Optimization Of Sensor Location For Link Flow Inferencein Traffic Networks

Traffic flow is the basis of policy making for traffic management and planning departments.It plays an important role in numerous transportation applications,such as network performance assessment,pavement management,and traffic congestion management.As the main source of real-time traffic information,the quantity and location of traffic sensors greatly influence quality of information obtained and application effects.Therefore,how to optimize the layout of sensors to maximize the traffic information obtained with the least cost is the focus of network sensor location problem.In this thesis,on the basis study of domestic and foreign research about network sensor location problem,an optimized placement approach is proposed to obtain the traffic flow in all links on the network.Firstly,in the paper traffic information collection technology especially fixed traffic detection equipment are introduced,and a link flow inference method based on the flow conservation at the node is proposed.Based on this,two models are proposed to solve network sensor location problem.Taking into account the differences in the layout costs of different links,an optimization model with minimum cost that satisfy constraints related to complete link flow inference is established.On the other hand,based on uncertainty of sensors,a multi-objective optimization model with minimum layout cost and uncertainty in link flow estimation is set up.Secondly,considering the difference of initial condition of the road network,the proposed models are corrected to satisfy different network sensor location problems.Finally,the Nguyen-Dupuis network is used to demonstrate the applicability and adaptability of the proposed models.The results show that the sensor deployment cost and the uncertainty in link flow estimation are reduced by optimizing the layout of sensors in network.And the Sioux-Fall network is used to demonstrate the applicability of the models on the large-scale network.The models proposed in this paper are based on network topology and flow conservation at nodes instead of historical traffic information and path enumeration,which reduces the computational complexity.The proposed approach provides a new way to solve the network sensor location problem,and it can be applied to large-scale networks.It provides reference to practical engineering applications,and it can be used to collect real-time dynamic traffic flow to provide basis data support for the operation of intelligent transportation systems.