Urban road signal intersection saturated time headway generalization estimation method and system

Abstract

The invention relates to the technical field of urban road traffic, and discloses an urban road signal intersection saturated time headway generalization estimation method and system, and the method comprises the steps: obtaining license plate recognition data, and carrying out the extraction to obtain an intersection signal period; matching the license plate identification data with the floating vehicle data, and extracting saturated time headway samples of vehicles queued at the intersection in all intersection signal periods; screening out saturated time headway influence factors; using the saturated time headway sample as input, and constructing a probability estimation model of the saturated time headway and various saturated time headway influence factors based on an NGBoost model according to the saturated time headway influence factors; and taking each saturated time headway influence factor as the input of the probability estimation model, and outputting a generalization estimation result of the saturated time headway. The method has the beneficial effects that generalization estimation can be carried out on the saturated time headway according to the existing number plate identification data and the floating car data which are easy to obtain.

Description

Technical Field

[0001] The invention belongs to the technical field of urban road traffic, and relates to a generalized estimation method and system for saturated headway time at a signal intersection on an urban road. Background Art

[0002] Urban road signalized intersections are key nodes where the contradiction between traffic supply and demand is prominent in urban road networks. Accurately describing and understanding the supply capacity characteristics of signalized intersections is of great significance for realizing refined and intelligent intersection traffic management and control. In traffic engineering, headway is one of the important indicators for measuring traffic flow status. Saturated headway reflects the efficiency of vehicles passing through intersections under saturated flow conditions and is an important representation of intersection supply characteristics.

[0003] The existing methods for estimating saturated headway time are as follows: (1) HCM-based methods: by collecting a large amount of data to calculate the relevant correction parameters of the saturated flow rate, the saturated flow rate is estimated, and then the saturated headway time is obtained. The calculation of the correction coefficient involves factors such as the number of lanes, the rate of large vehicles, the slope, the impact of buses, the area type, turning vehicles, pedestrians and non-motor vehicles. (2) Data-driven methods: divided into threshold methods and model methods. The former determines whether the headway time is in a saturated state by setting a threshold. The setting of the threshold is usually based on the median or average value of the collected samples; the latter classifies the headway time by establishing a relevant model to extract the saturated headway time, such as the hidden Markov chain model method, the Gaussian mixture model method, and the K-means clustering method. However, the methods represented by HCM systematically consider various factors affecting the saturated headway time. In real applications, a large amount of data needs to be collected for the calibration of model parameters. Many parameters are difficult to fully obtain in real urban roads. However, the data-driven saturated headway estimation model is limited to the saturated headway estimation under specific conditions and scenarios, and lacks in-depth analysis of the factors affecting the saturated headway, which leads to insufficient generalization of the model.

[0004] At present, the traffic data packets widely available on urban roads in my country are divided into two categories: license plate recognition data and floating vehicle data. Among them, license plate recognition data refers to the use of traffic monitoring cameras to detect vehicles leaving at intersections, and obtain vehicle license plates through automatic text recognition technology, including detector number, detection time, license plate, vehicle type, lane and other information. Floating vehicle data refers to the location data collected by satellite positioning vehicles (such as taxis), including time, license plate, passenger status, location, speed and other information. How to use these real observable data to analyze the influencing factors of saturated headway and establish a generalized estimation model is a problem that needs to be studied in depth and solved. Summary of the invention

[0005] The purpose of the present invention is to provide a generalized estimation method and system for saturated headway time at urban road signal intersections, which can perform generalized estimation of saturated headway time based on existing and easily accessible license plate recognition data and floating vehicle data, and provide a scientific basis for the characterization of supply characteristics of urban signal intersections.

[0006] In order to solve the above technical problems, the present invention is implemented by adopting the following technical solutions.

[0007] In a first aspect, the present invention provides a generalized estimation method for saturated headway time at a signalized intersection on an urban road, comprising:

[0008] Selecting several potential influencing factors of saturated headway, the potential influencing factors include road grade, lane type, vehicle type and weather conditions, judging the potential influencing factors, and then obtaining the influencing factors of saturated headway;

[0009] Input the screened influencing factors of saturated headway into a pre-built probability estimation model to obtain a generalized estimation result of the saturated headway, wherein the generalized estimation result of the saturated headway is the mean value and probability distribution of the saturated headway;

[0010] Wherein, the construction method of the probability estimation model is:

[0011] Based on the NGBoost model, a probability estimation model of the functional relationship between the saturated headway and the influencing factors of the saturated headway in the same period is established.

[0012] In combination with the first aspect, further, a specific method for constructing the probability estimation model is:

[0013] First, initialize the probability estimation model parameters , is the parameter of the target distribution estimated by the probability estimation model. For the normal distribution, .

[0014] Secondly, build a base learner. NGBoost uses base learners (such as decision trees) to gradually approximate the parameters of the target distribution.

[0015] Finally, for each round of iterative updates, perform the following steps:

[0016] Calculate the negative log-likelihood (NLL):

[0017]

[0018] in, is the negative log-likelihood, is the target variable, are the parameters estimated by the probability estimation model in the current iteration;

[0019] Compute the negative log-likelihood with respect to The gradient is:

[0020]

[0021] in, is the negative log-likelihood about The gradient of are the parameters estimated by the probability estimation model in the current iteration, is the negative log-likelihood;

[0022] Compute the Fisher information matrix:

[0023]

[0024] in, is the Fisher information matrix, is the expectation operator, is the target variable, are the parameters estimated by the probability estimation model in the current iteration;

[0025] Natural Gradient Update:

[0026]

[0027] in, is the natural gradient, is the Fisher information matrix, is the negative log-likelihood about The gradient of

[0028] Fitting Natural Gradients Using Base Learners :

[0029]

[0030] in, is the prediction increment of the base learner in the current iteration, is the current iteration number, are the characteristics of the training data, is the natural gradient, is the base learner;

[0031] Update the parameters of the estimated target distribution :

[0032]

[0033] in, It is The update parameters of the step, is the learning rate, is the prediction increment of the base learner in the current iteration.

[0034] In combination with the first aspect, further, the training method of the probability estimation model is: using the extracted saturated headway samples to train and test the probability estimation model, and determining the optimal hyperparameter combination of the probability estimation model based on grid search.

[0035] The optimal hyperparameter combination of the probability estimation model is as follows:

[0036] (1) Learning rate:

[0037] Control the weight adjustment of each tree and affect the model convergence speed.

[0038] (2) Number of base learners (n estimators):

[0039] Specifies the number of base learners to be built, affecting the complexity of the model.

[0040] (3) Minibatch fraction:

[0041] It is used to specify the proportion of training data used in each iteration step during the training process, which determines the amount of data used in each iterative update.

[0042] (4) Column sampling ratio (col sample):

[0043] Used to specify the subsample ratio of feature columns to use when building each base learner, controlling the number of features that each base learner can see.

[0044] In combination with the first aspect, further, the method for extracting the saturated headway time sample is:

[0045] Acquire historical license plate recognition data, use the historical license plate recognition data to identify vehicle queues, and extract intersection signal cycles based on the vehicle queues;

[0046] Get historical floating vehicle data;

[0047] The historical license plate recognition data is matched with the historical floating vehicle data to extract saturated headway time samples of vehicles queuing at intersections under all intersection signal cycles.

[0048] In combination with the first aspect, further, a specific method for obtaining the intersection signal cycle is:

[0049] Pre-process the historical license plate recognition data to obtain the headway time;

[0050] Based on the headway time gap, a headway time gap threshold is set to identify a vehicle queue;

[0051] Starting from the signal cycle determined by the head vehicle of each vehicle queue, the historical license plate recognition data is divided into cycles to obtain the intersection signal cycle.

[0052] In combination with the first aspect, further, the step of preprocessing the historical license plate recognition data includes dividing the license plate recognition data according to the lanes of the intersection, removing the license plate recognition data of repeated detection, and subtracting the continuous detection time of the vehicle to calculate the headway time.

[0053] In combination with the first aspect, further, based on the headway time gap, a specific method for setting a headway time gap threshold value to identify a vehicle queue is as follows:

[0054] For each headway, if it is less than the headway threshold gap, then the vehicle is added to the vehicle queue; otherwise, a new vehicle queue is added and each vehicle queue is numbered;

[0055] Filter out the vehicle queues whose number of vehicles is less than the vehicle number threshold num;

[0056] Merge adjacent vehicle queues that do not meet the minimum red light interval, and update the headway and vehicle queue number;

[0057] Starting from the signal cycle determined by the head vehicle of each vehicle queue, the license plate recognition data is divided into cycles to obtain the intersection signal cycle.

[0058] In combination with the first aspect, further, the specific method of matching the historical license plate recognition data with the historical floating vehicle data and extracting the saturated headway time samples of vehicles queuing at the intersection under all intersection signal cycles is:

[0059] For each intersection signal cycle, the historical floating vehicle data in the parking state is matched with the historical license plate recognition data one by one based on the vehicle license plate and time;

[0060] The headway time of the matched vehicle and its preceding vehicle within the same intersection signal cycle is extracted as the saturated headway time sample.

[0061] In combination with the first aspect, further, a specific method for obtaining the influencing factors of the saturated headway is as follows:

[0062] Select several potential influencing factors of the saturated headway; the potential influencing factors include road grade, lane type, vehicle type, weather conditions, etc., which should generally be easy to obtain and highly generalizable.

[0063] Perform VIF test on the potential influencing factors to determine whether there is collinearity in the characteristics; the specific method is:

[0064] (1) Take each potential influencing factor as the response variable and perform linear regression on all other potential influencing factors.

[0065] (2) Calculation of regression model The value, i.e., the coefficient of determination, indicates to what extent the remaining potential influencing factors can explain the variability of this potential influencing factor.

[0066] (3) Calculate VIF value:

[0067]

[0068] in, is the coefficient of determination of the above regression model.

[0069] (4) Screen out features with collinearity based on the VIF value. The larger the VIF value, the stronger the linear relationship between the potential influencing factor and other potential influencing factors, and further correlation test is performed.

[0070] Obtain the candidate influencing factors of saturated headway; perform correlation test on potential influencing factors with collinearity based on Spearman correlation coefficient, screen out the features with high correlation with other potential influencing factors, and obtain the candidate influencing factors; the calculation formula of Spearman correlation coefficient is as follows:

[0071]

[0072] in, is the Spearman correlation coefficient, It is The difference between the ranking of potential factors affecting the headway of two saturated vehicles is is the number of saturated headway samples.

[0073] Based on the KS test, we judge whether the alternative influencing factors have a significant impact on the saturated headway. The KS test mainly compares two empirical cumulative distribution functions ECDF to judge whether two samples with different characteristic values ​​of a certain alternative influencing factor come from the same distribution. The KS test value is the maximum absolute difference between the two empirical cumulative distribution functions. The larger the value, the greater the difference between the sample and the hypothesized distribution. The KS test value is calculated as follows:

[0074]

[0075] in, is the KS test quantity, The sample size of saturated headway is The empirical cumulative distribution function of The sample size of saturated headway is The empirical cumulative distribution function of represents the size of the first sample, represents the size of the second sample, Represents the value of the variable (i.e., saturated headway).

[0076] The influencing factors of the saturated headway are determined, and the influencing factors of the saturated headway are those that satisfy no obvious collinearity between the features after the collinearity test and the KS test and have a significant impact on the saturated headway.

[0077] In a second aspect, the present invention provides a generalized estimation system for saturated headway time at a signalized intersection on an urban road, comprising:

[0078] A saturated headway time influencing factor module is configured to select a number of potential influencing factors of the saturated headway time, wherein the potential influencing factors include road grade, lane type, vehicle type and weather conditions, and judge the potential influencing factors to obtain the saturated headway time influencing factors;

[0079] The saturated headway probability estimation module is configured to input the screened saturated headway influencing factors into a pre-built probability estimation model to obtain a generalized estimation result of the saturated headway;

[0080] Wherein, the construction method of the probability estimation model is:

[0081] Based on the NGBoost model, a probability estimation model of the functional relationship between the saturated headway and the influencing factors of the saturated headway in the same period is established.

[0082] In combination with the second aspect, further, the saturated headway generalization estimation system provided by the present invention further includes:

[0083] An intersection signal cycle extraction module is configured to obtain historical license plate recognition data, identify vehicle queues using the historical license plate recognition data, and extract the intersection signal cycle based on the vehicle queue;

[0084] A floating vehicle data acquisition module is configured to acquire historical floating vehicle data;

[0085] The saturated headway time sample extraction module is configured to match the historical license plate recognition data with the historical floating vehicle data, and extract the saturated headway time samples of vehicles queuing at the intersection under all intersection signal cycles.

[0086] In a third aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-mentioned saturated headway generalization estimation method.

[0087] In a fourth aspect, the present invention provides a computer device, comprising:

[0088] Memory for storing computer programs;

[0089] A processor is used to execute the computer program to implement the steps of the above-mentioned saturated headway generalization estimation method.

[0090] In a fifth aspect, the present invention provides a computer program product, comprising a computer program, which, when executed by a processor, implements the steps of the above-mentioned saturated headway generalization estimation method.

[0091] The present invention divides the signal period based on license plate recognition data, and identifies the vehicles in the queue at the intersection by matching the historical license plate recognition data with the historical floating vehicle data, so as to realize the extraction of saturated headway time samples. The collinearity between the factors is judged based on the VIF test and the correlation test, and the influence of each factor on the saturated headway time is judged based on the KS test, and the factors affecting the saturated headway time are screened out. The relationship model between the saturated headway time and the factors affecting the saturated headway time is constructed based on the NGBoost model, so as to realize the generalized estimation of the mean and probability distribution of the saturated headway time under dynamic and static factors such as different road grades (such as main roads, secondary roads, and branch roads), lane types (such as left turn lanes, straight left lanes, straight lanes, and straight right lanes), vehicle types (such as small cars and large cars), and weather conditions (such as rainfall and snowfall).

[0092] Compared with the prior art, the present invention has the following beneficial effects:

[0093] (1) The present invention uses license plate recognition data and floating vehicle data that are widely available on urban roads in my country to extract saturated headway samples at signalized intersections, reveal the key influencing factors of saturated headway, and construct a generalized estimation model with the help of machine learning methods, providing a scientific basis for the characterization of the supply characteristics of urban signalized intersections.

[0094] (2) Compared with the estimation method based on HCM, the proposed method takes into account the easily accessible influencing factors of saturated headway time, which significantly improves the generalization of the estimation model in practical applications.

[0095] (3) Compared with the data-driven method, the probability estimation model based on NGBoost constructed in the present invention further analyzes the influencing factors of the saturated headway and can output the probability distribution of the saturated headway, thereby improving the robustness of the model and providing good interpretability for the estimation results.

[0096] (4) The present invention can provide scientific and reliable intersection traffic supply characteristic parameters for urban road traffic flow modeling, traffic signal timing optimization, etc. BRIEF DESCRIPTION OF THE DRAWINGS

[0097] Figure 1 It is a flowchart of the steps of the generalized estimation method of saturated headway time of the present invention;

[0098] Figure 2 Schematic diagram of the flow of the generalized estimation method of saturated headway time of the present invention;

[0099] Figure 3 This is a schematic diagram of the signal period extraction result in the present invention;

[0100] Figure 4 Schematic diagram of the comparison of the probability density distribution of saturated headway time of different lanes in the present invention, wherein Figure (4a) is the left turn lane, Figure (4b) is the through lane 1, and Figure (4c) is the through lane 2;

[0101] Figure 5 It is a schematic diagram of the estimation result of the saturated headway probability distribution in the present invention;

[0102] Figure 6 6a is a schematic diagram for comparing the saturated headway estimation intervals at different confidence levels in the present invention, wherein Figure (6a) is an 80% confidence interval, Figure (6b) is an 85% confidence interval, and Figure (6c) is a 90% confidence interval. DETAILED DESCRIPTION

[0103] The technical solution of the present invention is described in detail below through the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical solution of the present invention, rather than limitations on the technical solution of the present invention. The embodiments of the present invention and the technical features in the embodiments may be combined with each other unless there is a conflict.

[0104] The term "and / or" is only a description of the association relationship between related objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " generally indicates that the related objects are in an "or" relationship.

[0105] Example 1

[0106] like Figure 1 As shown, the generalized estimation method for saturated headway time at a signalized intersection on an urban road proposed by the present invention comprises the following steps:

[0107] Step S1, obtaining historical license plate recognition data, using the historical license plate recognition data to identify vehicle queues, and extracting intersection signal cycles based on vehicle queues

[0108] The method for extracting intersection signal cycle is as follows:

[0109] Step S11, pre-processing the historical license plate recognition data to obtain the headway time, including: dividing the license plate recognition data according to the lanes of the intersection, removing the license plate recognition data of repeated detection, and subtracting the continuous detection time of the vehicle to calculate the headway time.

[0110] Step S12, starting from the signal cycle determined by the head vehicle of each vehicle queue, the historical license plate recognition data is divided into cycles to obtain the intersection signal cycle.

[0111] (1) Based on the headway time obtained from historical license plate recognition data, a headway time threshold gap is set to identify vehicle queues. The specific method is as follows: for each headway time, if it is less than the headway time threshold gap, the vehicle is added to the queue; otherwise, a new vehicle queue is added and each vehicle queue is numbered.

[0112] (2) Filter out vehicle queues whose number of vehicles is less than the vehicle number threshold num.

[0113] (3) Merge adjacent vehicle queues that do not meet the minimum red light interval and update the headway and queue number.

[0114] (4) Starting from the signal cycle determined by the head vehicle in each vehicle queue, the license plate recognition data is divided into cycles.

[0115] The pseudo code of the intersection signal cycle extraction algorithm based on the vehicle queue in step S12 is as follows:

[0116] In the above pseudo code, m represents the number of the vehicle queue. represents a vehicle queue numbered m, n represents the total number of vehicles passing a lane in a continuous period, i represents the i-th vehicle, i=1,2,3…,n, represents the headway time of the i-th vehicle, and red represents the minimum red light interval.

[0117] Step S2, matching the historical license plate recognition data with the historical floating vehicle data, extracting the saturated headway time samples of vehicles queuing at the intersection under all intersection signal cycles. The specific steps are as follows:

[0118] Step S21, for each intersection signal cycle, based on the vehicle license plate and time, the historical floating vehicle data in the parking state is matched with the historical license plate recognition data one by one.

[0119] Step S22, extracting the headway time of the matched vehicle and its preceding vehicle in the same cycle as a saturated headway time sample.

[0120] Step S3, perform collinearity discrimination based on VIF test and correlation test, perform feature influence degree discrimination based on KS test, and select appropriate saturation headway influencing factors. The selection process of saturation headway influencing factors is as follows:

[0121] Step S31, select several potential influencing factors of saturated headway, including but not limited to road grade, lane type, vehicle type, weather conditions, etc., which should generally be easy to obtain and highly generalizable.

[0122] Step S32: Perform VIF test on each influencing factor to determine whether there is collinearity among the features:

[0123] (1) Take each potential influencing factor as the response variable and perform linear regression on all other potential influencing factors.

[0124] (2) Calculation of regression model The value, i.e., the coefficient of determination, indicates to what extent the remaining potential influencing factors can explain the variability of this potential influencing factor.

[0125] (3) Calculate VIF value:

[0126]

[0127] in, is the coefficient of determination of the above regression model.

[0128] (4) Screen out features with collinearity based on the VIF value. The larger the VIF value, the stronger the linear relationship between the potential influencing factor and other potential influencing factors, and further correlation test is performed.

[0129] Step S33, obtaining candidate influencing factors of saturated headway. Based on the Spearman correlation coefficient, a correlation test is performed on the influencing factors with collinearity, and features with high correlation with other factors are screened out. The calculation formula of the Spearman correlation coefficient is as follows:

[0130]

[0131] in, is the Spearman correlation coefficient, It is The difference between the ranking of potential factors affecting the headway of two saturated vehicles is is the number of saturated headway samples.

[0132] Step S34, based on the KS test, determine whether the candidate features have a significant impact on the saturated headway. The KS test mainly determines whether two samples come from the same distribution by comparing two empirical cumulative distribution functions (ECDF). The KS test value is the maximum absolute difference between the two empirical cumulative distribution functions. The larger the value, the greater the difference between the sample and the hypothesized distribution. The KS test value is calculated as follows:

[0133]

[0134] in, is the KS test quantity, The sample size of saturated headway is The empirical cumulative distribution function of The sample size of saturated headway is The empirical cumulative distribution function of represents the size of the first sample, represents the size of the second sample, Represents the value of the variable (i.e., saturated headway).

[0135] Step S35, determining the influencing factors of the saturated headway, the influencing factors of the saturated headway are those that satisfy the conditions that there is no obvious collinearity between the features and that have a significant impact on the saturated headway after the collinearity test and the KS test. Generally speaking, if the KS test value , it is usually considered that there is a significant effect; otherwise, there is no significant effect.

[0136] In step S4, the saturated headway extracted in step S2 is used as input, and according to the influencing factors of the saturated headway determined in step S3, a probability estimation model of the saturated headway and various influencing factors is established based on the NGBoost model. Based on grid search, the model parameters are determined, and with various influencing factors as input, the model outputs a generalized estimation result of the saturated headway.

[0137] The specific steps are as follows:

[0138] Step S41: Based on the NGBoost model, a probability estimation model of the functional relationship between the saturated headway and the influencing factors of the saturated headway is established. First, the probability estimation model parameters are initialized. , is the parameter of the target distribution estimated by the probability estimation model. For the normal distribution, . Next, we build a base learner. NGBoost uses base learners (such as decision trees) to gradually approximate the parameters of the target distribution. Finally, for each round of iterative updates, we perform the following steps:

[0139] (1) Calculate the negative log-likelihood (NLL):

[0140]

[0141] in, is the negative log-likelihood, is the target variable, are the parameters estimated by the probability estimation model in the current iteration.

[0142] (2) Calculate the negative log-likelihood about The gradient is:

[0143]

[0144] in, is the negative log-likelihood about The gradient of are the parameters estimated by the probability estimation model in the current iteration, is the negative log-likelihood.

[0145] (3) Calculate the Fisher information matrix:

[0146]

[0147] in, is the Fisher information matrix, is the expectation operator, is the target variable, are the parameters estimated by the probability estimation model in the current iteration.

[0148] (4) Natural gradient update:

[0149]

[0150] in, is the natural gradient, is the Fisher information matrix, is the negative log-likelihood about gradient.

[0151] (5) Fitting natural gradient using base learners :

[0152]

[0153] in, is the prediction increment of the base learner in the current iteration, is the current iteration number, are the characteristics of the training data, is the natural gradient, is the base learner.

[0154] (6) Update the parameters of the estimated target distribution:

[0155]

[0156] in, It is The update parameters of the step, is the learning rate, is the prediction increment of the base learner in the current iteration.

[0157] Step S42, input the saturated headway extracted in step S2, use the extracted saturated headway to train and test the model, and find the optimal hyperparameter combination of NGBoost based on grid search. The important hyperparameters of the model are as follows:

[0158] (1) Learning rate:

[0159] Control the weight adjustment of each tree and affect the model convergence speed.

[0160] (2) Number of base learners (n estimators):

[0161] Specifies the number of base learners to be built, affecting the complexity of the model.

[0162] (3) Minibatch fraction:

[0163] It is used to specify the proportion of training data used in each iteration step during the training process, which determines the amount of data used in each iterative update.

[0164] (4) Column sampling ratio (col sample):

[0165] Used to specify the subsample ratio of feature columns to use when building each base learner, controlling the number of features that each base learner can see.

[0166] In step S43, the learned NGBoost model is a generalized estimation model of saturated headway. In practical applications, various influencing factors are used as input, and the model outputs the probability distribution of saturated headway.

[0167] Example 2

[0168] This embodiment selects an area near a high-speed railway station in a certain city as an example scenario of the method. The total coverage area of ​​the area is about 25 square kilometers, including about 30 urban roads and 50 signalized intersections. This embodiment extracts the saturated headway time of 11 intersections in the area from July to August 2018 and further realizes generalized estimation.

[0169] Taking the north entrance of a certain intersection as an example, the data of August 8, 2018 was selected. After data preprocessing, the signal cycle of the intersection was extracted based on the vehicle queue. Figure 3 shown.

[0170] Furthermore, the headway time is divided based on the signal cycle, and the license plate recognition data is matched with the floating vehicle data to obtain the saturated headway time of each lane of the entrance road in July 2018. The probability density distribution is as follows: Figure 4 As shown, Figure (4a) is the left turn lane, Figure (4b) is the through lane 1, and Figure (4c) is the through lane 2. The data statistical results are shown in Table 1.

[0171] After collinearity judgment and influence degree judgment, the influencing factors of saturated headway are as follows: queue position, rainfall, travel time, road grade, lane type, front vehicle model, current vehicle model, and central dividing strip.

[0172] Based on the NGBoost model, a probability estimation model for saturated headway is constructed. The extracted saturated headway is used for model training to estimate the normal distribution probability density of saturated headway. The results are as follows: Figure 5 As shown. Estimation intervals under different confidence levels are constructed respectively, and the results are as follows Figure 6 As shown, Figure (6a) is the 80% confidence interval, Figure (6b) is the 85% confidence interval, and Figure (6c) is the 90% confidence interval.

[0173] The coverage of the estimated interval (PICP) and the normalized average width (PINAW) are selected to evaluate the probability estimation results. The above indicators evaluate the model performance from two aspects: interval reliability and interval width. Figure 6 As can be seen from Table 2, the estimated interval of the model covers a large proportion of the actual observations, the interval reliability is high, and the interval width is small, so the estimation effect is good.

[0174] Example 2

[0175] Based on the same inventive concept as that of Example 1, this example introduces a generalized estimation system for saturated headway time at a signalized intersection on an urban road, comprising:

[0176] A saturated headway time influencing factor module is configured to select a number of potential influencing factors of the saturated headway time, the potential influencing factors including road grade, lane type, vehicle type and weather conditions, and identify the potential influencing factors to obtain the saturated headway time influencing factors;

[0177] The saturated headway probability estimation module is configured to input the screened saturated headway influencing factors into a pre-built probability estimation model to obtain a generalized estimation result of the saturated headway;

[0178] Among them, the construction method of the probability estimation model is:

[0179] Based on the NGBoost model, a probability estimation model of the functional relationship between the saturated headway and the influencing factors of the saturated headway in the same period is established.

[0180] In a specific implementation of this embodiment, the saturated headway generalization estimation system of the present invention further includes:

[0181] An intersection signal cycle extraction module is configured to obtain historical license plate recognition data, identify vehicle queues using the license plate recognition data, and extract the intersection signal cycle based on the vehicle queue;

[0182] A floating vehicle data acquisition module is configured to acquire historical floating vehicle data;

[0183] The saturated headway time sample extraction module is configured to match the historical license plate recognition data with the historical floating vehicle data, and extract the saturated headway time samples of vehicles queuing at the intersection under all intersection signal cycles.

[0184] Example 3

[0185] Based on the same inventive concept as other embodiments, this embodiment introduces a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the above-mentioned saturated headway generalization estimation method are implemented.

[0186] Example 4

[0187] Based on the same inventive concept as other embodiments, this embodiment introduces a computer device, including: a memory for storing a computer program; a processor for executing the computer program to implement the steps of the above-mentioned saturated headway generalization estimation method.

[0188] Example 5

[0189] Based on the same inventive concept as other embodiments, this embodiment introduces a computer program product, including a computer program, which implements the steps of the above-mentioned saturated headway generalization estimation method when executed by a processor.

[0190] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0191] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0192] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0193] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0194] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the enlightenment of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the purpose of the present invention and the claims, which all fall within the protection of the present invention.

Claims

1. A generalized estimation method for saturated headway time at a signalized intersection on an urban road, characterized in that: include: Selecting several potential influencing factors of saturated headway, the potential influencing factors include road grade, lane type, vehicle type and weather conditions, judging the potential influencing factors, and then obtaining the influencing factors of saturated headway; Input the screened influencing factors of saturated headway into the pre-built probability estimation model to obtain the generalized estimation result of saturated headway; Wherein, the construction method of the probability estimation model is: Based on the NGBoost model, a probability estimation model of the functional relationship between the saturated headway and the influencing factors of the saturated headway in the same period is established.

2. The generalized estimation method for saturated headway time at urban road signalized intersections according to claim 1 is characterized in that: The training method of the probability estimation model is: using the extracted saturated headway time samples to train and test the probability estimation model, and determining the optimal hyperparameter combination of the probability estimation model based on grid search; The method for extracting the saturated headway time samples is as follows: Acquire historical license plate recognition data, use the historical license plate recognition data to identify vehicle queues, and extract intersection signal cycles based on the vehicle queues; Get historical floating vehicle data; The historical license plate recognition data is matched with the historical floating vehicle data to extract saturated headway time samples of vehicles queuing at intersections under all intersection signal cycles.

3. The generalized estimation method for saturated headway time at a signalized intersection on an urban road according to claim 2, characterized in that: The specific method for obtaining the intersection signal cycle is: Pre-process the historical license plate recognition data to obtain the headway time; Based on the headway time gap, a headway time gap threshold is set to identify a vehicle queue; Starting from the signal cycle determined by the head vehicle of each vehicle queue, the historical license plate recognition data is divided into cycles to obtain the intersection signal cycle.

4. The generalized estimation method for saturated headway time at a signalized intersection on an urban road according to claim 3, characterized in that: Based on the headway time gap, the specific method of setting the headway time gap threshold value to identify the vehicle queue is: For each headway, if it is less than the headway threshold gap, then the vehicle is added to the vehicle queue; otherwise, a new vehicle queue is added and each vehicle queue is numbered; Filter out the vehicle queues whose number of vehicles is less than the vehicle number threshold num; Merge adjacent vehicle queues that do not meet the minimum red light interval, and update the headway and vehicle queue number; Starting from the signal cycle determined by the head vehicle of each vehicle queue, the license plate recognition data is divided into cycles to obtain the intersection signal cycle.

5. The generalized estimation method for saturated headway time at urban road signalized intersections according to claim 2 is characterized in that: The specific method of matching the historical license plate recognition data with the historical floating vehicle data and extracting the saturated headway time samples of vehicles queuing at the intersection under all intersection signal cycles is as follows: For each intersection signal cycle, the historical floating vehicle data in the parking state is matched with the historical license plate recognition data one by one based on the vehicle license plate and time; The headway time of the matched vehicle and its preceding vehicle within the same intersection signal cycle is extracted as the saturated headway time sample.

6. The generalized estimation method for saturated headway time at urban road signalized intersections according to claim 1 is characterized in that: The method to obtain the factors affecting the saturated headway is: Based on the VIF test and correlation test, several potential influencing factors of saturated headway are selected for collinearity judgment, and the alternative influencing factors of saturated headway are obtained. Based on the KS test, the characteristic influence degree of the alternative influencing factors of saturated headway is judged, and then the influencing factors of saturated headway are obtained.

7. A generalized estimation system for saturated headway time at urban road signalized intersections, characterized in that: include: A saturated headway time influencing factor module is configured to select a number of potential influencing factors of the saturated headway time, wherein the potential influencing factors include road grade, lane type, vehicle type and weather conditions, and judge the potential influencing factors to obtain the saturated headway time influencing factors; The saturated headway probability estimation module is configured to input the screened saturated headway influencing factors into a pre-built probability estimation model to obtain a generalized estimation result of the saturated headway; Wherein, the construction method of the probability estimation model is: Based on the NGBoost model, a probability estimation model of the functional relationship between the saturated headway and the influencing factors of the saturated headway in the same period is established.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the generalized estimation method for saturated headway time described in any one of claims 1 to 6 are implemented.

9. A computer device, characterized in that: include: Memory for storing computer programs; A processor is used to execute the computer program to implement the steps of the generalized saturated headway estimation method according to any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that: When the computer program is executed by a processor, the steps of the generalized saturated headway estimation method according to any one of claims 1 to 6 are implemented.

Images