Development of a user walking behavior-based positioning strategy for dockless shared transportation systems

Shared transportation systems offer innovative solutions to transportation and parking problems, especially in metropolitan areas. Micromobility solutions, which are an important component of these systems, are becoming increasingly important due to urban traffic congestion, difficulties in finding parking spaces, and challenges in accessing public transport. These systems are categorised into two main types—docked and dockless—regardless of the type of vehicle used (e-scooter, bicycle, motorcycle, car, etc.). Dockless (stationless or free-floating) models offer users unique flexibility by allowing them to pick up and drop off vehicles at any time and place. However, this flexibility also brings a number of operational challenges for operators. Chief among these is the supply-demand imbalance, which can be summarised as the inability to find the required number of vehicles in high-demand areas, while vehicles remain idle in low-demand areas. In scientific and academic studies, the primary aim is to accurately predict user demand, followed by transferring vehicles from low-demand to high-demand areas. In this rebalancing step, the number of vehicles to be placed in different areas and where these vehicles will come from are determined, taking into account the current active vehicle status in the service area. An important additional component at this stage is the placement/positioning strategy. This strategy can be defined as the spatial optimisation of candidate vehicle placement points to meet user demands. Various models are used in the literature for these studies, such as the Maximal Covering Location Problem (MCLP), Location Allocation Problem, and Facility Location Problem. The main purpose of MCLP is to identify service points that will cover the maximum possible demand within a specified service distance. In shared transportation systems, this coverage can be expressed as the distance or time users are willing to walk to access those vehicles. Studies in the literature have overlooked accurately modelling users' walking tolerance and willingness, as well as integrating another factor—slope—into the model. This study aims to determine and analyse users' walking durations before rides. Based on this analysis, walking behaviours depending on both distance and slope were modelled, and a placement model more successful than those in the literature was developed. A Walking Behaviour Based MCLP (WB-MCLP) model is proposed to cover demand points and ensure that future demands can access vehicles with minimal walking times. The developed WB-MCLP model was compared with the DT-MCLP (Distance Tolerance MCLP) model in the literature, the improved version of this model (developed in this thesis) with an added slope parameter—DT-MCLP+S (Distance and Slope Tolerance MCLP)—as well as random placement and genetic algorithm strategies. Demand points obtained from historical trip data were used to generate filtered candidate points in grid form with different sizes (250, 500, 750, and 1000 m). During candidate point generation and model testing, the repeated k-fold cross-validation method was used. The results showed that the proposed method achieved more successful outcomes than other models in many cities and regions by effectively representing slope and walking behaviour. In summary, this study developed an innovative approach to improving the efficiency of shared transportation systems by comprehensively analysing the interaction between user behaviours, environmental factors, and operational strategies. The findings proved that incorporating both distance and slope factors into the model more realistically reflects user walking behaviour, providing significant improvements in the optimisation of stationless shared e-scooter systems. This study offers valuable insights to mobility service providers, urban planners, and researchers seeking to optimise free-floating shared transportation systems.