This thesis introduces new operational management policies for two types of on-demand shared mobility systems, the now well-studied one-way carsharing systems and a new innovative semi-autonomous last-mile transportation system. In addition, two associated simulation tools are specifically developed to evaluate the performance of the introduced policies. All research projects within this thesis have been led within partnerships with major transportation companies, i.e. car manufacturers and a public transportation operator, and address problems that they have met in real operations. The car-sharing related part of this work starts with the description of a modular event-based carsharing simulation framework. The chosen structure allows to test for multiple system configurations under various reservation and relocation policies. The simulation framework is used for the first time in a prospective study that looks at introducing one-way rentals in a station-based car-sharing system that only allowed round-trips until then. Parallel to this switch, a relocation mechanism and new rental regulations are introduced to smoothen the transition for both users and operator. Results show that the introduced relocation policy succeeds in limiting the accumulation of vehicles at stations and efficiently redistributes them over the network. The focus is then set on another type of one-way station-based carsharing systems implementing a complete journey reservation policy. A new staff-based proactive relocation policy based on Markov chain dynamics is introduced. This policy explicitely utilizes reservation information from the complete journey reservation policy to better predict the future states of the stations. It is compared to a state-of-the art relocation policy and a centralistic relocation model assuming full knowledge of the demand. The dynamic policies are applied in a real field test experiment in the city of Grenoble with full authority regarding the management of the system. Using the previously developed carsharing simulation tool, additional system configurations are analyzed and compared. Numerical results highlight the improvements obtained with the proposed proactive relocation policy. The last part of this thesis deals with a transportation system whose pattern has been rarely described in the literature: the Multi-Layered Personal Transit System (MuLPeTS). In this system, passengers travel on board autonomous trailers from their origin to their destination. No transfer is required. To cross and serve zones where autonomous driving is not yet possible, trailers have to join convoys led by human-driven vehicles. The system is described in details and positioned with respect to the existing literature. The operational problem related to the MuLPeTS is defined and solved by an approach ready to be applied in reality. An extensive simulation experiment is performed on a prospective implementation of the system with real-world data. Results provide meaningful insights on the system requirements, advantages and limitations and highlight good configurations in terms of system design and policy choices.