Dynamic Pricing in Free-Floating Carsharing Systems - A Model Predictive Control Approach

Mobility Sharing Systems (MSS) gained popularity throughout the last decade. Especially in one-way systems where users do not need to specify neither destination nor intended usage time, system operators regularly face system imbalances, where demand and vehicles are out of sync. There are two ways to encounter these situations: operator-based reallocation and user-based reallocation of vehicles. While the former utilizes paid staff, the latter tries to affect demand for vehicles to control the system towards a favorable distribution. In this paper, two approaches for user-base reallocation are presented. The myopic pricing policy only considers the current vehicle distribution as well as the demand expected in the current time-frame. The dynamic pricing policy incorporates the expected demand and the current and expected vehicle distribution within a prediction horizon. For this, a one-way MSS is modeled as Markov Decision Process featuring the vehicle distribution as states. The complexity of of the solution is reduced by applying an approximation to the state transition. Utilizing a Model Predictive Control-approach, optimal prices for the current state are determined. The impact of both proposed pricing policies is evaluated in a framework-based case study against a static pricing policy. Both dynamic policies yield an increase in profit by over 300%, while vehicle availability and average number of rentals increase by up to 8.15% and 45.8%. After all, the vehicle imbalance, measured as Root Mean Square Error, was reduced without staff- based reallocation by 11.9%.     «

Mobility Sharing Systems (MSS) gained popularity throughout the last decade. Especially in one-way systems where users do not need to specify neither destination nor intended usage time, system operators regularly face system imbalances, where demand and vehicles are out of sync. There are two ways to encounter these situations: operator-based reallocation and user-based reallocation of vehicles. While the former utilizes paid staff, the latter tries to affect demand for vehicles to control the...     »