Dynamic fleet redistribution in carsharing

Abstract

The paper deals with the problem of dynamic redistribution of unused cars in order to achieve a harmonization of supply and demand in carsharing service. The optimization criterion is minimization of losses from demand dissatisfaction with minimal costs for operating vehicles. The paper obtained analytical results regarding the policy of optimal movement of unused vehicles. The proposed solution structure can be applied to practical problems due to its computational efficiency, as well as the ability to handle time-dependent requirements. The initial formulation of the problem is formulated in terms of stochastic dynamic programming. The first stage considers a simplified two-stage model of distributed robust optimization (DRO), which serves as an illustration of our approach. The following demonstrates a computationally efficient multi-stage model with an extended linear decision rule (ELDR). The result obtained in computational experiments shows that the scalable ELDR numerical solution is close to the optimal solution of the dynamic programming problem.