Dynamic Airport Taxi-sharing Problem Under The Uncertainty Of Passengers’ Arrival Time

In recent years,online taxi-sharing services are favored by more and more travelers under the supports of mobile internet,and mobile payment technologies.Such services arrange a single vehicle to deliver multiple passengers with similar trips.Passengers in the same vehicle share the travel cost and enjoy better travel experience than that in metro and bus.At the same time,drivers obtain more incomes when the idle seats in their cars are fully utilized.In addition,such traveling mode can effectively improve the efficiency of resources usage,relieve traffic congestion,and reduce pollution.As an important transportation hub of a city,airports are usually located at remote places with large flow of passengers,which offer the most important application scenarios of online taxi-sharing services.In China,serval online travel agents have launched online taxi-sharing platforms,such as "Superbus" of Ctrip,the "Tiehang hotline" of Sichuan Airlines,especially for passengers travel demands in many cities’ airports.To further improve the operations of airport taxi-sharing platforms,and customer satisfaction,a dynamic matching strategy and route planning algorithm for airport carpooling services need to be researched.Based on the analysis of operational data of an airport taxi-sharing platform in Chengdu Shuangliu International Airport(CTU),this thesis proposes a dynamic prospective matching(PM)policy to consolidate passengers together with considering the uncertain arrival times of passengers.Accordingly,a near-optimal deterministic model is formulated for the problem under the PM policy.An improved Differential Evolution(DE)algorithm is developed to resolve the problem efficiently.The main work and contributions of this thesis are summarized as follows.(1)A PM policy is designed to improve the performance of the passengers’ consolidation,which makes decision considering the already arrived passengers as well as the coming passengers with random arrival times.(2)A two-stage stochastic programming model is formulated for the matching-and-vehicle-routing integration problem under the proposed PM policy.A Bayesian updating method is used to estimate the passengers’ arrival times and then simplifies the problem to a near-optimal deterministic model.(3)An improved DE algorithm is proposed for the vehicle routing planning subproblem.(4)We establish a simulation experiment bases on real data collected from platform supplying online taxisharing services in CTU airport of Chengdu city.The experimental results validate the effectiveness and efficiency of the proposed PM policy,model,and algorithm.The results of this study can not only used for the airport online taxi-sharing problem,but also be applied to express delivery,food,retail distribution and other fields.