This research concerns a routing and dispatching problem in the context of the on-demand Personal Rapid Transit transportation system. The objective of the problem is to efficiently optimize the total empty movement related to the vehicles as well as the waiting time of passengers in a dynamic settings. To achieve this, a multi-agent model is developed to manage efficiently the Personal Rapid Transit system. The proposed model involves three types of agent as well as the implementation of an efficient decentralized management strategy in order to handle demand of the passengers. Tests are made on a realistic use’case and results reveal the viability of the proposed model and strategy.