A method for predicting highway congestion dissipation time that considers traffic event information

Abstract

This invention discloses a method for predicting highway congestion dissipation time that considers traffic event information. It utilizes detailed information from detected abnormal events to learn a decision tree model, thereby constructing a decision tree model that can predict the duration of the event. Simultaneously, it uses event reporting information to construct a capacity estimation model. Based on this, a congestion dissipation time prediction model based on a fuzzy traffic wave model is built. Combining the outputs of the above two parts with the fuzzy traffic wave model, a final congestion dissipation time prediction model for the event scenario is obtained. This invention effectively overcomes the limitations of existing congestion dissipation time prediction methods that do not adequately consider the specific circumstances of abnormal events. Furthermore, the use of a fuzzy traffic wave model fully considers the speed distribution differences during vehicle queuing, which are not adequately taken into account by traditional traffic wave models, thus achieving higher prediction accuracy and generalization ability for congestion dissipation time.

Description

Technical Field

[0001] This invention belongs to the field of intelligent transportation information technology, and specifically relates to a method for predicting the dissipation time of highway congestion that takes into account traffic event information. Background Technology

[0002] With the construction of intercity expressways and the increase in the number of vehicles owned by households, the prediction of the time it takes for traffic congestion to dissipate on expressways under abnormal events is particularly important, as it can provide technical support for subsequent expressway management.

[0003] Existing traffic prediction methods can predict road traffic flow information, analyze traffic conditions, and provide the predicted road traffic status to road navigation algorithms and traffic management authorities for corresponding guidance and control. In the event of abnormal events on highways, accurate predictions of highway congestion dissipation times can, on the one hand, provide traffic management departments and travelers with relevant road congestion information to reduce subsequent vehicles entering congested sections and alleviate traffic pressure on congested highways; on the other hand, it can save travelers' travel time and improve their travel experience.

[0004] Although many methods for predicting congestion dissipation time have been proposed, these methods still have some limitations. First, traditional methods are mainly based on historical data and statistical models, without considering the circumstances of events. This means that after an abnormal event occurs, methods using historical data and statistical models cannot adequately account for the differences between different types of events.

[0005] For example, existing patent CN117012031A discloses a method for predicting highway congestion dissipation time based on deep meta-learning. This method develops a congestion dissipation time prediction approach based on the concept of deep meta-learning, but it does not adequately consider information regarding abnormal events on highways. Furthermore, because it requires the use of vehicle trajectory data output by dense highway detection equipment, the distribution of such equipment on conventional highways does not meet the requirements of this patent. Additionally, the manual reporting of specific event information after an abnormal event occurs is not effectively utilized. Existing technologies do not adequately consider the differences between various abnormal events and are not well-suited for use on road sections with poorly distributed detection equipment.

[0006] Research on congestion dissipation time prediction that considers specific event circumstances is of great research and application value. How to accurately predict highway congestion dissipation time is a technical problem that those skilled in the art need to continuously solve and optimize. Summary of the Invention

[0007] In view of this, the purpose of this invention is to provide a method for predicting highway congestion dissipation time that considers traffic event information. This invention aims to solve the problem that existing methods cannot adequately consider the different impacts of event types on congestion dissipation time after an abnormal event occurs, resulting in low accuracy in congestion dissipation time prediction.

[0008] To achieve the above objectives, this invention provides a method for predicting highway congestion dissipation time that considers traffic event information, comprising the following steps:

[0009] S1. Collect abnormal event data on highways and extract the characteristic relationship between traffic events and traffic congestion;

[0010] S2. Collect upstream and downstream ETC data and detector data corresponding to abnormal events on highways to construct a traffic flow dataset;

[0011] S3. Based on the characteristic relationship between traffic events and traffic congestion obtained in step S1, construct an event duration prediction model;

[0012] S4. Based on the event duration prediction model constructed in step S3, construct a congestion dissipation time prediction model based on the fuzzy traffic wave model.

[0013] S5. Perform machine learning on the event duration prediction model constructed in step S3, and then input the output of the machine learning event duration prediction model into the congestion dissipation time prediction model based on the traffic wave model constructed in step S4, so as to obtain the prediction result of congestion dissipation time.

[0014] Furthermore, step S1 includes the following sub-steps:

[0015] S1.1 Collects information on abnormal events on the highway;

[0016] The abnormal event information of the highway includes the event type, the vehicles involved, the road section where the event occurred, the time of the event, the impact of the event on traffic, and the duration of the event.

[0017] S1.2 Preprocess the data information collected in step S1.1, that is, clean the collected data information;

[0018] S1.3 Perform hot encoding on the preprocessed data information from step S1.2 to convert the text information into encoded information.

[0019] Furthermore, step S2 includes the following sub-steps:

[0020] S2.1 Collect upstream and downstream ETC data and detector data corresponding to abnormal events on highways, and extract traffic flow characteristic information of the road segments corresponding to abnormal events on highways;

[0021] S2.2 Perform data preprocessing on the traffic flow feature information collected and extracted in step S2.1, that is, clean the extracted data information and remove outliers;

[0022] S2.3 Based on the data information preprocessed in step S2.2, extract vehicle and traffic flow feature data;

[0023] S2.4 Match the traffic flow feature data extracted in step S2.3 with the abnormal event information to create a traffic flow dataset under abnormal events.

[0024] Furthermore, step S3 includes the following sub-steps:

[0025] S3.1 Based on the main influencing factors of the event on the duration of traffic congestion obtained in step S1, construct an evaluation system for the severity of the event, and then evaluate the traffic events on the highway.

[0026] S3.2 Based on highway traffic incident evaluation indicators, construct an event duration prediction model based on a decision tree model;

[0027] S3.3 Input the dataset after hot encoding of highway abnormal event information in step S1.3 into the time duration prediction model based on decision tree to train the model and obtain the highway abnormal event duration prediction model after training.

[0028] The input to the highway abnormal event duration prediction model is highway abnormal event information, and the output is the predicted duration of the event.

[0029] Furthermore, step S4 includes the following sub-steps:

[0030] S4.1 Construct a queue length prediction model based on a fuzzy traffic wave model;

[0031]

[0032] In the formula, Let E0 be the average equivalent queue length in the multi-lane road event segment at time t; E0 is the standard number of vehicles between segments A and B at the initial time; A x (i,t) represents the number of vehicles joining the event segment at time t in the i-th lane participating in the queue, as determined using fuzzy sets; A xl (i,t) represents the number of vehicles leaving the queuing lane at time t, determined using fuzzy sets; A xn (i,t) represents the number of vehicles in the i-th lane that has not joined the queue at time t, determined using fuzzy sets; M represents the total number of lanes in the road. The average congestion density of vehicles on the road; The optimal vehicle density for traffic on the road; L is the length between road segments A and B.

[0033] S4.2 Construct a road capacity estimation model under abnormal events;

[0034] C s =C0·f n ·f HV ·f L

[0035] In the formula, C s C0 represents the regional traffic capacity of the expressway under abnormal traffic events; f represents the basic traffic capacity of the expressway. n f is the correction factor for the number of closed lanes on the highway segment's capacity; HV f is the correction factor for the capacity of highway sections relative to the proportion of large vehicles; L This is the correction factor for the length of the closed driving lane on the highway's traffic capacity;

[0036] S4.3 The traffic flow dataset collected in step S2 and the output of the highway abnormal event duration prediction model constructed in step S3.3 are used as inputs and fed into the queue length prediction model constructed in step S4.1 to obtain relevant information on highway congestion under abnormal events.

[0037] Furthermore, in step S4.2, the relevant information on highway congestion under abnormal events includes the maximum number of vehicles in the queue and the congestion dissipation time.

[0038] Furthermore, the formula for calculating the congestion dissipation time is as follows:

[0039]

[0040] In the formula, N U (i,t) represents the number of vehicles joining the event segment in lane i at time t; t0 represents the output of the event duration prediction model; N D (i,t) represents the number of vehicles leaving the event segment at time t in the i-th lane; C s (t) represents the real-time road capacity.

[0041] Beneficial effects:

[0042] This invention proposes a method for predicting highway congestion dissipation time that considers traffic event information. It fully utilizes detailed information on abnormal events reported manually or detected by event detectors to learn a decision tree model, thereby constructing a decision tree model that can predict event duration. Simultaneously, it uses event reporting information to construct a capacity estimation model, assessing the road capacity under the event by utilizing the event's impact on road traffic. Based on this, a congestion dissipation time prediction model based on a fuzzy traffic wave model is constructed. Combining the outputs of the above two parts and the fuzzy traffic wave model, a final congestion dissipation time prediction model for the event is obtained. This invention effectively overcomes the limitations of existing congestion dissipation time prediction methods that do not adequately consider the specific circumstances of abnormal events. Furthermore, the use of the fuzzy traffic wave model fully considers the speed distribution differences during vehicle queuing, which are not adequately considered when using traditional traffic wave models, thus achieving higher prediction accuracy and generalization ability for congestion dissipation time.

[0043] Other advantages, objectives, and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination, or may be learned from practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description

[0044] Figure 1 This is a flowchart of a method for predicting highway congestion dissipation time that takes into account traffic event information, according to the present invention.

[0045] Figure 2 The flowchart shows the overall training process for the congestion dissipation time prediction model. Detailed Implementation

[0046] To make the technical solutions, advantages, and objectives of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the described embodiments of the present invention without creative effort are within the protection scope of this application.

[0047] like Figure 1 As shown, this invention provides a method for predicting highway congestion dissipation time that considers traffic event information, comprising the following steps:

[0048] S1. Collect abnormal event data on highways and extract the characteristic relationship between traffic events and traffic congestion;

[0049] S1.1 Collect information on abnormal events on highways, including event type, participating vehicles, road segment where the event occurred, time of the event, impact on traffic, and duration of the event;

[0050] S1.2 Preprocess the data information collected in step S1.1, that is, clean the collected data information;

[0051] S1.3 Perform hot encoding on the preprocessed data information from step S1.2 to convert text information into encoded information, in preparation for subsequent machine learning.

[0052] The encoding conversion results for each data item are as follows:

[0053] Event type: ["Disabled vehicle blocking lane", "Road traffic incident", "Road maintenance work"] = [100,010,001]

[0054] Participating vehicle information: ["Only n small cars participated", "There were also large cars participating, a total of n"] = [0n, 1n]

[0055] The incident occurred on the following road section: ["Regular road section", "Inside the tunnel", "Steep slope section"] = [100,010,001]

[0056] Event time period: ["Daytime off-peak period", "Nighttime off-peak period", "Peak period"] = [100,010,001]

[0057] S2. Collect upstream and downstream ETC data and detector data corresponding to abnormal events on highways to construct a traffic flow dataset;

[0058] S2.1 Collect upstream and downstream ETC data and detector data corresponding to abnormal events on highways, such as vehicle passage information detected by ETC detectors, in order to extract traffic flow characteristic information of the road segment corresponding to the abnormal event on the highway.

[0059] S2.2 Perform data preprocessing on the traffic flow feature information collected and extracted in step S2.1, that is, clean the extracted data information and remove outliers;

[0060] S2.3 Based on the data information preprocessed in step S2.2, extract vehicle and traffic flow feature data;

[0061] S2.4 Match the traffic flow feature data extracted in step S2.3 with the abnormal event information to create a traffic flow dataset under abnormal events.

[0062] The formula for calculating the road traffic density corresponding to ETC is as follows:

[0063]

[0064] In the formula, k is the traffic density of the road segment; Q A (t) represents the traffic flow at upstream gantry A during time t; Q B (t) represents the traffic flow through downstream gantry B in time t; L is the length between road segments A and B; E0 is the standard number of vehicles between road segments A and B at the initial time, calculated using the following formula:

[0065]

[0066] In the formula, E0 is the standard number of vehicles between road segments A and B at the initial time; For upstream gantry A in time The number of standard vehicles that pass through vehicle X during the time period between gantries A and B;

[0067] S3. Based on the characteristic relationship between traffic events and traffic congestion obtained in step S1, construct an event duration prediction model;

[0068] S3.1 Based on the main influencing factors of the event on the duration of traffic congestion obtained in step S1, construct an evaluation system for the severity of the event, and then evaluate the traffic events on the highway.

[0069] The evaluation indicators are as follows:

[0070] Event severity: ["Severe", "Ordinary", "Slight"] = [100,010,001]

[0071] The severity of an event is encoded and converted according to the above encoding format using the actual event duration, and the duration of subsequent events is predicted based on this.

[0072] S3.2 Based on highway traffic incident evaluation indicators, construct an event duration prediction model based on a decision tree model;

[0073] The steps involved in using a decision tree model to predict the duration of an event are as follows:

[0074] I. The processed training data collected in S3.1 will be used as the model input for the decision tree to train the nonlinear decision tree regression model and generate the final model used to predict the duration of the event;

[0075] II. The decision tree regression model recursively divides each region into two sub-regions in the input space where the training dataset is located, and makes decisions on the output of each sub-region, thus constructing a binary decision tree;

[0076] III. Steps for generating a decision tree:

[0077] (1) Select the optimal splitting variable j and splitting point s, where the calculation formula is as follows:

[0078]

[0079] By traversing variable j, scanning the splitting point s for the fixed splitting variable j, and selecting the pair (j,s) that minimizes the target of the above formula;

[0080] (2) Use (j,s) obtained above to divide the region and determine the corresponding output value;

[0081]

[0082] In the formula, R1(j,s)={x|x (j) ≤s},R2(j,s)={x|x (j) >s};

[0083] (3) Continue to call (1) and (2) the above calls to the two sub-regions until the stopping condition is met;

[0084] (4) Divide the input control into M regions, R1, R2, ..., R m Generate a decision tree;

[0085] (5)

[0086] In the formula, I is the indicator function.

[0087] S3.3 Input the dataset after hot encoding of highway abnormal event information in step S1.3 into the time duration prediction model based on decision tree to train the model and obtain the highway abnormal event duration prediction model after training.

[0088] The input to the highway abnormal event duration prediction model is highway abnormal event information, including: the location of the abnormal event, the vehicles involved in the abnormal event, the time of the abnormal event, the upstream and downstream traffic flow at the time of the abnormal event, and the impact of the abnormal event on traffic, etc. The model output is the predicted duration of the event.

[0089] S4. Based on the event duration prediction model constructed in step S3, construct a congestion dissipation time prediction model based on the fuzzy traffic wave model.

[0090] S4.1 Constructing a queue length prediction model based on a fuzzy traffic wave model

[0091] Determining membership functions is a key issue in fuzzy theory. When applying fuzzy theory to solve practical problems, the primary challenge is establishing membership functions that accurately reflect reality. Therefore, the first step here should be to establish the fuzzy set membership functions for vehicles belonging to abnormal traffic conditions.

[0092] The first part defines the fuzzy subsets. This method uses an assignment method to determine the membership function of the fuzzy subsets. The assignment method generally sets the membership function according to three directions: small, medium, and large. Here, the number of fuzzy subsets is set to three, namely x1, x2, and x3, which correspond to the small, medium, and large fuzzy sets, respectively. In this method, they represent the sets of vehicles that have not entered the queue, vehicles in the queue, and vehicles that have left the queue, respectively. The specific expression of the fuzzy set is shown below:

[0093]

[0094]

[0095]

[0096] In the formula, x n For a fuzzy subset; u n For subintervals; f nn For u n For fuzzy set A n The degree of subordination.

[0097] The second part involves determining the membership function of the fuzzy subset. The general method of assignment used in this approach is as follows:

[0098] First, it is assumed that the speed distribution of queuing vehicles during congestion follows a normal distribution random process. Therefore, the normal distribution membership function formula is also used to determine the membership function. Based on the normal distribution membership function formula, the specific membership degree value of each normally distributed random number to the three fuzzy sets is determined, as shown in the following formula:

[0099] Vehicles that did not enter the queue:

[0100] Vehicles in queue:

[0101] Vehicles that have left the queue:

[0102] In the formula, A(x) is the membership function of the normal fuzzy distribution; a is a constant; k is a fixed value that needs to be calibrated later.

[0103] After data calibration, the upper and lower bound fuzzy sets and their membership functions are determined based on the specific k value. First, the upper and lower bound fuzzy sets are defined as follows:

[0104]

[0105] In the formula, Let x be the mean of the fuzzy set. i It is a fuzzy subset of the domain.

[0106]

[0107] In the formula, S 2 For fuzzy set variance; x i It is a fuzzy subset of the domain.

[0108] After establishing the fuzzy set indicating whether vehicles are queuing, and inputting the fuzzy parameters into the traffic wave model, this method selects the MAEQL queue length model as the improved base model, and its calculation expression is as follows:

[0109]

[0110] In the formula, N represents the average equivalent queue length in the multi-lane road event segment at time t; U (i,t) represents the number of vehicles joining the event segment in lane i at time t; N D (i,t) represents the number of vehicles leaving the event segment at time t in the i-th lane; M represents the total number of lanes in the road. The average congestion density of vehicles on the road; This represents the optimal vehicle density for traffic on the road. The model's applicability is... Let be the average single-lane vehicle density between the road segments at time t;

[0111] The fuzzy traffic parameters are input into the MAEQL model, resulting in the following calculation expression for the improved model:

[0112]

[0113] In the formula, A x (i,t), A xl (i,t), A xn (i,t) represent the number of vehicles joining the event segment at time t in the i-th lane participating in the queue, as determined using fuzzy sets; A xl (i,t) represents the number of vehicles leaving the queuing lane at time t, determined using fuzzy sets; A xn (i,t) represents the number of vehicles in the i-th lane that has not joined the queue at time t, determined using fuzzy sets.

[0114] S4.2 Constructing a Road Capacity Estimation Model under Abnormal Events

[0115] The formula for calculating the impact of the event on traffic is as follows:

[0116] C s =C0·f n ·f HV ·f L

[0117] In the formula, C s C0 represents the regional traffic capacity of the expressway under abnormal traffic events; f represents the basic traffic capacity of the expressway. n f is the correction factor for the number of closed lanes on the highway segment's capacity; HV f is the correction factor for the capacity of highway sections relative to the proportion of large vehicles; L This is the correction factor for the length of the closed driving lane on the highway's traffic capacity;

[0118] The highway event information obtained from hot coding in S1.3 is input into the formula, and the real-time road capacity C is obtained according to the processing status of real-time events. s (t).

[0119] S4.3 The traffic flow dataset collected in step S2 and the output of the highway abnormal event duration prediction model constructed in step S3.3 are used as inputs and fed into the queue length prediction model constructed in step S4.1 to obtain relevant information on highway congestion under abnormal events, such as the maximum number of vehicles in the queue and the congestion dissipation time.

[0120] The output calculation formulas are as follows:

[0121] Maximum queue length:

[0122] Congestion dissipation time:

[0123] In the formula, t0 is the output of the event duration prediction model; C s (t) represents the real-time road capacity.

[0124] S5. Perform machine learning on the event duration prediction model constructed in step S3, and then input the output of the machine learning-based event duration prediction model into the congestion dissipation time prediction model based on the traffic wave model constructed in step S4, thereby obtaining the prediction result of congestion dissipation time. The training process is as follows: Figure 2 As shown.

[0125] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the present invention, and all such modifications or substitutions should be covered within the protection scope of the present invention.

Claims

1. A method for predicting highway congestion dissipation time considering traffic event information, characterized in that, Includes the following steps: S1. Collect abnormal event data on highways and extract the characteristic relationship between traffic events and traffic congestion; S2. Collect upstream and downstream ETC data and detector data corresponding to abnormal events on highways to construct a traffic flow dataset; S3. Based on the characteristic relationship between traffic events and traffic congestion obtained in step S1, construct an event duration prediction model; S4. Based on the event duration prediction model constructed in step S3, construct a congestion dissipation time prediction model based on the fuzzy traffic wave model. S4.1 Construct a queue length prediction model based on a fuzzy traffic wave model; In the formula, Let E0 be the average equivalent queue length in the multi-lane road event segment at time t; E0 is the standard number of vehicles between segments A and B at the initial time; A x (i,t) represents the number of vehicles joining the event segment at time t in the i-th lane participating in the queue, as determined using fuzzy sets; A xl (i,t) represents the number of vehicles leaving the queuing lane at time t, determined using fuzzy sets; A xn (i,t) represents the number of vehicles in the i-th lane that has not joined the queue at time t, determined using fuzzy sets; M represents the total number of lanes in the road. The average congestion density of vehicles on the road; The optimal vehicle density for traffic on the road; L is the length between road segments A and B. S4.2 Construct a road capacity estimation model under abnormal events; C s =C0·f n ·f HV ·f L In the formula, C s C0 represents the regional traffic capacity of the expressway under abnormal traffic events; f represents the basic traffic capacity of the expressway. n f is the correction factor for the number of closed lanes on the highway segment's capacity; HV The correction factor for the proportion of large vehicles on the capacity of highway sections; f L This is the correction factor for the length of the closed driving lane on the highway's traffic capacity; S4.3 The traffic flow dataset collected in step S2 and the output of the abnormal event duration prediction model constructed in step S3 are used as inputs and fed into the queue length prediction model constructed in step S4.1 to obtain relevant information on highway congestion under abnormal events. In step S4.3, the relevant information on highway congestion under abnormal events includes the maximum number of vehicles in the queue and the congestion dissipation time. The formula for calculating the congestion dissipation time is: In the formula, N U (i,t) represents the number of vehicles joining the event segment in lane i at time t; N D (i,t) represents the number of vehicles leaving the event segment in lane i at time t; t0 represents the output of the event duration prediction model; C s (t) represents the real-time road capacity; S5. Perform machine learning on the event duration prediction model constructed in step S3, and then input the output of the machine learning event duration prediction model into the congestion dissipation time prediction model based on the traffic wave model constructed in step S4, so as to obtain the prediction result of congestion dissipation time.

2. The method for predicting highway congestion dissipation time considering traffic event information according to claim 1, characterized in that: Step S1 includes the following sub-steps: S1.1 Collects information on abnormal events on the highway; The abnormal event information of the highway includes the event type, the vehicles involved, the road section where the event occurred, the time of the event, the impact of the event on traffic, and the duration of the event. S1.2 Preprocess the data information collected in step S1.1, that is, clean the collected data information; S1.3 Perform hot encoding on the preprocessed data information from step S1.2 to convert the text information into encoded information.

3. The method for predicting highway congestion dissipation time considering traffic event information according to claim 2, characterized in that, Step S2 includes the following sub-steps: S2.1 Collect upstream and downstream ETC data and detector data corresponding to abnormal events on highways, and extract traffic flow characteristic information of the road segments corresponding to abnormal events on highways; S2.2 Perform data preprocessing on the traffic flow feature information collected and extracted in step S2.1, that is, clean the extracted data information and remove outliers; S2.3 Based on the data information preprocessed in step S2.2, extract vehicle and traffic flow feature data; S2.4 Match the traffic flow feature data extracted in step S2.3 with the abnormal event information to create a traffic flow dataset under abnormal events.

4. The method for predicting highway congestion dissipation time considering traffic event information according to claim 3, characterized in that, Step S3 includes the following sub-steps: S3.1 Based on the main influencing factors of the event on the duration of traffic congestion obtained in step S1, construct an evaluation system for the severity of the event, and then evaluate the traffic events on the highway. S3.2 Based on highway traffic incident evaluation indicators, construct an event duration prediction model based on a decision tree model; S3.3 Input the dataset after hot encoding of highway abnormal event information in step S1.3 into the event duration prediction model based on decision tree to train the model and obtain the highway abnormal event duration prediction model after training. The input to the highway abnormal event duration prediction model is highway abnormal event information, and the output is the predicted duration of the event.

Images