Centralized and decentralized autonomous dispatching strategy for dynamic autonomous taxi operation in hybrid request mode

Abstract

The combination of self-driving technology and taxis has a considerable potential in improving service quality and economic efficiency that can completely change traditional taxi operation. This study accordingly focuses on the autonomous taxi (aTaxi) dispatching problem in hybrid request mode where travelers can either request immediate rides or reserve taxi services ahead of time. In this study, a centralized dispatcher and decentralized autonomous dispatchers are designed to plan short-term and long-term routes for aTaxis in real time, respectively. The centralized dispatcher integrates vehicle-to-passenger assignment with empty vehicle rebalance to guarantee solution quality. Decentralized autonomous dispatchers distribute a portion of the calculation to aTaxis, thus reducing the centralized dispatcher’s workload. The dispatching strategy ensures that the response to traveler requests can be made immediately, and if such requests are accepted, the travelers will receive the expected services. Finally, experiments are conducted based on the actual road network and the trip data of Manhattan to investigate the dispatching strategy. The results confirm the high performance of the proposed dispatching strategies in terms of service quality and economic efficiency. It is also found that the increase in reservation rates can enhance the robustness of the method with respect to errors in predicted requests. Moreover, when the number of vehicles is adequate and reservations are made longer ahead of time, the completion rate of requests and the revenue improve.