Optimal Methods Of Traffic Sensors Placement On Highways

At present,most of the existing traffic sensors on the highway in China are laid at the time of the construction of the highway,which can meet the needs of road traffic data collection at that time.However,with the continuous increase of highway mileage,the rapid increase of traffic volume and the structure of the road network has become more and more complicated,a series of traffic problems have become increasingly prominent.The data obtained by the existing sensors are not ideal in terms of accuracy and timeliness,and cannot meet the needs of road traffic management.Therefore,it is necessary to optimize the placement of traffic sensors on highways.Based on the problem,this paper takes the estimation accuracy of the travel time of the road as the goal,and optimizes the placement of the traffic sensors from the two aspects of adding fixed sensors and increasing probe vehicles.Firstly,this paper reviews and summarizes the research status at home and abroad from several aspects: fixed traffic sensors layout research,travel time calculation research,and probe vehicles' sample size research.Moreover,this paper introduces common fixed traffic sensors and mobile traffic detectors.And,the paper conducts a comparative analysis from several aspects: detectable data types,data accuracy,cost,environmental adaptability.Based on the characteristics of different types of detectors,this paper proposes the principle and process of the combination of traffic sensors.Secondly,the paper divides the road in order to provide the point for the fixed traffic sensors.Considering the order between the sections of the highway,this paper proposes a method of segment length division based on unsupervised learning theory—Fisher ordered clustering,and introduces the principle and steps of the method.When applying the ordered clustering model to divide the road,considering some clustering indicators such as traffic parameters having time-distribution characteristics,so the paper defines the diameter of the class by constructing the dispersion squared function of the Frobenius norm form.This paper takes the Wuxi-Suzhou section of the Nanjing-Shanghai highway as the research object,and uses Fisher ordered clustering model to divide the road sections based on speed,traffic events,road line types and congestion data.The result shows that when the number of road segments is 30,the division result tends to be stable,so the paper divides the research road into 30 segments.Then,this paper optimizes the layout of the road sensors from adding fixed sensors and adding probe vehicles.In the research of adding fixed sensors,this paper establishes an integer programming model,which aims at the minimum estimation error of the road's travel time,adding the number of sensors and cost as constraints.The model is solved by genetic algorithm.In the research of adding probe vehicles,this paper pre-sets various proportions,and obtains the road data under different proportions through VISSIM simulation.The BP neural network fusion model is used to estimate the travel time of the road.The paper selects the optimal sample size of the probe vehicles,which is the minimum probe ratio that meets the accuracy requirements of travel time estimation.Finally,the method proposed in this paper is used to solve the optimal number and position of fixed traffic sensors,the optimal sample size of probe vehicles on the Wuxi-Suzhou section of the Nanjing-Shanghai highway.The results show that the optimal number of fixed sensors is 5,and the average absolute relative error of the estimated travel time is 0.08938.The optimal sample size of probe vehicles is 3%,and the average absolute relative error of the values is 0.01650.For highways with fixed traffic sensors,adding probe vehicles is a better strategy than adding fixed traffic sensors.