examine the viability of executing energy storage optimization strategies using the idle vehicles of a driverless fleet in order to create additional profit. We do so by first producing a mathematical model for this optimization scenario. We treat the idle car batteries as a collective entity, adjusting our model to account for the changes in stored energy and storage capacity that occur as vehicles become idle or are called into action. We then employ a backward approximate dynamic programming (ADP) algorithm to generate a value function approximation (VFA) based policy. We apply this policy to the modeled problem in various simulations, comparing its performance to a simple policy function approximation (PFA). The results from these simulations demonstrate that this energy storage optimization technique is a viable way for a fleet operator to noticeably increase the profitability of their fleet.