A method for monitoring traffic state of a basic section of a highway by setting mileposts

Abstract

The present application belongs to the technical field of road facility equipment layout in traffic engineering, more particularly, relates to a method for monitoring the traffic state of a basic road section of an expressway. The method comprises the following steps: firstly, randomly setting detectors at different intervals on the expressway to obtain original vehicle trajectory data; then, pre-processing the original data; and then, calculating the running speed, speed difference, headway, headway time and vehicle density of each vehicle at each time in the detection interval based on the original data. Based on the fuzzy C-means clustering traffic state clustering analysis, the interval traffic data collected in 1 min is clustered and divided into traffic states, and each category is given a label to provide a training data set for the state discrimination algorithm. The traffic state discrimination model is trained, the precision rate is selected as the machine learning evaluation index, and the traffic state discrimination model based on the random forest is selected. The determination of the detector layout interval based on the traffic state transition.

Description

A method for setting up mileage markers to monitor traffic conditions on basic sections of highways Technical Field

[0001] This invention belongs to the technical field of road facility and equipment layout in traffic engineering, and more specifically, relates to a method for setting up mileage markers to monitor the traffic status of basic sections of expressways. Background Technology

[0002] Real-time monitoring of highway traffic conditions helps traffic managers understand and control traffic conditions, enabling them to take appropriate measures to address congestion. Accurate traffic condition monitoring can proactively and reasonably provide effective travel guidance to travelers, improving the service level of the highway network and serving as an important part of the proactive traffic management system.

[0003] Traffic operation status classification standards can generally be divided into two main categories: absolute measurement standards and relative measurement standards. Absolute measurement standards refer to standards with fixed values ​​over a large range. For example, the U.S. Road Capacity Manual uses traffic volume, average travel speed, saturation, and corresponding load coefficients as comprehensive measurement indicators to classify traffic operation status into six categories (A to F). Similarly, my country's "Interim Technical Requirements for Highway Network Operation Monitoring and Service" uses average travel speed as a measurement indicator to classify traffic operation status into five categories: smooth flow, basically smooth flow, light congestion, moderate congestion, and severe congestion. These standards provide quantitative criteria for traffic status classification, enabling the identification of traffic operation status through measurement indicators. However, because traffic flow is affected by various factors such as roads, traffic conditions, weather, and time of day, and traffic flow itself has a certain degree of uncertainty, a unified absolute measurement standard cannot reflect the actual traffic status on basic road sections under different spatiotemporal conditions. Relative measurement standards do not have fixed measurement indicators; they classify traffic status based on the actual traffic flow operation of the road combined with existing standards to describe highway sections. This paper classifies the traffic status of basic highway sections into four states: smooth flow, stable flow, congested flow, and blocked flow.

[0004] The new generation of highway traffic detectors, composed of multiple sensor modules including video and radar, can detect traffic flow information and surrounding environmental information such as weather in real time. Traditional induction coil detectors can directly count traffic volume and occupancy, and then infer traffic information such as vehicle speed, vehicle length, platoon length, and vehicle type. Compared to these detectors, the new detectors can detect the trajectory of all vehicles on the road, calculating headway, distance between vehicles, speed difference between vehicles in front and behind, and vehicle density within the detector's coverage area based on vehicle position and speed. Deploying detectors along the entire road can obtain the spatiotemporal trajectories of vehicles across the entire route, enabling accurate detection of real-time traffic conditions. However, considering the deployment cost, detectors need to be deployed at certain intervals. By utilizing the transmissibility of traffic flow, the duration of traffic conditions within a section, the time of state changes, and the state transmission time between upstream and downstream sections can be analyzed to find an optimal deployment interval. Summary of the Invention

[0005] Therefore, a method for deploying mileage markers to monitor the traffic conditions of basic highway sections is needed. This method involves the rational deployment of smart mileage markers for monitoring the traffic conditions of basic highway sections, enabling the smart mileage markers to accurately monitor road traffic conditions at a lower cost.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] A method for setting up mileage markers to monitor traffic conditions on basic sections of highways includes the following steps:

[0008] S1. First, detectors are set at different intervals on the highway to obtain raw data of vehicle trajectories. Then, the raw data is preprocessed and then the raw data is processed to calculate the vehicle's running speed, speed difference, headway, headway and time distance, and vehicle density of the detection interval at each moment.

[0009] S2. Traffic state clustering analysis based on fuzzy C-means clustering: Traffic states are divided into clusters based on 1-minute interval traffic data, and labels are assigned to each category to provide a training dataset for the state discrimination algorithm. The processed feature variables are used as training samples. Let each sample point be (x1,x2,x3,x4,x5), where x1,x2,x3,x4,x5 represent interval speed, speed difference, headway, headway time distance, and vehicle density, respectively. Fuzzy clustering algorithm is used to cluster traffic states, determine the category to which each sample belongs, and assign category labels to the samples. Then, the traffic flow parameter variable characteristics of each category represent a specific traffic state.

[0010] S3. Traffic state discrimination based on random forest: Cluster analysis is performed on historical traffic data to obtain datasets of four traffic states. 60% of the data is extracted as the training set and the remaining 40% of the data is used as the test sample to train the traffic state discrimination model, thus obtaining the traffic state discrimination model based on random forest.

[0011] S4. Determining the detector deployment interval based on traffic state transition.

[0012] Further optimization of this technical solution includes the following formulas for calculating speed, speed difference, headway, headway distance, and vehicle density in the specified interval:

[0013] 1) Speed

[0014]

[0015] 2) Speed ​​difference

[0016] vdeviationi = v i+1 -v i

[0017]

[0018] 3) Headway

[0019] h i =x i -x i+1

[0020]

[0021] 4) Headway

[0022]

[0023]

[0024] 5) Vehicle density

[0025]

[0026] Within one second, vehicles on the road are separated by lanes and their positions are arranged sequentially according to their direction of travel, where i is the number of the preceding vehicle, i+1 is the number of the following vehicle, and x... i The position of the vehicle in front, x i+1 For the position of the rear vehicle, v i v is the speed of the vehicle in front. i+1Let V be the speed of the following vehicle, and V be the average speed over that time interval. deviation Let H be the average speed difference in the time interval, H be the average headway (m) in the time interval, HT be the average headway (s) in the time interval, and D be the vehicle density (veh / km) in the time interval.

[0027] This technical solution is further optimized so that the traffic states are smooth, stable, congested, and blocked.

[0028] Further optimization of this technical solution includes the following steps in the fuzzy C-means clustering algorithm process:

[0029] S21. Set initial parameters, including the number of clusters w = 4, the fuzzy weighting exponent m, and the convergence condition: the number of iterations is T. max The iteration termination threshold ε;

[0030] S22. Randomly select initial cluster centers And calculate the initial membership matrix U0;

[0031] S23. Calculate the cluster centers c = {c1, c2, c3, c4} and update the membership matrix U;

[0032] S24. Calculate the objective function, if ||c|| t+1 -c t If ||≤ε or the maximum number of iterations is reached, the calculation terminates; otherwise, return to step 3 to continue iterating.

[0033] S25. Output the sample membership matrix to obtain the traffic state dataset.

[0034] Further optimization of this technical solution involves the following steps for constructing the traffic state discrimination model based on random forest:

[0035] S31. Construct the training set for each traffic state decision tree learner. For each traffic state learner, use the bootstrap resampling method to randomly extract a portion of traffic flow data samples from the original dataset with replacement as the sample subspace of the decision tree, i.e., perform traffic flow data sampling.

[0036] S32. Construct an attribute subspace. For each node of the decision tree, first randomly select a given number of traffic flow parameter variables to form an attribute subspace, then calculate the split value of each attribute, and then use the better result as the split criterion for that node.

[0037] S33. Construct a decision tree as a traffic state learner. Construct a decision tree for each traffic flow dataset sample subspace as a corresponding decision tree. No pruning is required for each decision tree. Finally, a traffic state learner is formed.

[0038] S34. Construct a random forest algorithm by combining the decision trees established in step S33 into a random forest algorithm and testing it with test data.

[0039] S35. Decision voting: When using the random forest algorithm to determine traffic status, each traffic status learner votes on the test data, calculates all voting results, and finally uses a majority voting strategy to determine the final result of the test data.

[0040] In a further optimization of this technical solution, the steps are arranged at intervals as follows:

[0041] T = min{(t 1i -t 2i ),(t 1i -t 3i ),(t 1i -t 4i ),(t 1i -t 5i )}

[0042] T is the minimum detection time, t 1i = In interval 1, the start time of traffic state i, t 2i = In interval 2, the start time of traffic state i, t 3i = In interval 3, the start time of traffic state i, t 4i = In interval 4, the start time of traffic state i, t 5i = The start time of traffic state i in interval 5.

[0043] Unlike existing technologies, this invention studies a deployment scheme for intelligent kilometer markers used in traffic condition monitoring on basic highway sections. This invention analyzes the spatiotemporal trajectory data of vehicles detected by the novel intelligent kilometer markers, introduces algorithms such as Fuzzy Clustering (FCM) and Random Forest (RF) to establish an FCM-RF traffic condition discrimination model, and uses confusion matrix and accuracy as performance evaluation indicators. Different deployment intervals for the detectors are set to comprehensively evaluate the differences in time variation. Research shows that the FCM-RF combined algorithm can effectively identify the traffic conditions on basic highway sections, with a classification accuracy of over 97%, and can effectively determine traffic flow conditions; a deployment interval of 200m yields good results. Attached Figure Description

[0044] Figure 1 is a flowchart of the method for setting up mileage markers to monitor the traffic conditions of basic sections of expressways.

[0045] Figure 2 is a schematic diagram of the deployment of virtual detectors on highways. Detailed Implementation

[0046] To explain in detail the technical content, structural features, objectives, and effects of the technical solution, the following description is provided in conjunction with specific embodiments and accompanying drawings.

[0047] A method for setting up mileage markers to monitor traffic conditions on basic sections of highways, the method comprising the following steps:

[0048] S1. First, detectors are randomly set at different intervals on the highway to obtain raw vehicle trajectory data. Then, the raw data is preprocessed to delete empty and unreasonable values. The raw data is then processed to calculate the vehicle's running speed, speed difference, headway, headway time distance, and vehicle density at each moment in the detection interval.

[0049] 1) Speed

[0050]

[0051] 2) Speed ​​difference

[0052] vdeviation i =vi +1 -v i

[0053]

[0054] 3) Headway

[0055] h i =x i -x i+1

[0056]

[0057] 4) Headway

[0058]

[0059]

[0060] 5) Vehicle density

[0061]

[0062] Within one second, vehicles on the road are separated by lanes and their positions are arranged sequentially according to their direction of travel.

[0063] i is the number of the preceding vehicle, and i+1 is the number of the following vehicle.

[0064] x i The position of the vehicle in front, x i+1 For the position of the rear vehicle,

[0065] v i v is the speed of the vehicle in front. i+1 For the speed of the following vehicle,

[0066] V is the average velocity over that time interval.

[0067] V deviation This represents the average velocity difference over the interval at that time.

[0068] H represents the average headway (m) during that time interval.

[0069] HT represents the average headway (s) for that time interval.

[0070] D represents the vehicle density (veh / km) during that time interval.

[0071] S1. Traffic state clustering analysis based on fuzzy C-means clustering: Traffic states are clustered according to 1-minute interval traffic data, and labels are assigned to each category to provide a training dataset for the state discrimination algorithm. The processed feature variables are used as training samples, with each sample point (x1, x2, x3, x4, x5), where x1, x2, x3, x4, and x5 represent interval speed, speed difference, headway, headway time, and vehicle density, respectively. Fuzzy clustering algorithm is used to cluster traffic states, determining the category to which each sample belongs and assigning a category label. Then, the traffic flow parameter variables of each category are analyzed to represent a specific traffic state. Finally, four traffic states are obtained: smooth flow, stable flow, congested flow, and blocked flow.

[0072] The S2 fuzzy C-means algorithm primarily uses fuzzy clustering to cluster traffic states, resulting in a dataset with four traffic state categories. This provides a relative standard for classifying traffic flow data. The detailed steps of the fuzzy C-means clustering algorithm include the following.

[0073] (1) Set initial parameters. This includes the number of clusters w = 4, the fuzzy weighting exponent m, and the convergence condition: the number of iterations is T. max The iteration termination threshold ε;

[0074] (2) Randomly select initial cluster centers And calculate the initial membership matrix U0;

[0075] (3) Calculate the cluster centers c = {c1, c2, c3, c4} and update the membership matrix U;

[0076] (4) Calculate the objective function, if ||c|| t+1 -ct If ||≤ε or the maximum number of iterations is reached, the calculation terminates; otherwise, return to step 3 to continue iterating.

[0077] (5) Output the sample membership matrix to obtain the traffic state dataset.

[0078] S3. Traffic state discrimination based on random forest: Cluster analysis is performed on historical traffic data to obtain datasets of four traffic states. 60% of the data is extracted as the training set and the remaining 40% of the data is used as the test sample to train the traffic state discrimination model. Accuracy is selected as the machine learning evaluation index to obtain a traffic state discrimination model based on random forest with an accuracy greater than 97%.

[0079] Random forest algorithm is a special type of combined traffic state discrimination learner that uses k decision trees as base classifiers.

[0080] This can be achieved through ensemble learning. The algorithm first uses bootstrap resampling to extract and train multiple subsets of samples from the original sample set, and then models a decision tree for each subset. Each decision tree corresponds to a traffic state learner, with each output treated as a vote. Finally, the traffic state learners are ensembled together to determine the final traffic state through a voting process. The idea behind building a random forest is as follows:

[0081] (1) Construct the training set for each traffic state decision tree learner. For each traffic state learner, the bootstrap resampling method is used to randomly extract a portion of traffic flow data samples from the original dataset with replacement as the sample subspace of the decision tree, i.e., traffic flow data sampling is performed.

[0082] (2) Constructing the attribute subspace. For each node of the decision tree, firstly, a given number of traffic flow parameter variables are randomly selected to form the attribute subspace. Then, the splitting value of each attribute is calculated, and the better result is used as the splitting criterion for that node.

[0083] (3) Construct decision trees as traffic state learners. Each traffic flow dataset sample subspace is used as a corresponding decision tree for construction. No pruning is required for each decision tree, and finally, a traffic state learner is formed.

[0084] (4) Construct the random forest algorithm. Combine the decision trees established in step 3 into a random forest algorithm and test it with test data.

[0085] (5) Decision voting. When using the random forest algorithm to determine traffic conditions, each traffic condition learner votes on the test data, calculates all the voting results, and finally uses a majority voting strategy to determine the final result of the test data.

[0086] S4. Determination of Detector Deployment Interval Based on Traffic State Transition. Deploying detectors along the entire road segment can acquire the spatiotemporal trajectories of vehicles and accurately detect real-time traffic conditions. However, considering the deployment cost, detectors need to be deployed at certain intervals. Using traffic flow fluctuation theory, the duration of traffic states within a given interval, the time of state change, the time difference between state changes in upstream and downstream intervals, and the traffic state transfer time are analyzed. The duration of traffic states, the time of state change, and the transition time between intervals are analyzed at intervals of 300m, 500m, 700m, and 1000m. The detection time increases with the interval distance. The optimal deployment interval is reached when the detection time reaches its marginal time.

[0087] T = min{(t 1i -t 2i ),(t 1i -t 3i ),(t 1i -t 4i ),(t 1i -t 5i )}

[0088] T is the minimum detection time, t 1i = In interval 1, the start time of traffic state i, t 2i = In interval 2, the start time of traffic state i, t 3i = In interval 3, the start time of traffic state i, t 4i = In interval 4, the start time of traffic state i, t 5i = The start time of traffic state i in interval 5.

[0089] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Unless otherwise specified, an element defined by the phrase "comprising..." or "including..." does not exclude the presence of additional elements in the process, method, article, or terminal device that includes said element. Additionally, in this document, "greater than," "less than," "exceeding," etc., are understood to exclude the stated number; "above," "below," "within," etc., are understood to include the stated number.

[0090] Although the above embodiments have been described, those skilled in the art, once they understand the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the above descriptions are merely embodiments of the present invention and do not limit the scope of patent protection of the present invention. Any equivalent structural or procedural transformations made using the content of the present invention's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A method for setting up mileage markers to monitor traffic conditions on basic sections of highways, characterized in that, The steps include: S1. First, by setting up detectors at different intervals on the highway, the raw data of vehicle trajectories is obtained. Then, the raw data is preprocessed, and then the raw data is processed to calculate the vehicle's running speed, speed difference, headway, headway and time distance at each moment in the detection interval and the vehicle density in that interval. S2. Traffic state clustering analysis based on fuzzy C-means clustering: Traffic states are divided into clusters based on 1-minute interval traffic data, and labels are assigned to each category to provide a training dataset for the state discrimination algorithm. The processed feature variables are used as training samples. Let each sample point be (x1, x2, x3, x4, x5), where x1, x2, x3, x4, x5 represent interval speed, speed difference, headway, headway time distance, and vehicle density, respectively. Fuzzy clustering algorithm is used to cluster traffic states, determine the category to which each sample belongs, and assign category labels to the samples. Then, the traffic flow parameter variable characteristics of each category represent a specific traffic state. S3. Traffic state discrimination based on random forest: Cluster analysis is performed on historical traffic data to obtain datasets of four traffic states. 60% of the data is extracted as the training set, and the remaining 40% is used as the test sample to train the traffic state discrimination model, resulting in a traffic state discrimination model based on random forest. S4. Determination of detector deployment interval based on traffic state transition; The deployment interval in the above steps is as follows: For the minimum detection time, t 1i =In interval 1, the start time of traffic state i, t 2i =In interval 2, the start time of traffic state i, t 3i =In interval 3, the start time of traffic state i, t 4i =In interval 4, the start time of traffic state i, t 5i =In interval 5, the start time of traffic state i; the deployment interval = speed.

2. The method for setting up mileage markers to monitor traffic conditions on basic sections of highways as described in claim 1, characterized in that, The formulas for calculating speed, speed difference, headway, headway, and vehicle density in this section are as follows: 1) Speed 2) Speed ​​difference 3) Headway 4) Headway 5) Vehicle density Within one second, vehicles on the road are separated by lanes and their positions are arranged sequentially according to their direction of travel. The number of the vehicle in front. It's the number of the car behind. Position of the vehicle in front. For the position of the rear vehicle, The speed of the vehicle in front. For the speed of the following vehicle, The average velocity over that time interval. This represents the average velocity difference over the interval at that time. This represents the average headway (m) during that time interval. The average headway (s) for that time interval. This represents the vehicle density (veh / km) for that time interval.

3. The method for setting up mileage markers to monitor traffic conditions on basic sections of highways as described in claim 1, characterized in that, The traffic conditions are described as smooth, stable, congested, and blocked.

4. The method for setting up mileage markers to monitor traffic conditions on basic sections of highways as described in claim 1, characterized in that, The fuzzy C-means clustering algorithm includes the following steps: S21, setting initial parameters, including the number of clusters. The fuzzy weighted index is Convergence condition: The number of iterations is The iteration termination threshold ɛ; S22, randomly select initial cluster centers. And calculate the initial membership matrix. S23, Calculate cluster centers And update the membership matrix. S24. Calculate the objective function, if If the maximum number of iterations is reached, the calculation is terminated; otherwise, return to step 3 to continue iterating. S25: Output the sample membership matrix to obtain the datasets for each traffic state.

5. The method for setting up mileage markers to monitor traffic conditions on basic sections of highways as described in claim 1, characterized in that, The steps for constructing the traffic state discrimination model based on random forest are as follows: S31, construct the training set for each traffic state decision tree learner. For each traffic state learner, a portion of traffic flow data samples are randomly extracted from the original dataset with replacement using the bootstrap resampling method as the sample subspace of the decision tree, i.e., traffic flow data sampling is performed; S32, construct the attribute subspace. For each node of the decision tree, a given number of traffic flow parameter variables are randomly selected to form the attribute subspace. Then, the splitting value of each attribute is calculated, and the better result is used as the splitting criterion for the node. S33. Construct decision trees as traffic state learners. Each traffic flow dataset sample subspace is used as a corresponding decision tree for construction. No pruning is required for each decision tree, ultimately forming the traffic state learner. S34. Construct a random forest algorithm. The decision trees established in step S33 are combined into a random forest algorithm, which is then tested using test data. S35. Decision voting. When using the random forest algorithm to determine traffic states, each traffic state learner votes on the test data, and all voting results are calculated. Finally, a majority voting strategy is used to determine the final result for the test data.

Images