Online Taxi-hailing Scheduling Considering Both The Interests Of Platform And Each Individual Driver

In recent years,on-demand ride-hailing platforms are developing rapidly.Optimizing the dispatching method to improve social benefits with consideration of the characteristics of ondemand ride-hailing is an important part of vehicle scheduling research.Previous researches mainly focus on the benefit of the platform,such as maximizing the number of finished requests,but seldom consider the interest of each individual driver.The benefits and the development of the platform will be affected if the interest of each individual driver cannot be guaranteed.Hence,this thesis studies the taxi-dispatching problem to improve each driver's interest while optimizing the benefit of the platform.This thesis firstly considers the different order-taking decision of the drivers,and optimizes the request allocation method to improve the scheduling effect at each dispatch time.This thesis then proposes a more long-term conducive vehicle scheduling method using future demand information.Finally,this thesis proposes a long-term conducive vehicle scheduling method considering the different service time of drivers.Firstly,this thesis studies the scheduling optimization problem considering the driver's order-taking decision.The drivers are rational and their ultimate goal is to maximize their own benefits,so the drivers do not necessarily accept the allocation requests provided by the platform.In addition,different drivers have different decision-making scheme.Previous researches on vehicle dispatching usually assume that the drivers always accept the allocation by limiting the dispatching distance.This thesis considers the driver's order-taking decision model,and proposes a method based on network flow to maximize the total value of the finished requests in the way of price adjustment mechanism.The experimental results show that our method can effectively improve the total value of the finished requests.Secondly,this thesis studies the long-term scheduling optimization problem considering forecasting demand information.Previous studies on scheduling optimization with predicted demand information usually consider the overall interests of drivers rather than the interest of each individual driver.In addition,previous methods usually obtain long-term dispatching scheme by firstly assigning orders and then scheduling empty vehicles.This thesis considers pre-matching future demands and proposes a heuristic algorithm based on minimum cost and maximum flow to maximize the total passenger-carrying time of drivers while improving the minimum passenger-carrying time of each driver.The experimental results show that our method can effectively improve the platform interest while guaranteeing the minimum interest of each individual driver.Finally,this thesis studies the long-term scheduling optimization problem considering the different service time of the drivers.Previous researches on vehicle dispatching usually assume that the service vehicles are fixed and they provide service during the whole optimization period.However,the service time of on-demand ride-hailing platform drivers is usually unfixed.Using the distribution information of the future available drivers can improve long-term vehicle dispatching.This thesis proposes a vehicle-allocation algorithm based on bipartite graph matching and an empty vehicle-dispatching algorithm based on greedy algorithm to maximize the drivers' total passenger-carrying time while improving the minimum passengercarrying time ratio of each individual driver.The experimental results show that our method can effectively improve the platform interest while guaranteeing the minimum interest of each individual driver.