Multi-parameter Sensing And Sensor Network Optimization For Road Traffic Information Acquisition

The technologies of road traffic information acquisition provide the important data for intelligent transportation systems (ITS), especially those technologies to detect the road network information in dots, sections or large-scale areas by using sensor network technologies. Sensor network has been an important way to acquire road traffic information of ITS, which is more and more prominent in many modern cities of the world. For the temporal and spatial correlation, the traffic information of road sections and networks will be derived from the traffic data of some key points in the road by making use of some spatio-temporal models, traffic flow models or correlation models. In addition, due to the constraints of space-time correlation, project investment and construction cost, it has been an important research direction of ITS to study the methods of traffic information acquisition by using sensor network technologies. The researches of those problems referred to theories, techniques and methods of traffic information acquisition sensor network (TIASN) have important theoretical and practical significance.Given the actual demands of realtime, accuracy and completeness for traffic information acquisition, the paper has focused on sloving those key problems which are:(1) how to obtain as much parameters of traffic information as possible using a single sensor of TIASN (micro-level);(2) how to realize the node information standardization and topology optimization of TIASN and its subnets (meso-level); and (3) how to obtain optimal large-scale deployment schemes of TIASN and sensors (macro-level). The main contributions of the dissertation are as follows:(1) In view of the sensor node of TIASN, a novel multi-parameter sensing method of traffic information by a kind of multi-function magnetic sensor (MFS) is proposed.The MFS obtains the vehicle waveform signals and some vehicle detection algorithms which can be used in different traffic scenarios are put forward. Using these algorithms, we can compute the time when a vehicle is closing to the MFS or the vehicle is departing from the MFS. According to different hardware configurations of MFS, two kind of vehicle speed estimation methods are proposed, and the estimation results have been validated by field video data. For vehicle classification, some on-line classification models and algorithms are given using the time domain or frequency domain features in order to adapt to different hardware configurations of MFS. (2) A kind of semantic code method to realize information standardization and a series of topology optimization methods to building low-power, low-cost TIASN are put forward.In view of information standardization and subnet topology optimization of TIASN, the architecture and function of TIASN are analyzed; and then the paper designs the semantic codes of TIASN which can provide the reference for traffic information sharing and standardization. Besides, by the methods of physical topology and communication topology optimization, the economic efficiency of sensor network deployment and communication efficiency of information transmission can be ensured for TIASN. The multi-knapsack model for physical topology optimization and the power-weight graph model for communication topology optimization are proposed to obtain low-power, low-cost and high-efficiency TIASN.(3) A spatial distribution model of traffic information credibility (TIC) by point sensors and corresponding TIC measures are given and some sensor location problem (SLP) models and methods by TIC measures are presented.By analyzing the spatial distribution of TIC, the thesis proposes a3D TIC spatial distribution model and defines three TIC measures for different spatial size which are sensor information credibility (ICF), road ICF and network ICF. And then an empirical study is done for calibrating ICF by field traffic information data in Beijing. Consideration of the influence factors of SLP and TIC measures, a minimum investment model (MIM) is built to meet the demands of traffic information completeness and availability. By introducing the sensor added-value factor, we establish a maximum benefit model (MBM) to obtain maximum benefit of study road. To analyze the influence pattern of SLP factors and transform the discrete models into continuous models, the simplified formulae of MIM and MBM are shown. Then, the models and their corresponding algorithms are tested and verified by numerical examples. Finally, some theoretical problems with different ICFs and road endpoint constrains are studies. The results show that the methods proposed in this paper can optimize the large-scale deployment of TIASNs in Engineering.According to actual demands of traffic information acquisition, the technologies and methods proposed in this thesis have been tested and verified by field applications or numerical examples. By given the solution schemes in engineering, the availability and validity of those methods have been done. What's more, on the basis of the current work and research findings, the paper also shows the problems of future research.