Smart highway is an extension and innovation of the traditional highway electro-mechanical system. The normal operation of the traditional highway electro-mechanical system is mainly for guaranteeing the fast, safe and efficient operation of highway. The smart highway emphasizes the application and transmission of traffic data for the highway management and service. It adopted the fixed traffic sensors, i.e., loop detector, microwave sensor, video, and geomagnetic vehicle detector, and probe vehicles including the GPS/GSM based vehicles, to obtain traffic data on the vertical and horizontal section of the road in real time all day and night. Today, freeways in China are already deployed with certain amount of traffic sensors. However, these sensors are unable to meet the requirements for the dynamic traffic data in reliability, anti-interference, data accuracy and timeliness. In addition, traffic data obtained by these sensors is not satisfactory since the number of sensors is limited and the layout scheme is not reasonable. Therefore, this article focuses on studying the combination of multi-type sensors on the freeway to improve the accuracy of traffic data.
The development of traffic detection technologies, the methods for deploying traffic sensors and for seeking the sample size of probe vehicles have been reviewed extensively based on the detailed investigation of the related literatures. Most of the studies focused on the placement of single-type sensors, few studies explored the combination placement of multi-type sensors; most studies focused on the placement of sensors on the freeway without sensors, but few studies investigated the method of adding another type of sensors to the deployed freeway; though the combination placement of multi-type sensors has been considered in some literature, they did not explore the specific locations of these sensors. With the consideration of these factors, the object and the content of this article are determined, followed by the structure of the article.
Based on the characteristics of the fixed sensors, 15 schemes which combined two types of fixed sensors on a freeway section are proposed. For probe vehicles, this research combined the fixed sensors with the probe vehicles. 4 schemes of such kind are listed. Each scheme is analyzed from several aspects, such as detection data, advantages and disadvantages, and feasibility and so on. Based on the conditions impacting sensor working which may include traffic volume, the road physical structure, the climate of the road area, the principles of the combination placement of multi-type sensors are proposed.
Three methods, the integer programming-genetic algorithm (IP-GA), the clustering algorithm and the two-stage algorithm are applied to solve the multi-type fixed sensor deployment problem. After analyzing numerical examples, the optimal method for deploying multi-type sensors is determined, which is the IP-GA method. Based on this method, the approach of adding another type of fixed sensors on the way that deployed with fixed sensors already is discussed. A single objective optimization model was constructed to determine the locations of the add-on sensors. The genetic algorithm is used to solve the model. The analysis results show that: data estimation error can be reduced by increasing the number of sensors. However, optimizing the locations of sensors is more effective in improving data accuracy than by just increasing the number of sensors. There exists an optimal sensor number, by increasing which will not greatly improve data accuracy, but by reducing which will affect the data accuracy largely.
Based on the method of deploying single-type fixed sensors on a freeway and the method of choosing the sample size of probe vehicles, the method of combing the fixed-mobile sensors to detect traffic data is designed. For the road deployed with loop detectors, only the sample size of probe vehicles need to be determined. By using the simulation analysis method, the optimal sample size is selected. For fusing loop detector data with the probe data, the BP neural network method is adopted. The results show that under the premise of obtaining the same precision of data, the sample size of probe vehicles on the freeway with fixed detectors is about 2% lower than the one on the freeway without fixed sensors. For the freeway without fixed sensors, a method is proposed to combine these two kinds of sensors to detect traffic sensors. Firstly, the number of fixed sensors and the sample size of probe vehicles are chosen by analyzing the scheme of deploying any one type of sensors on a freeway. Numbers of schemes are combined based on these values. By comparing data errors of each scheme, the optimal combination scheme is selected.
By analyzing the data that can be collected on the freeway, this study chose the real-time monitoring data (two types of fixed sensor detection data) and the GPS location data to estimate the traffic state on a freeway. A link spilt method——the clustering algorithm——is proposed to separate the freeway into links for state estimation. By applying the adaptive weighted fusion algorithm and BP neural network algorithm, these three kinds of data are fused. The analysis results show that the BP neural network algorithm is superior to the adaptive weighted fusion algorithm. According to the highway service level standard in our country, traffic state is divided into four kinds: free-flow, unimpeded flow, general congestion, and serious congestion. By combining the speed and density together, traffic state is estimated. By using the speed data, the density data, and the combinantion of these two data separately, traffic state is estimated by referring to the highway service standard. The results indicate that only using the velocity data or the density data can result in false judgement of traffic condition.