Research And Verification Of Data-driven Route Planning On Multi-modal Transportation Network

With the rapid development of urban economy,increasing concerns are being raised about serious traffic problems nowadays.Giving priority to the development of public transportation is considered to be the key to alleviate traffic congestion.The Advanced Traveler Information System(ATIS)provides travelers with public transportation information and multi-modal trans-port route planning to encourage users to use public transportation,thus to alleviate congestion and air pollution,as well as the quality of individual life.Bike sharing systems(BSS),with the advantages of environmental friendly,energy saving,high accessibility,autonomy and flexibility,provide short-distance travel solutions for the city,facilitating passengers to transfer to other trans-port modes.As an alternative way to avoid congested roads,BSS highly improves the operational efficiency of the bus network and alleviates road congestion.However,existing works haven’t fully incorporated bike sharing systems within ATIS,and the uncertainty of traffic conditions and multi-modal routing makes it challenging to accurately estimate the travel time.Based on massive public transportation historical data,this thesis uses data analysis and machine learning techniques to mine dynamic parameters in public transportation networks to characterize traffic uncertainty.A stochastic multi-modal public transportation network including public bikes is constructed and an optimal path recommendation is provided to the user based on the multi-criteria path planning algorithm to reduce user’s travel time and improve the trip reliability.The specific contributions are as follows:Firstly the thesis proposed a general framework for modeling public traffic uncertainties aim-ing at the traffic stochastic problem.K-means clustering algorithm and statistical method are used to obtain its parameters,and the stock distribution,availability and returnability of bike sharing system stations are considered.Secondly,a directed multi-edge graph is introduced combining time-expanded network and time-dependent networks to construct a stochastic,time-event-dependent transport network.It ex-tends the traditional multi-modal network by integrating the public bicycle service into the tradi-tional network,and simplifies the network complexity.Finally,a multi-criteria real-time route planning algorithm is proposed,which can solve the aggregation problem of travel time between different modes of transportation,and optimize the travel time,reliability and transfer times.We evaluated the system based on the massive historical data of London,and the results show that the system in this thesis is better than the common path planning system in terms of travel time,travel reliability and travel time prediction accuracy.