Optimization Of Ridesourcing Services For The Future Deployment Of Connected And Autonomous Vehicles In Urban Areas

On-demand ridesourcing services have become increasingly popular due to their convenience.There are some debates claiming that ridesourcing services could increase congestion and pollution.Ridesplitting,a new shared mobility service,is a more sustainable travel mode for improving traffic efficiency and reducing air pollution.Therefore,the motivation of this study is to propose an optimization framework for the shared mobility system(SMS).The SMS ecosystem can be modeled with four specific research domains: city infrastructures,users,shared mobility system,and connected and autonomous vehicles(CAVs)technologies.Critical results and conclusions could be summarized into the following four aspects:1)City infrastructures Exploring nonlinear effects of the built environment on ridesplitting serviceWe use a machine learning method,gradient boosting decision tree model,to identify the impacts of the built environment on the ridesplitting ratio.The results show that most built environment features have a strong nonlinear effect on the ridesplitting ratio.Partial dependence plots could provide policy implications for the shared mobility system to improve the ridesplitting service.2)Users Improving the demand prediction of ridesourcing services with an Optimized Spatiotemporal Encoder-Decoder Neural NetworkWe propose a novel ridesourcing demand prediction framework.We use a combination of a graph convolutional network and a read-first long short-term memory network and a dynamic spatiotemporal attention mechanism.The experimental results show that the proposed framework outperforms state-of-the-art models for both singlestep demand prediction and multi-step prediction.3)Ridesourcing system Optimizing ridesourcing system based on a shareability network on a city levelWe optimize the ridesourcing system based on a shareability network and evaluate the gap between the potential and actual level of ridesplitting.We explore the potential of ridesplitting during peak hours using empirical ridesourcing data in Chengdu,China.The results show that the percentage of potential cost savings,time savings and shared trips could reach up to 18.47%,25.75% and 90.69%,while the actual level is 1.22%,2.38% and 7.85%.CAVs are also applicable in the proposed system.4)CAVs technologies Proposing a resilient and robust control strategy for connected and autonomous vehicles to resist the threat of cyberattacksIn anticipation of applying the previous models and solutions to CAVs,cyberattacks are an urgent problem for CAVs.We explore how cyberattacks work on CAVs.A resilient and robust control strategy(RRCS)for cyberattacks is developed and its impacts on the mixed traffic flow are explored.Finally,sensitivity analyses are conducted in different platoon compositions,vehicle distribution and cyberattack intensities.The results show that the proposed RRCS of cyberattacks is robust and can resist the negative threats of cyberattacks on the CAV platoon.The originality and innovative aspects of this dissertation could be summarized according to 2 perspectives.For the value of theory and methodology,the proposed framework for the SMS could provide a systematic methodology for the modelling and simulation.The proposed artificial intelligent algorithms could provide a better understanding for the researches of travel behaviors analysis and spatiotemporal modelling.For the value of practical application,the proposed shared mobility system could help improve ridesplitting service to build a low carbon transport,which could incorporate CAVs for the future mobility.