Research On Supply And Demand Matching Strategy For The Evolution Of Ridesharing Behavior In An Open Transportation Environment

Ridesharing is considered an effective means of transportation to solve environmental,energy,and traffic congestion problems.it requires an efficient online supply and demand matching strategy,which can dynamically search and select suitable drivers and passengers.Existing work mainly focuses on the static allocation of vehicles with known travel information,usually to minimize the total travel distance or travel cost,using passenger preferences to establish a stable matching model between drivers and passengers,which can improve the stability of single matching in the short term sex and success.However,in an open-traffic travel environment,passengers may change their preferences for travel services over time or generate new preferences.For example,in different travel scenarios such as entertainment,commuting,and business,passengers will personalize time,location,and vehicle type,which may gradually generate travel service evaluation information.However,the existing short-term matching strategy usually assumes that passengers ’ preference information remains unchanged,which makes it impossible to express and use new information for model training and dynamic adjustment of strategies to achieve a dynamic balance of supply and demand.Moreover,when the matching strategy needs to use old and new information simultaneously,the existing model training methods need to re-add or adjust parameters to expand the model,which will significantly increase the time complexity and increase service response time.They will also severely limit the matching strategy in large-scale applications and the promotion of large-scale ridesharing scenarios.Because of the above problems,to improve the adaptability of the matching model to changes in passenger preference characteristics and realize personalized and long-term stable supply and demand matching in an open traffic travel environment,this paper proposes a behavior(preference)evolution-oriented ridesharing supply and demand matching optimization strategy(Behavior Extrapolation for Ridesharing Matching,referred to as BERM in this paper),the travel behavior is considered to be the choice of time and place and personal preference to meet the travel requirements.Based on the static known behavior features,the proposed matching model can describe the evolution of behavior features space and transfer the new and old features to adaptively generate the vector representation of the new features.The matching optimization strategy proposed in this paper can combine the new and old features to search the travel groups efficiently and realize dynamic online supply and demand matching.The contributions of this paper are summarized as follows:(1)According to the information of drivers and passengers,the reasonable meeting points and time constraints are determined quickly based on the R-Tree index structure.The travel network is constructed to model the spatiotemporal correlation of drivers and passengers,and the spatiotemporal vector representation of drivers and passengers is generated dynamically.(2)The static travel network model is transformed into a bipartite graph model with the characteristics of drivers and passengers and their travel behavior,and the evolution process of the behavior characteristics is described.At the same time,the multi-head attention mechanism is used to dynamically fuse the new and old feature vectors of drivers and passengers to realize the re-matching based on correlation degree and update the matching results.(3)Build a dynamic ridesharing simulation system based on the micro-traffic flow simulator to realize the random generation of order requirements and vehicle trajectories and a travel mode that multiple people can share.The matching strategy proposed in this paper is verified visually in the simulation system,and it can be obtained that the BERM model can dynamically adapt to changes in travelers’ behavior preferences.At the same time,this paper uses the open-source data set provided by the well-known platform to verify the effectiveness and accuracy of the matching strategy.The experimental results show,the precision of BERM is increased by 5.2%-7.5% on average,which proves the proposed performance of the model is better than the benchmark models.