This thesis develops an innovative mesoscopic relocation model for free-floating Carsharing (FFCS) systems. In FFCS systems, one-way trips might cause imbalances between the spatial-temporal distributions of vehicles and demand. The model's purpose is finding vehicle relocation strategies which eliminate these imbalances in a cost-efficient or even profit-increasing way. Major emphasis is placed on creating a model which is tailored to the needs and dynamics of FFCS systems and overcomes the flaws of existing approaches. The model is described in detail and tested both within a real-world system and a simulation. The elaboration of the beginnings, the development and the impact of Carsharing (CS) provides the dissertation's theoretical background. An extensive literature review on existing solution approaches for the vehicle relocation problem in one-way Vehicle Sharing (VS) systems is also given. The lack of approaches for FFCS systems limits the focus to station-based systems. Moreover, the reasons as well as the frequency and the degree of asymmetries between vehicle supply and demand in FFCS systems are analyzed. Booking data of a system in Munich, Germany are examined spatially and temporally both on a macroscopic zone level and on a microscopic hexagon level. The combination of different analysis techniques succeeds in proving vehicle imbalances. The development of the new relocation model is the core result of this research. Some steps of the model act on macroscopic zones of the business area which are derived by solving a facility location problem based on booking data. Other steps are more detailed as they are based on microscopic hexagons. Once before the first application of the model and later on as-needed, the model input data is (re-)initialized. Booking data is analyzed and zones are categorized in historical surplus or shortage zones enabling the calculation of target vehicle distributions for different target periods. If vehicle supply and demand deviate from each other, five macroscopic or microscopic steps using optimization and rule-based methods are applied. A mixed integer program is solved for finding macroscopic zone to zone relocations which as far as possible maximize profit and lead to optimal vehicle numbers per zone. These macroscopic relocations are detailed microscopically. Based on rules which incorporate expert knowledge, individual vehicles and end hexagons are chosen for relocations. Specific vehicles are prioritized so that relocations are immediately combined with service trips like the unplugging of electric vehicles (EVs) to unblock charging stations, the recharging of EVs and the refueling of internal combustion engine vehicles (ICEVs). As vehicle imbalances also occur within zones, vehicle movements within zones are suggested analogously. Additionally, service trips which are not handled by previous steps are detected. Intra zone relocations and service trips are chosen such that they can be best combined with inter zone relocations. The last step of the model generates efficient schedules for the relocation workers such that as many relocations as possible can be conducted within a given time window. For the technical implementation of the relocation model, Microsoft Excel, ArcGIS, MATLAB, the TOMLAB Modeling Language TomSym and the TOMLAB Optimization Solver CPLEX are used. The model is also implemented into a FFCS system in Munich, Germany. Four field tests were conducted under real-world conditions within different stages of development of the model and for different degrees of model automation. Test three and four represent the final version of the model. The tests had positive impacts on the key performance indicators underlining that vehicle relocations make sense. Summing up, test three succeeded best in joining a high degree of automation, efficiency of the relocations and promising overall results like a 5.8% increase in net profit. Additionally to the real-world case study, a simulation framework for the developed relocation model is built up within this work. This enables the identification of optimal application frequencies, optimal application times and optimal parameters of the relocation algorithm. The simulation is based on booking data and captures user-based vehicle movements as well as relocation movements suggested by the relocation model. The real dynamics of the FFCS system in Munich are simulated. The simulator's results assist the operator on determining the frequency and amount of relocations. The expected relocation number can be translated into required staff size and thus into approximate costs for the relocations. The operator is able to react proactively in relocations. The process of transferring the relocation model to another FFCS system in Berlin is also described. The only effort is the initialization of the model input data. No further adjustments are necessary for the remaining steps of the model. Moreover, some model components were applied as decision support for user-based relocation strategies. Within daily operation, the fleet manager of the Munich system used the output vehicle list of some model steps for marking vehicles in unfavorable locations with reduced prices. Thereby, the fleet manager's efficiency was increased. In conclusion, the major contribution of this research lies in the development of the practice-ready relocation model with low computational time. The model fulfills the requirements arising due to changed dynamics and operation mode of FFCS systems. The uniqueness of the model lies in its mesoscopic, comprehensible structure combining macroscopic and microscopic steps. This enables the output of relocation recommendations which are detailed but not pseudo exact. The rule-based steps incorporating expert knowledge increase the model's transparency. Thus, fleet managers applying the algorithm feel more confident with respect to recommended relocations. Additionally, the rule-based steps still leave room for manual adjustments to the relocations. Therefore, the developed relocation model has a high potential of increasing the FFCS operator's acceptance for fully automated relocation models. Testing this efficient approach in real world and within a simulation has shown positive impacts on the system's key performance indicators. The relocation model contributes in improving vehicle availability and level of service, two important factors for the success and attractiveness of the new FFCS systems.
«This thesis develops an innovative mesoscopic relocation model for free-floating Carsharing (FFCS) systems. In FFCS systems, one-way trips might cause imbalances between the spatial-temporal distributions of vehicles and demand. The model's purpose is finding vehicle relocation strategies which eliminate these imbalances in a cost-efficient or even profit-increasing way. Major emphasis is placed on creating a model which is tailored to the needs and dynamics of FFCS systems and overcomes the fla...
»