Demand-responsive transportation systems diversify local transport supply with providing flexibility of routing and scheduling as well as affordable price. The classical Dial-a-ride problem enables to assign users to vehicles by minimization of a cost function under constraints. However, the projected itinerary of the vehicles is calculated on a global approach considering automatically each new users independently and a dispatcher tests a match between the user and all available service vehicles. Thus, matching calculation time is restricting for the large operational services. The objective of this paper is a reformulation of the problem with more flexibility and an aggregated solution to optimize calculation time. This modelling is based on a mobility system of transit vehicles with their own dispatcher and station infrastructures. The vehicles are shared dynamically and make extra-stops to collect travelers with various destinations. In addition, the users of the service have not any transfer and are transported directly to their destination stations. We constructed an agent-based model which maximizes the utility of each vehicle while satisfying its own constraints. The minimum travel and waiting times of the users and maximum occupancy of vehicles are the default ones. We propose then a new formulation of the problem on the maximization of the vehicle utility. Passengers’ assignment to vehicles is controlled directly by the vehicle drivers on an aggregated level by origin destination pair. A case study illustrates the theoretical model and compares the efficiency of several service choices. A fleet of small vehicles seems to minimize the user waiting time.