| With the increasingly development of car-hailing service in the past decade,ridesharing has been one of the popular services that allows passengers to share similar ride routes with others.It provides a convenient and economical travel service.However,there are two issues caused by the burgeoning car-hailing service.The first is the less served ratio in a ridesharing trip,which makes a low utilization rate for vehicles.The main reason is that the existing ride-matching methods don’t consider potential travel demand,which leads more missing travel requests.The second is transportation congestion due to a lot of vehicles cross the city.Although existing methods could calculate the minimum fleet size according to travel demand,these methods couldn’t work with the ridesharing service.To solve the above issues,the main research of this thesis is summarized as follows:(1)To predict the potential travel demand,an ensemble method with regional similarity and diversity enhancement named EMSD is proposed in this thesis,which predicts the travel demand for target regions based on similar regions.What’s more,the diversity of EMSD is enhanced by training base models based on subset of training data.The experiments show that the performance of EMSD is 1.1% and 2.3% higher than DEMSC model on the two datasets,respectively.(2)Based on the travel demand prediction using EMSD,this thesis proposes a ride-matching based on the prediction of travel demand called RTDP.The ride-matching of RTDP maximizes the potential travel demand along vehicle path,under the constraints of passenger waiting time,detour time,and traveling vehicle capacity.It matches multiple vehicles and multiple passengers with maximum demand for all vehicles.The experiments show that the sharing rate of RTDP is 8.5% and 7.5% higher than the baseline algorithm respectively on the data of San Francisco and Wuhan,and the served rate is 9.5% and 6.7% respectively.(3)This thesis proposes a minimum fleet planning for ridesharing named MFPR.The proposed method calculates the minimum fleet size based on a trip graph.Given the ridematching result for the passengers,the rides shared by multiple passengers are combined into one ridesharing trip,then the trip graph is generated to represent that multiple passengers can be served by the same vehicle consecutively.Based on the trip graph,the minimum fleet size is calculated by the Hopcroft-Karp algorithm.The experiments show that the minimum fleet method proposed can respectively save 50.6% and 75.7% of the vehicle on the data of Wuhan and San Francisco.In summary,the research in this thesis improves the utilization rate of vehicle,reduces the fleet size,and relief the traffic congestion.Therefore,the research in this thesis has practical significance on current car-hailing service and shared autonomous vehicles in the future. |
