Book section Open Access
Bike-sharing transportation systems have been well studied from a top-down viewpoint, either for an optimal conception of the sys- tem, or for a better statistical understanding of their working mecha- nisms in the aim of the optimization of the management strategy. Yet bottom-up approaches that could include behavior of users have not been well studied so far. We propose an agent-based model for the short time evolution of a bike-sharing system, with a focus on two strategical pa- rameters that are the role of the quantity of information users have on the all system and the propensity of user to walk after having dropped their bike. We implement the model in a general way so it is applicable to every system as soon as data are available in a certain format. The model of simulation is parametrized and calibrated on processed real time-series of bike movements for the system of Paris. After showing the robustness of the simulations by validating internally and externally the model, we are able to test different user-based strategies for an increase of the level of service. In particular, we show that an increase of user information can have significant impact on the homogeneity of repartition of bikes in docking stations, and, what is important for a future implementation of the strategy, that an action on only 30% of regular users is enough to obtain most of the possible amelioration.