Study On Optimal Location Of Multiple Traffic Sensors In Complex Environment

The optimal problem of traffic sensors location is a long-term concern in the field of operations research.Locating the traffic sensor provides an effective way to obtain useful traffic information.But there is always a game between minimizing the cost of sensor location and maximizing the access to traffic information.Therefore,the optimal location of sensors in the traffic network is a research hot spot in the field of traffic network modeling in recent years.Compared with other traffic information,origin-destination(OD)demand and travel time are two very important input parameters of traffic system,which provide the main basis for urban traffic planning and management.However,there are many uncertainties in OD demand and travel time in real life due to weather,peak period and sensor failure.Moreover,the cost,accuracy and performance of different types of sensors are different.Therefore,based on different purposes of OD demands and travel time estimation in complex environment,this dissertation studies the optimal location of multiple traffic sensors,such as counting sensor,automatic vehicle identification(AVI)sensor,mobile sensor,and the combination of fixed sensor and mobile sensor.The related models and algorithms are the embodiment of operations research in cross-application of traffic planning and management.The specific work is as follows:Chapter 1 discusses the research background and significance of the traffic sensor location optimization problem,and describes the main research content of this dissertation.Chapter 2 introduces the development status of various traffic sensors in detail,and analyzes the advantages and disadvantages of different types of sensors.Then,based on the subject of this dissertation,the literature of traffic sensor location based on OD estimation and travel time estimation is sorted out,and the deficiency of current literature is pointed out,so as to determine the research direction of this dissertation.In Chapter 3,aiming at the uncertainty problem of OD demand estimation in urban road network,most existing literature only estimate the mean value of OD demand without considering the randomness of OD demand.Therefore,a counting sensor location model based on OD mean and covariance estimation is established.First,two new measures are defined to capture the maximum possible absolute error of the mean and the covariance of the estimated OD demand.Then,the bounds of these two measures are mathematically deduced,and a bi-objective optimization model is formulated to minimize the two upper bounds simultaneously.Moreover,a surrogate-assisted genetic algorithm is proposed to solve this model.Finally,the effectiveness of the model and algorithm is verified by two different-scale traffic networks.In Chapter 4,based on the OD estimation of urban road network in Chapter 3,in order to ensure the uniqueness of OD demand,some location optimization models of AVI sensor considering sensor failure and budget constraint are proposed.Unlike the traditional transition through the uniqueness of the route flow,this dissertation uses the rich traffic information collected by AVI sensors to directly realize the uniqueness of OD demand.First,an innovative AVI sensor location model is proposed to minimize the number of sensors to determine OD demand uniquely considering sensors failure and the observation sequence of AVI sensors.Moreover,under budget constraints,a sensor location model is developed to estimate OD demand,which maximizes the information value of differentiating OD pairs under sensor failures.We also design several greedy heuristic algorithms to solve the two sensor location models.Three numerical experiments show that the proposed models and algorithms can effectively determine the AVI sensor locations to recognize the OD demand and its uniqueness in the event of uncertain sensor failures.In Chapter 5,the application background of sensor location is extended from the OD estimation in urban road network in Chapter 3 and Chapter 4 to the travel time estimation on freeway.In relation to the stochasticity issue of travel time on freeway,the fixed sensors used in the existing literature can not guarantee the timeliness of data collection,so this dissertation presents a dynamic location model of mobile sensor.Firstly,under the assumption that there are no traffic sensors on freeway,the whole freeway and observation time are discretized,and a bi-objective optimization model of the mobile sensor location is proposed,in which the estimation error of travel time is minimized and the observed traffic flow is maximized.Secondly,on the premise that fixed sensors have been located on freeway,mobile sensors are dynamically added.For the data fusion of different types of sensors,a dynamic adaptive BP neural network group algorithm(DBPG)is presented.By introducing dynamic proportion factors,the update of each BP neural network is affected by both itself and the optimal network.Moreover,the contribution of each network for group study is constantly adjusted.Then,using the fusion data of multi-source sensor based on DBPG,a combination location model of mobile and fixed sensor is established.In addition,a simulated annealing algorithm is designed to solve the two location optimization models.The experimental results show that in most cases,the mobile sensor is superior to the fixed sensor,but the combination of two is better.