Optimization of profits in one-way free-floating car-sharing services, with a user-based relocation strategy that apply dynamic pricing and urban area demand defined gathering real vehicle-sensor data.

Rapid growing in urbanization and miles driven in the city will triple urban mobility by 2050. This explosion in demand requires switching to Mobility-as-a-Service (MaaS) models, such as Car-sharing. However, a critical issue for Car-sharing one-way free-floating services is the imbalance problem that requires to solve the conflict between the positioning of vehicles “at the right place and time” and the freedom for customers to return vehicles where and when they want. To better understand the impact of the imbalance problem, we propose to use a grid partition of the served city into zones with different demand potentials. To this aim as first step of the research real data related to vehicle positions of three Car-sharing services have been collected for approximately three months in the cities of Rome, Milan, Turin and Florence (Italy). In the experimental results data of the city of Rome have been used. This part of the research focuses on analysing user behaviour by using the number of stops in selected city zones (Stop Density) and the duration of any stop (Average Stop Duration); in fact, all the stops of each vehicle belonging to any car-sharing operator, are uniquely associated and mapped to exactly one cell of the city grid representing the Urban Areas, also tracking stop start/end time and trip start/end time. This spatial association is used to calculate Stop Density and Average Stop Duration of each urban area and to map stops to specific time-slots. Consequently, in each urban area, the Urban Area Value is calculated as a function of Stop Density and Average Stop Duration belonging to the urban area; the results of this research confirm that Urban Area Value is high where high values of Stop Density and low value of Average Stop Duration occurs. Urban Areas are ranked using the Urban Area Value calculated by considering all Car-sharing services operating in the eco-system; a spatial analysis with a thermographic map of Urban Area Value allows to visualize the existence of city zones with crucial different demand potentials. The analysis derived from such Urban Area Value and from a time-slot dynamic of the Urban Areas Values themselves, that suggested to split the standard operating day in five hourly ranges, is then used to construct a flexible and dynamic pricing mathematical programming model that has been used to derive an optimal setting of tariffs and to perform a validation phase. In this model the trip fare is defined, based on a trip planning trigger, applying a bonus/malus mechanism to a basic tariff, which considers vehicle service cost, staff relocation saving and the difference of demand value between origin and destination Urban Areas. If the user desired destination is planned in an urban area which is adjoining urban areas with higher values, alternatives with lower fees are proposed. This approach is applicable, in the reality, to several Car-sharing operators and mobility-sharing aggregators such as Urbi. The model and the outcomes of Urban Area Values have been validated in a study based on real data collected in the city of Rome (Italy) during an observation period of 49 days from April 28th to June 16th, in 2016, and where 287.975 stops observation referring to 1.271 distinct vehicles have been collected. All the stops have been observed in the city of Rome whose grid representation has been partitioned in 636 cells. These results have been presented to the 2017 COMPSAC Conference, July 7th, 2017 in the Workshop “Smart Sharing Mobility in Smart Cities” 1. These data have been used to construct an integer linear programming model where only a grid of 25 cells has been considered over the same period of 49 days. The resulting model (which has 84.500 variables and 87.750 constraints) has been solved using AMPL/CPLEX and validated by simulating a trip demand over an observed period. The result of this pricing scheme seems to produce interesting results with a business applicability in urban car–sharing market. The thesis is organized as follows. Chapter 1 is focused on the analysis of main challenges of urban mobility, and the role that car-sharing systems can play. Chapters 2, 3, 4 are devoted to the introduction and a systematic review of the literature. In Chapter 5 the data collection and cleaning are described and the final Data set is presented. Chapter 6 includes the grid partition of a city and the procedure to evaluate the Urban Area Value. Chapter 7 presents a review of the up-to-date pricing models for Car sharing that are used for defining some parameters in the optimization model presented in Chapter 8. Finally, in Chapter 9 the results obtained on the available Data set for the city of Rome are presented.