On-Demand Ride Services System Analysis And Platform Dispatching Optimization

With-the further integration of the "Internet+" strategy and the traditional transportation industry,various kinds of on-demand service platforms emerge as the times require,especially in cities such as Beijing,Shanghai,Guangzhou,Shenzhen,Hangzhou and other cities with a prominent gap between traffic demand and supply.Shared Mobility users can obtain short-term use rights of vehicles on demand without owning the vehicles.However,the related issues of Shared Mobility like system analysis and dispatching optimization,have attracted the attention of academia in recent years,and scholars' understanding about the operation mechanism of Shared Mobility is still quite limited.For example,we are not clear:the impact of Shared Mobility on the transportation system,how the perception and behavior of participants as independent decision-makers will influence their shared travel intention,and how to establish an efficient dispatching model.In view of these,this paper focuses on studying the travel behavior mechanism and its external impact,and constructs an efficient dispatching model based on Reinforcement Learning,in order to match the gap between supply and demand of the Shared Mobility system,and also help to optimize the urban traffic structure and improve the efficiency of road resources use.This paper mainly includes the following three aspects:(1)Mechanism analysis and impact assessment of Ridesharing behavior:By combining SP survey and RP survey,data sets such as sharing rate of commuting modes,number of trips,we analyze the transfer amount of other travel modes to Ridesharing modes,and study the influencing factors of choosing Ridesharing.Based on the real-world data and questionnaire survey data of Ridesharing users,this paper analyzed the behavior of Ridesharing and its impact on urban traffic system quantificationally.(2)Road network travel time reliability analysis based on Ridesourcing order data:This paper establishes a multiple traffic network travel time reliability evaluation index,considers the probability distribution of OD pair travel time rate in the road network,and uses the real-world Ridesourcing data of China's largest TNC platform for the case study.The network travel time reliability index can better evaluate regional and urban traffic conditions,and provide advice and guidance for travelers.It can be applied to different levels of users,including travelers and traffic managers,and can also help to adjust individual travel strategy and improve traffic operation efficiency.(3)On-demand ride services:systematic dispatching optimization via Monte Carlo tree search:breaking through the previous dispatching rules that only search from the perspective of space,this paper expands the set of available vehicles from two dimensions of space and time,which can improve the dispatching rate and shorten the waiting time of riders.This paper establishes a multiplier dispatching model from multiple perspectives,and establishes a dispatching tree structure according to the multiplier dispatching relationship,and then uses the improved Monte Carlo tree search algorithm to solve the model.The results show that the proposed optimization dispatching model and reinforcement learning algorithm can achieve efficient dispatching,especially in the case of unbalanced supply and demand of the TNC platform.