Research On Real-Time Taxi Ride-Sharing Problem Based On Demand-Driven

With the rapid development of social economy,the improvement of people's living standard and increasing demand for urban passenger transport,taxi is increasing popular among passengers for its fast,convenient and comfortable characteristics,and owing to the principle of supply control,the number of taxi has not increased for many years.However,it may propose the challenge that the contradiction between supply and demand of taxi traffic becomes increasingly prominent and the sustainable development of taxi traffic may be hindered as a result.As an effective method of traffic demand management(TDM),taxi ride-sharing can offer many benefits,such as improving taxi transportation capacity and efficiency,promoting the formation of a benign supply and demand relationship in the taxi market,alleviating the problems of "difficult response and high expense" in the urban cities,improving traffic congestion,alleviating environmental pollution and reducing energy waste.The paper is organized as following.First,we utilize traffic demand and supply theory,analyze influence of taxi ride-sharing on different participants,such as passengers,taxi enterprises,taxi market and propose demand-driven self-organization module and technique for taxi ride-sharing.Second,in order to describe the whole process of taxi ride-sharing,we divide real-time taxi ride-sharing problem into real-time taxi ride-sharing match problem and real-time trip assignment problem,where the sub problems are connected with a feasible trip set.Objectives for the sub problems are maximizing successful match rate and minimizing system cost respectively.Third,we consider a two-stage real-time taxi ride-sharing algorithm for the above problem.In the first stage,we generate multiple possible trips,examine possible trips using travel function and then collect a feasible trip set.In the second stage,we change the trip assignment problem into 0-1 integer programming and solve the problem using hybrid particle swarm algorithm.Finally,we carry out a series of experiments with 100 passengers and 40 taxis based on real GPS data in Beijing to demonstrate the feasibility and efficiency of the proposed models and algorithms.Moreover,we conduct sensitivity analysis with respect to significant parameters to test their influences on the taxi ride-sharing mechanism.The simulation tests show that the real-time taxi ride-sharing model can describe the taxi sharing problem in a proper way.Taken consideration of taxi ride-sharing,the total cost of the system is reduced by 10.29%,the passenger expense rate is dropped by 8.03%,the expense rate of operation is deduced by 13.48%,and the average number of passengers is increased by 1.5 times.Furthermore,the results of sensitivity analysis demonstrate that the passenger request size and arrival rate have a limited impact on the success matching rate of taxi ride-sharing ride,whereas different passenger expectations for the waiting time and delay time and the amount of delay time has a great impact on the success matching rate.