Urban trunk road-oriented signal control problem diagnosis method

Abstract

The invention relates to a signal control problem diagnosis method for an urban trunk road, and the method comprises the steps: taking the average vehicle control delay of the urban trunk road as an evaluation index, quantifying the benefits of different signal control schemes, and obtaining a benefit matrix; based on the scene to be diagnosed, defining a cost function, and obtaining a cost matrix; based on the benefit matrix and the cost matrix, generating a recommendation system special value matrix, and performing complementation and deviation correction on the recommendation system special value matrix; quantifying the similarity between the scene to be diagnosed and each typical traffic demand scene in the recommendation system characteristic value matrix, and calculating the reference value weight of the typical traffic demand scene to the scene to be diagnosed according to the similarity proportion; and based on the reference value weight, calculating unreasonable setting probabilities of various signal control parameters in the scene needing to be diagnosed, and realizing signal control problem diagnosis according to the unreasonable setting probabilities. Compared with the prior art, the urban trunk signal control problem can be diagnosed only by depending on small sample historical data.

Description

Technical Field

[0001] This invention relates to the field of traffic information, and in particular to a method for diagnosing signal control problems on urban arterial roads. Background Technology

[0002] Traditional traffic signal control optimization typically relies on specific evaluation metrics (such as average vehicle delay, number of stops, and maximum queue length). Experts analyze traffic problems and select appropriate control measures (such as signal control, traffic organization, and channelization design) to optimize the solution. Among these, "problem diagnosis" mainly relies on management experience and expert knowledge. Existing traffic control systems lack effective coordination between state estimation, problem diagnosis, and problem-oriented control optimization, making it difficult to form an autonomous closed-loop optimization mechanism.

[0003] In recent years, research on problem diagnosis can be broadly categorized into three types based on different research objects and classification criteria: classification based on the control object and purpose, scenario classification based on operational evaluation characteristics, and scenario classification based on the rationality of control strategies and parameter settings. The first type focuses on clearly defining the control object and its boundaries, such as bus signal priority control or urban expressway ramp control. The second type combines multiple evaluation indicators to provide a comprehensive description of the macroscopic operational status, often used to analyze traffic state mechanisms, such as low-saturation intersections and frequently congested hotspots. The third type emphasizes establishing the correspondence between control optimization strategies and parameter settings, directly locating optimization targets, and stressing the autonomy and precision of signal control operations, such as the rationality of green light ratios or public transport cycles.

[0004] Currently, although scholars both domestically and internationally have conducted some research on the diagnosis of control strategies and parameter rationality, most studies focus on single signal control parameters and primarily rely on internet-based vehicle trajectory data. A systematic and unified diagnostic framework has not yet been established, making it difficult to achieve comprehensive diagnosis of multiple signal control parameters. Furthermore, it is challenging to diagnose problems on urban arterial roads where data collection is incomplete. While engineering applications increasingly emphasize signal control problem diagnosis, and existing products can provide an overall evaluation of control parameter rationality and traffic operation status, limitations remain in the scope and systematic coverage of their diagnoses, making it difficult to meet the comprehensive diagnostic needs in complex traffic environments. Chinese patent CN116311981B discloses a traffic problem diagnosis method for signal-controlled intersections on urban roads. Although it enriches the types of problems that can be diagnosed and achieves problem-oriented optimization control, it is still imperfect in terms of defining the concept of "diagnosis" and the diagnosis method for the rationality of control parameters. At the same time, it cannot perform well in the face of unfamiliar arterial roads with sparse data. In the future, the focus should be on solving the problems of clarifying the diagnostic objects and building a complete diagnostic system, especially in the scenario of urban arterial roads, and establishing a method that covers the rationality diagnosis of multiple types of parameters to improve the comprehensiveness and applicability of signal control problem diagnosis. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the existing technology and provide a diagnostic method for signal control problems on urban arterial roads. It improves upon the shortcomings of existing urban arterial road management systems, such as the lack of state estimation to problem diagnosis and problem-oriented control optimization, and the inability to form an autonomous closed loop. It enriches the types of problems that can be diagnosed and can accurately and autonomously diagnose operational problems on urban arterial roads.

[0006] The objective of this invention can be achieved through the following technical solutions:

[0007] A method for diagnosing signal control problems on urban arterial roads, the method comprising:

[0008] Obtain the current diagnostic scenario that requires diagnosis;

[0009] Considering the potential traffic demand and control parameters of urban arterial roads, typical traffic demand scenarios and signal control scenarios are generated. The typical traffic demand scenarios and signal control scenarios are then arbitrarily combined to obtain combined scenarios.

[0010] Using the average vehicle delay on urban arterial roads as an evaluation indicator, the average vehicle delay value of combined scenarios is generated through numerical calculation or simulation. By comparing the average vehicle delay of each combined scenario with the average vehicle delay of the scenario to be diagnosed, the benefit matrix is ​​obtained.

[0011] Based on the diagnostic scenario, a cost function is defined, and the signal control cost of each combined scenario is calculated using the cost function to obtain the cost matrix.

[0012] Based on the benefit matrix and cost matrix, a recommendation system feature matrix is ​​generated, and the feature matrix is ​​completed and corrected using the matrix singular value decomposition method that considers bias.

[0013] The similarity between the scenario requiring diagnosis and each typical traffic demand scenario in the feature value matrix of the recommendation system is quantified, and the reference value weight of the typical traffic demand scenario to the scenario requiring diagnosis is calculated based on the similarity ratio.

[0014] Based on the aforementioned reference value weights, the probability of unreasonable settings of various signal control parameters in the scenario to be diagnosed is calculated, and the signal control problem is diagnosed based on the probability of unreasonable settings.

[0015] Furthermore, the traffic demand scenario is represented by a path flow matrix, which is composed of the flow of each path on the urban arterial road; the signal control scenario is formed by a combination of different signal control parameters, which include common cycle length, green ratio, coordinated phase difference, phase sequence and variable lane function.

[0016] Furthermore, the process of generating the aforementioned typical traffic demand scenario includes:

[0017] Each boundary point of the urban arterial road is regarded as a traffic origination point and a traffic attraction point, and the traffic generation volume of each traffic origination point and the traffic attraction volume of each traffic attraction point are set.

[0018] Considering the different states of low and high traffic volume on urban arterial roads, a percentage fluctuation of traffic generation or attraction is set, the range of variation of traffic generation at each traffic generation point and traffic attraction at each traffic attraction point is obtained, and several sets of boundary generation and attraction combinations are constructed.

[0019] Based on the combination of attraction quantities generated at the aforementioned boundaries, the path flow matrix of urban arterial roads is obtained using a traffic distribution method.

[0020] Based on the combination of the attraction quantities at the boundary and their corresponding path flow matrices, a path flow tensor is constructed, and a preset number of typical traffic demand scenarios are obtained through a clustering algorithm.

[0021] The process of generating the signal control scenario includes:

[0022] Based on existing traffic standards, public cycle length parameters are set, including maximum cycle length and minimum cycle length;

[0023] The green light ratio is set based on the NEMA double-ring structure, and the green light ratio includes the green light duration and the red light duration;

[0024] Using the current main intersections of urban arterial roads as a reference, calculate the coordinated phase difference of other intersections. The minimum phase difference is 0 seconds, and the maximum is the common cycle length.

[0025] For the current urban arterial road scenario, all reversible lane functions are enumerated, all possible values ​​are traversed, and the parameters of the reversible lane function are obtained.

[0026] Based on the common cycle length parameter, green light ratio phase duration, intersection coordination phase difference, and variable lane function parameters, various signal control scenarios are generated.

[0027] Furthermore, the process of obtaining the benefit matrix includes:

[0028] Using the average vehicle delay per vehicle on urban arterial roads as the evaluation index, the average vehicle delay value for each combined scenario is generated based on numerical calculation or simulation. The control effectiveness index matrix X is obtained by combining the average vehicle delay values ​​of each combined scenario. M·N ;

[0029] Let the vehicle-average control delay in the scenario requiring diagnosis be x. o Based on the control benefit index matrix X M·N The benefit matrix is ​​calculated using the following expression:

[0030] E ij =X ij -x o

[0031] Among them, E ij X represents the value in the i-th row and j-th column of the benefit matrix, that is, the improvement in control benefit for the i-th traffic demand scenario under the j-th signal control scenario. ij This represents the value in the i-th row and j-th column of the benefit index matrix, that is, the average vehicle delay under the j-th signal control scenario for the i-th traffic demand scenario.

[0032] Furthermore, the expression for the cost function is:

[0033] COST ij =c1·δ(C ij ≠C O )+c2·δ(g ij ≠g O )+c3·δ(sq ij ≠sq O )+c4·δ(δ ij ≠δ O )+c5·δ(α ij ≠α O )

[0034] Among them, C ij ,g ij ,sq ij ,δ ij ,α ij Let C represent the common cycle length, green ratio, phase sequence, coordinated phase difference, and variable lane function parameters for the corresponding signal control scenario in the i-th row and j-th column, respectively. Let c1, c2, c3, c4, and c5 represent the costs of optimizing the common cycle length, green ratio, phase sequence, coordinated phase difference, and variable lane function, respectively. O ,g o ,sq o ,δ o and a o These represent the common cycle length, green light ratio, phase sequence, coordinated phase difference, and variable lane function, respectively, for reference in the diagnostic scenario. COST ij This represents the value in the i-th row and j-th column of the cost matrix, which is the cost of the i-th traffic demand scenario under the j-th signal control scenario.

[0035] Furthermore, the process of generating the recommendation system's feature value matrix based on the benefit matrix and cost matrix includes:

[0036] The benefit matrix and cost matrix are normalized respectively, and the benefit-cost ratio is calculated based on the normalized benefit matrix and cost matrix. The benefit-cost ratio is used as the feature value to construct the recommendation system feature value matrix.

[0037] Furthermore, the process of completing and correcting the feature value matrix of the recommendation system includes:

[0038] Set the bias values ​​of traffic demand scenario to signal control scenario and the bias values ​​of signal control scenario to traffic demand scenario, and construct an objective function that takes into account the bias values.

[0039] The objective function is solved iteratively using the gradient descent method to obtain the optimal latent vector and bias term, and the feature value matrix of the recommendation system is then completed and corrected.

[0040] Furthermore, the objective function considering the deviation value is:

[0041]

[0042] Where μ represents the average mean of the system, b i Let b represent the bias term for the i-th traffic demand scenario. j Let λ represent the bias term for the j-th signal control scenario, where λ is the regularization coefficient, and m ij Let represent the true characteristic values ​​of the i-th traffic demand scenario and the j-th signal control scenario, that is, the benefit-cost ratio of the i-th traffic demand scenario and the j-th signal control scenario. p represents the latent vector of the j-th signal-controlled scenario. i Let represent the latent vector of the i-th traffic demand scenario.

[0043] Furthermore, the expression for calculating the similarity between the scenario requiring diagnosis and each typical traffic demand scenario in the feature value matrix of the recommendation system is as follows:

[0044]

[0045] Where sim(i,k) represents the similarity function between the scenario i to be diagnosed and the k-th typical traffic demand scenario. express, express;

[0046] The formula for calculating the reference value weight of the typical traffic demand scenario for the scenario requiring diagnosis is as follows:

[0047]

[0048] Among them, w i,k Z represents the weight of scenario i to be diagnosed in relation to the k-th typical traffic demand scenario. iThis represents the nearest neighbor set of scenario i that needs to be diagnosed, and its size is the number of typical traffic demand scenarios.

[0049] Furthermore, the calculation expression for the probability of unreasonable settings of various signal control parameters in the scenario requiring diagnosis is as follows:

[0050]

[0051] X∈{C,g,sq,δ,α}

[0052] Where X represents the control parameter being analyzed, C, g, sq, δ, and α represent the common cycle length, green light ratio, phase sequence, coordinated phase difference, and variable lane function, respectively, and R X Represents the potential causal probability of parameter X, card(·) is a function to calculate the number of elements in a finite set, I X This represents the number of control schemes for sampling points where parameter X in the signal control sampling scheme set has been changed. k,j This represents the value of the signal control scheme at the j-th sampling point in the k-th typical traffic demand scenario for parameter X.

[0053] Compared with the prior art, the beneficial effects of the present invention include:

[0054] 1. This invention generates and completes the feature matrix of a recommendation system through typical traffic demand scenarios and signal control scenarios. It does not rely on full historical data, but only requires a small sample of original path traffic and signal timing data to diagnose signal control problems on urban arterial roads. This allows for autonomous and accurate diagnosis of signal control issues on urban arterial roads, improving the applicability and reliability of the diagnosis. Furthermore, this invention achieves autonomous diagnosis through a fully algorithmic design, requiring no subjective human intervention. Only basic data needs to be input to output quantitative diagnostic conclusions, improving the objectivity and consistency of diagnostic results and reducing reliance on high-end experts. This invention does not depend on the specific attributes of urban arterial roads; demand and control scenarios can be generated arbitrarily and can be transferred to other scenarios outside of urban arterial roads, demonstrating strong scalability.

[0055] 2. This invention incorporates core control parameters such as common cycle, green light ratio, phase sequence, coordinated phase difference, and variable lane function into a unified diagnostic framework. By quantifying the probability of unreasonableness of each type of parameter through causal probability calculation, it achieves precise problem localization.

[0056] 3. When generating the feature matrix of the recommendation system, this invention considers that due to the diversity of traffic demand scenarios and control variables, most combined scenarios lack sufficient traffic state data, resulting in data sparsity in the constructed feature matrix of the recommendation system. Therefore, it adopts matrix completion technology in recommendation systems, combined with urban arterial road traffic state extrapolation methods and scenario generation technology, to generate a sparse feature matrix representing the supply and demand relationship scenario. By setting a bias term to consider sampling bias, matrix completion is completed. This method has low data requirements, can effectively support subsequent scheme optimization, and is applicable to urban arterial roads with incomplete data collection, reducing the threshold for project implementation and significantly reducing implementation costs.

[0057] 4. Although this invention focuses on urban arterial roads, the traffic problem diagnosis process and its embedded methods have high applicability and can be extended to various research scenarios such as single-point intersections, urban road networks, and expressways, demonstrating strong scalability.

[0058] 5. This invention calculates the similarity between the diagnostic scenario and the typical scenario by Euclidean distance and assigns weights, and then calculates the probability of unreasonable settings of each type of control parameter based on the weights to locate the core problem parameters. This can solve the problem that new diagnostic scenarios are difficult to match with typical scenarios completely. At the same time, the diagnostic results are presented in the form of probabilities, which intuitively reflects the causal probability of each parameter, making it easier for engineers to understand and formulate optimization strategies. Attached Figure Description

[0059] Figure 1 This is a flowchart of the method of the present invention;

[0060] Figure 2 This is a schematic diagram of the vehicle fleet dispersion in an embodiment of the present invention;

[0061] Figure 3 This is a schematic diagram of the spatiotemporal region in an embodiment of the present invention;

[0062] Figure 4 This is a schematic diagram of key points of traffic waves in an embodiment of the present invention;

[0063] Figure 5 This is a schematic diagram of the traffic wave point set in an embodiment of the present invention;

[0064] Figure 6 This is a geometric schematic diagram of the verification scenario in an embodiment of the present invention;

[0065] Figure 7 This is a schematic diagram of the original signal timing in the verification scenario of this invention embodiment;

[0066] Figure 8 This is a flowchart of the reliability verification method for problem diagnosis conclusions in an embodiment of the present invention. Detailed Implementation

[0067] 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. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0068] Example 1

[0069] This embodiment discloses a method for diagnosing signal control problems on urban arterial roads, the method as follows: Figure 1 As shown, the process specifically includes steps S1-S7, and the details of each step are as follows:

[0070] Step S1: Obtain the current diagnostic scenario that needs to be diagnosed.

[0071] The scenarios requiring diagnosis include the current traffic demand scenario and the current signal control scenario. Step S1 is the step of collecting the current traffic information of the urban arterial road.

[0072] Step S2: Consider the potential traffic demand and control parameters of urban arterial roads, generate typical traffic demand scenarios and signal control scenarios, and arbitrarily combine the typical traffic demand scenarios and signal control scenarios to obtain combined scenarios.

[0073] Traffic demand scenarios are represented by a path flow matrix, which consists of the flow of each path on urban arterial roads; signal control scenarios are formed by combinations of different signal control parameters, including common cycle length, green ratio, coordinated phase difference, phase sequence, and reversible lane function.

[0074] The process of generating typical traffic demand scenarios includes:

[0075] The boundary points of urban arterial roads are regarded as traffic origination points and traffic attraction points, and the traffic demand between these points is independent of each other.

[0076] Path flow under the original traffic demand can provide reference information, and the correlation between path flows does not change significantly. Based on this, the traffic generation volume at each traffic origination point and the traffic attraction volume at each traffic attraction point can be set. The traffic generation volume at each boundary point of the studied urban arterial road is... Traffic attraction volume is Where X is the number of boundary points;

[0077] Considering the different states of low and high traffic volume on urban arterial roads, a percentage fluctuation α is set for traffic generation or attraction. Q Obtain the range of variation of traffic generation volume at each traffic origination point and traffic attraction volume at each traffic attraction point, and construct several sets of boundary generation and attraction volume combinations;

[0078] The range of variation can be expressed as:

[0079]

[0080] For each combination, the total boundary traffic generation should be equal to the total boundary traffic attraction. If they are not equal, the boundary traffic generation should be used as a reference to adjust the boundary traffic attraction until they reach a balance.

[0081] Based on the combination of attraction at the boundary, the path flow matrix of urban arterial roads is obtained using a traffic distribution method.

[0082] Based on the combination of boundary attraction quantities and their corresponding path flow matrices, assume Q L*X*X The obtained path flow tensor is given by K, where K represents the number of possible boundary traffic occurrence / attraction scenarios within the research scenario, and X is the number of boundary points. Based on this, a preset number of typical traffic demand scenarios are obtained through clustering algorithms.

[0083] This invention uses the Furniture method as an example to illustrate the main steps for obtaining the path flow matrix of urban arterial roads:

[0084] (1) Initialize the number of calculations m = 0;

[0085] (2) Given the current path traffic table And U in the future path flow table i V j X represents the combination of traffic generation / attraction. Where U... i V j These represent the generated traffic volume and attracted traffic volume in the future path traffic table, respectively; X represents the generated traffic volume in the future path traffic table.

[0086] (3) Calculate the growth rates of traffic generated and traffic attracted at each boundary point. in, This represents the growth coefficient of boundary point i in the m-th iteration; Let represent the attraction growth coefficient of boundary point j in the m-th iteration. In the first iteration, it is assumed that all attraction growth coefficients are 1, i.e.

[0087]

[0088] (4) Find the approximate value of the traffic volume in the (m+1)th distribution. The calculation rule for f is: if in the previous step but on the contrary,

[0089]

[0090] (5) Perform convergence judgment. ∈ is an arbitrarily given error constant, which is taken as 0.03 in this embodiment of the invention;

[0091]

[0092] (6) If the error accuracy requirement is met, the final result is obtained; otherwise, m = m + 1, and continue iterative calculation based on steps (3) to (5).

[0093] K-means clustering is a commonly used unsupervised learning algorithm, known for its high computational efficiency and strong interpretability. Therefore, this embodiment of the invention assumes the number of typical application scenarios is M, and uses K-means clustering as an example to illustrate how to obtain a preset number of typical traffic demand scenarios. The steps are as follows:

[0094] (1) Initialization: Randomly select M samples from the first dimension of the original tensor as the initial cluster centers;

[0095] (2) Sample allocation: Calculate the distance between each sample and each cluster center, and allocate the sample to the cluster containing the nearest cluster center;

[0096] (3) Update cluster centers: Calculate the mean of all samples in each cluster and use the mean as the new cluster center;

[0097] (4) Repeat steps (2) and (3) until the cluster centers no longer change or the predetermined number of iterations is reached;

[0098] (5) Convergence: When the cluster centers are stable or the number of iterations reaches a predetermined value, the K-means algorithm converges, and the typical traffic demand scenarios of type M are obtained.

[0099] The process of generating a signal control scenario includes:

[0100] Based on existing traffic standards, set public cycle length parameters, including maximum and minimum cycle lengths. For example, "Road Traffic Signal Control Methods Part 4: Implementation Requirements for Arterial Coordinated Signal Control Methods GA / T527.4-2018" recommends adjusting the public cycle length according to traffic flow characteristics at different times, and the maximum value should not exceed 180 seconds. The minimum cycle length within the urban arterial road area can be calibrated according to the minimum cycle length index calculation method introduced in "HCM 2010" to avoid potential traffic congestion.

[0101] The green light ratio is set based on the NEMA (National Electrical Manufacturers Association) dual-ring structure. The green light ratio is the phase sequence parameter, which includes the green light duration and the red light duration.

[0102] With a fixed cycle length, once the green light ratio for each flow direction is determined, the phase sequence is also determined. Combining the green light ratio and the common cycle length parameter, the duration of each phase can be calculated. For the green light ratio parameter, the green light duration must first meet the minimum green light duration requirement for pedestrian crossings. This requirement depends on the geometric conditions of the studied scenario. Furthermore, the "Road Traffic Signal Control Methods Part 4: Implementation Requirements for Arterial Coordinated Signal Control Methods GA / T 527.4-2018" recommends that the green light duration for coordinated directions should not be less than 30 seconds. These two conditions jointly constrain the minimum green light duration for each flow direction. On the other hand, the "Traffic Signal Control Guidelines" recommend that the maximum red light time for motor vehicles should not exceed 120 seconds. Therefore, for each flow direction, the maximum value of the green light ratio can be deduced from the recommended maximum red light time. Based on the minimum and maximum green light ratios for each phase, the sampling range of the green light ratio can be further reduced according to actual conditions (such as road geometry, pedestrian crossing speed, and other calibration parameters).

[0103] Taking the current main intersection of the urban arterial road as a reference (the global coordinated phase difference is 0), calculate the coordinated phase difference of the other intersections. The minimum phase difference range is 0 seconds (i.e., synchronous coordination), and the maximum value is the common cycle length.

[0104] For the current urban arterial road scenario, all variable lane functions are enumerated, and all possible values ​​are traversed to obtain the variable lane function parameters. Since lane functions are discrete variables, all variable lane functions can be enumerated for the urban arterial road scenario under study.

[0105] Based on the common cycle length parameter, green ratio phase duration, intersection coordination phase difference, and variable lane function parameters, multiple signal control schemes are generated. Each scheme is a signal control scenario. Since the control parameters are independent of each other, the number of sampling points for a potential signal control scheme is the product of the number of sampling points for each parameter. The density of sampling points in the control scheme depends on the calculation conditions. When the conditions are sufficient, the sampling point density of the control scheme can be increased to improve the reliability of the diagnostic results.

[0106] Step S3: Using the average vehicle delay on urban arterial roads as an evaluation indicator, the average vehicle delay value of the combined scenarios is generated through numerical calculation or simulation. By comparing the average vehicle delay of each combined scenario with the average vehicle delay of the scenario to be diagnosed, the benefit matrix is ​​obtained.

[0107] The process of obtaining the benefit matrix includes:

[0108] Using the average vehicle delay per vehicle on urban arterial roads as the evaluation index, the average vehicle delay value for each combined scenario is generated based on numerical calculation or simulation. The control effectiveness index matrix X is obtained by combining the average vehicle delay values ​​of each combined scenario. M·N ;

[0109] Let the vehicle-average control delay in the scenario requiring diagnosis be x. O Based on the control benefit index matrix X M·N The benefit matrix is ​​calculated using the following expression:

[0110] E ij =X ij -x O

[0111] Among them, E ij X represents the value in the i-th row and j-th column of the benefit matrix, that is, the improvement in control benefit for the i-th traffic demand scenario under the j-th signal control scenario. ij This represents the value in the i-th row and j-th column of the benefit index matrix, that is, the average vehicle delay under the j-th signal control scenario for the i-th traffic demand scenario.

[0112] To address the issue of generating average vehicle delay values ​​for various combined scenarios based on numerical calculations or simulations, this invention presents a traffic wave extrapolation model for urban arterial roads, the main steps of which are as follows:

[0113] (1) Input traffic volume calculation

[0114] Assume the path flow matrix of the urban arterial road under study is Q. X*X Where X represents the number of outer cross sections within the urban arterial road area, and q ij This represents the flow rate from the starting section i to the ending section j. The matrix is ​​defined as follows:

[0115]

[0116] Assuming the analyzed time period is T(s), which is typically set as the length of the control period, this method differs from traditional traffic wave reconstruction methods that assume a fixed arrival rate. It considers the short-term fluctuations in traffic volume, dividing the study period into multiple fixed-length intervals Δ. t The time period (s). Therefore, for any external cross section s, its flow vector is defined as follows:

[0117] q′ s,l =(q s ×Δ t / T*e T +∈ T )×θ l

[0118]

[0119] Where, q′ s,l q represents the traffic demand of lane l in section s within one hour. si Let represent the flow rate at the starting section 's' and the ending section 'i', and let 'e' represent the unit row vector whose length is the number of time periods within the studied period. ∈ T This represents the random error generated by short-term traffic fluctuations, where ∈ i ~U(-αq s· ,αq s· ), where α is the possible percentage fluctuation of traffic volume, and ∈ T The length of e T Consistent; θ l This represents the proportion of traffic volume in lane l within section s.

[0120] (2) Periodic arrival flow estimation

[0121] The traffic demand q′ of lane l in section s is obtained. s,l Subsequently, if we disregard differences in vehicle movement, the flow rate reaching the downstream parking line should be equal to q′. s,l The main difference lies in the overall time offset. However, due to variations in driver performance, vehicle travel time or speed can fluctuate during operation. Therefore, vehicles input into an urban arterial road scenario typically arrive at the stop line in platoons. Due to speed differences among vehicles, the spacing between vehicles in the platoon gradually increases, resulting in platoon dispersion, such as... Figure 2 As shown.

[0122] Extensive research has been conducted by scholars both domestically and internationally on the vehicle platoon discretization problem. This method models the vehicle platoon discretization process based on the Robertson model. Taking lane l in section s as an example, the traffic volume q′ calculated in step (1) is used... s,l , assuming q′ s,l,m This represents the traffic volume of lane l in section s during time period m. The vehicle fleet is discretized using a geometric distribution model, and the downstream arrival traffic volume is calculated by weighting the upstream arrival data for each time period. Therefore, the traffic volume arriving at the downstream stop line in each time period can be calculated.

[0123]

[0124] Where, q″ s,l,m q′ represents the total number of vehicles (PCU) arriving at the downstream stop line in lane l of section s within time period m, i.e., the downstream arrival flow after platoon discretization. s,l,m This represents the number of vehicles (PCU) that arrive at the starting point of the analyzed road segment within time period m in lane l, i.e., the upstream flow in the discrete model. F represents the travel time (i.e., the number of time segments) of the fastest vehicle in the convoy within the road segment containing section s. s The vehicle dispersion coefficient is used to characterize the degree of dispersion of vehicles during the driving process. The calculation formula is as follows:

[0125]

[0126] in, This represents the average travel time (i.e., the number of time segments) of the convoy on the road segment containing section s; It can be adjusted based on the average driving time, that is

[0127] The discrete platoon model can transform the traffic flow distribution at the starting point of a road segment into the traffic flow distribution at the stop line at the intersection. Compared with models that assume that the platoon arrives downstream at a fixed speed and uniformly, it better reflects the impact of individual differences on vehicle motion characteristics, thus more accurately reconstructing the actual traffic flow on the road segment.

[0128] (3) Traffic wave reconstruction

[0129] Based on the discrete traffic volume data for each time period obtained in step (2), this step further reconstructs the traffic waves for each time period. Unlike the traditional method of setting the arrival rate within a cycle to a fixed value, this method introduces the concept of a "spatial-temporal region," such as... Figure 3 As shown. The blue area represents a spatiotemporal region, that is, within a certain study period m, the average driving speed v of the road segment. s The flow of traffic reaches the downstream area. For example, for the m-th spatiotemporal region of section s, its range can be closed by the following four straight lines: front boundary, rear boundary, upper boundary, and lower boundary.

[0130]

[0131] Where x and t represent the ordinate and abscissa in the time-position two-dimensional coordinate system, respectively. s Indicates the location of section s, t m ,t m+1 Let x represent the starting times of the m-th and (n+1)-th spatiotemporal regions, respectively. s,up This indicates the location of the starting point of the road segment containing section s.

[0132] This method assumes that the vehicle arrival rate remains constant within each spatiotemporal region, but there are differences between different regions, and that the saturation dissipation rate of vehicles is constant. Therefore, the key point set of the dissipation wave... It includes two main points: the spatiotemporal coordinates of the stop line at the moment the green light turns on. The intersection of the converging wave and the dissipating wave The coordinates marking the start and end of the convoy's dissipation are respectively set. Simultaneously, the key point set of the rally wave... Including the intersection of the rallying wave with various spatiotemporal regions like Figure 4 As shown.

[0133] in, This indicates the starting time of lane l in section s during period k (with the red light turning on as the start time of the period); c s This represents the period (in seconds) of the intersection where section s is located; The green light duration (in seconds) for lane l in section s is represented. Let x and y represent the x-coordinate (s) and y-coordinate (m) of the end point of the dissipated wave of lane l in section s during period k, respectively. Let x and y represent the x-coordinate (s) and y-coordinate (m) of the intersection point of lane l in section s with the p-th spatiotemporal region in period k, respectively. Let S represent the x-coordinate (s) and y-coordinate (m) of the end point of the rally wave in section s of lane l during period k.

[0134] Then, the traffic q″ s,l,m Converted to flow rate, i.e., r″ s,l,m =q″ s,l,m / Δ t Based on this, the wave velocity of the rallied wave can be calculated, as shown in the following formula:

[0135]

[0136] in, The velocity of the condensed wave in the m-th spatiotemporal region of lane l in section s during period k is represented by r″. s,l,m k represents the arrival rate of lane l in the m-th spatiotemporal region within section s. q v represents the queuing density. s This represents the average driving speed from the starting point of the road segment to section s.

[0137] Considering different spatiotemporal regions, based on lane l in section s, each spatiotemporal region contains a corresponding rallying wave segment. These segments are connected end to end to form a complete rallying wave polyline, as shown below:

[0138]

[0139] Therefore, the starting point for calculating the rally wave is the spatiotemporal coordinates of the parking line position at the beginning of the cycle. When m = 0, its key points of the rally wave are... for Based on this, the rally wave segments are divided into two categories: the first category is the first rally wave segment within the period (i.e., m=1), whose calculation starting point is the spatiotemporal coordinates of the stop line at the beginning of the period; the second category is the rally waves after the first segment, whose calculation starting point is based on the calculation result of the previous rally wave segment, that is, derived through a recursive relationship. The key points of the (m+1)th rally wave segment can be recursively obtained based on the key points of the mth rally wave segment. The formula for calculating the key points of the rally wave is as follows:

[0140]

[0141] The spatiotemporal region numbering of the calculation starting point can be based on key points. The spatiotemporal region in which it is located is defined, that is Based on this formula, the location of the initial positioning point and its correspondence with the spatiotemporal region can be determined. Based on this, a mobilization wave is iteratively constructed, such as... Figure 5 As shown.

[0142] Subsequently, this invention assumes that the saturated vehicle traffic dissipates through the intersection at a fixed rate, thus allowing for the direct construction of a dissipation wave, as shown in the following equation. Wherein, This represents the dissipation wave velocity at section s.

[0143]

[0144] The intersection of the polyline of the assembling wave and the dissipating wave is the spacetime coordinate of the maximum queue length. Therefore, by combining the formulas for rallied waves and dissipated waves, the location of the maximum queue length can be calculated, as shown below.

[0145]

[0146] At the same time, the spatiotemporal region number where the maximum queue length is located can be calculated. And it satisfies the following constraints.

[0147]

[0148] Therefore, based on lane l in section s, the complete convective wave polyline, dissipation wave, and the location of the maximum queue length with period k can be obtained, such as... Figure 5 As shown.

[0149] (4) Dissipation traffic volume estimation

[0150] When the reconstructed cross section *s* is an external cross section, reconstruction can be performed directly using detector data. However, if cross section *s* is an internal cross section, its traffic input during its cycle depends not only on detector data but also on the signal timing of the upstream intersection. This is because the upstream signal timing affects the dissipation process of the upstream cross section, which in turn affects the arrival flow of the downstream cross section. Therefore, this step uses a dissipation traffic volume estimation method to provide the necessary input traffic volume data for the internal cross section in traffic wave reconstruction.

[0151] Based on the spatiotemporal coordinates of the maximum queue length calculated in step (3) The green light period for this section can be divided into two phases. The first phase is used for the saturation dissipation of queued vehicles, and the second phase is used for the arrival and departure of free traffic. Therefore, it is first necessary to calculate the boundary point between these two phases, as shown below.

[0152]

[0153] in, This represents the moment when the last vehicle in lane l in section s passes the stop line during period k, which is the dividing point between the two time periods.

[0154] Regarding the dissipation process of lane l in the studied cross section s during period k, it is assumed that the spatiotemporal region at the beginning of the period is numbered as follows: The spatiotemporal region where the maximum queue length is located is defined as... Vehicles arriving during this time period will queue in lane l of section s and will proceed in full capacity after the green light turns on. The total number of these vehicles can be expressed as follows:

[0155]

[0156] in, q″ represents the number of vehicles queuing in lane l within section s during period k. s,l,k,m This represents the number of vehicles arriving in lane l within the m-th spatiotemporal region of period k in section s.

[0157] Assume that the spatiotemporal region where the green light of lane l in section s turns on is numbered as follows: Furthermore, the spatiotemporal regions where the two sub-time period separators within the green light period are located are numbered as follows: Then queued vehicles During the time period to The flow is released evenly within the green light period, and this portion constitutes the dissipation flow during the first segment of the green light period; while the remaining portion of the green light period, i.e., from... to During this period, these vehicles will not stop, and their flow rate is related to the downstream arrival flow rate q″ calculated based on the fleet discrete model. s,l,k,m They are equal, as shown below.

[0158]

[0159] in, This represents the dissipation flow of lane l in section s within the m-th spatiotemporal region of period k.

[0160] Therefore, for the downstream internal cross-section that needs to be reconstructed, its three potential upstream input flows (left turn, straight, right turn) can be based on the above. The calculation formula is used to calculate the total flow rate, as shown below:

[0161]

[0162] Where, q′ s+1,l,m Let represent the flow rate of lane l reaching the starting point of the road segment in the m-th spatiotemporal region within the downstream internal section s+1, where s, x, and y represent the upstream section numbers where straight-through, left-turn, and right-turn traffic merges into the downstream flow direction at the upstream intersection, respectively. These represent the dissipation flow rates of the three flow directions in the m-th spatiotemporal region.

[0163] To obtain the flow rate q′ of the downstream internal cross section s+1,l,m Then, steps (1) to (2) can be repeated to reconstruct the traffic wave. The calculation process is the same as before, except that the flow rate input to the periodic arrival flow estimation step is different. Therefore, similar calculations can be performed on all downstream sections of the urban arterial road in sequence to form a continuous closed-loop modeling process and generate the average vehicle delay value for each combined scenario.

[0164] Step S4: Based on the scenario requiring diagnosis, define a cost function, use the cost function to calculate the signal control cost of each combined scenario, and obtain the cost matrix.

[0165] Assume the costs of optimizing the common cycle length, green light ratio, phase sequence, coordinated phase difference, and reversible lane function are c1, c2, c3, c4, and c5, respectively, and the original signal control scheme, i.e., the signal control scheme for the scenario requiring diagnosis, is S. O ={C O ,g o ,sq o ,δ o ,α o}, where C O ,g o ,sq o ,δ o ,α o These represent the common cycle length, green light ratio, phase sequence, coordinated phase difference, and variable lane function under the original signal control scheme, respectively. Based on this, a cost function can be constructed to calculate the cost of each combined scenario compared to the original control scheme.

[0166] The expression for the cost function is:

[0167] COST ij =c1·δ(C ij ≠C O )+c2·δ(g ij ≠g O )+c3·δ(sq ij ≠sq O )+c4·δ(δ ij ≠δ O )+c5·δ(α ij ≠α O )

[0168] Among them, C ij ,g ij ,sq ij ,δ ij ,α ij The values ​​of c1, c2, c3, c4, and c5 represent the common cycle length, green ratio, phase sequence, coordinated phase difference, and variable lane function parameters for the corresponding signal control scenario in the i-th row and j-th column, respectively. COST represents the cost of optimizing the common cycle length, green ratio, phase sequence, coordinated phase difference, and variable lane function. ij This represents the value in the i-th row and j-th column of the cost matrix, which is the cost of the i-th traffic demand scenario under the j-th signal control scenario.

[0169] Step S5: Based on the benefit matrix and cost matrix, generate the recommendation system feature matrix, and use the singular value decomposition method of the matrix considering bias to complete and correct the feature matrix of the recommendation system.

[0170] In the feature matrix of the recommender system, rows represent traffic demand scenarios, and columns represent control parameter scenarios. Each value in the matrix reflects the degree of preference for different control schemes under different traffic demand scenarios. Since the feature value is the benefit-cost ratio, the higher the value, the stronger the preference for that control scheme.

[0171] The process of generating the recommendation system's special value matrix based on the benefit matrix and cost matrix includes:

[0172] The benefit matrix and cost matrix are normalized respectively, and the benefit-cost ratio is calculated based on the normalized benefit matrix and cost matrix. The benefit-cost ratio is used as the feature value to construct the recommendation system feature value matrix.

[0173] The process of normalizing the benefit matrix includes:

[0174] Extract the minimum value E of all elements in the benefit matrix. min =min{E ij|i=1,2…M,j=1,2,…,N} and the maximum value E max =max{E ij |i=1,2…M,j=1,2,…,N}, and calculate according to the following formula:

[0175]

[0176] Among them, E′ ij This represents the normalized value of the control benefit for the i-th traffic demand scenario under the control scheme at the j-th sampling point, ranging from -1 to 1. The sign of this normalized value is consistent with the trend of benefit change, i.e., a negative value indicates a decrease in benefit, and a positive value indicates an increase in benefit.

[0177] The process of cost matrix normalization includes:

[0178] Calculate the minimum and maximum values ​​of all elements in the cost matrix, i.e., the cost. min =min{cost ij |i=1,2…M,j=1,2,…,N} and cost max =max{cost ij |i=1,2…M,j=1,2,…,N}, then, calculations are performed based on the following formula:

[0179]

[0180] Among them, cost′ ij This represents the normalized cost value of the i-th traffic demand scenario under the control scheme at the j-th sampling point, and its value ranges from 0 to 1.

[0181] V ij Let represent the characteristic value of the i-th traffic demand scenario under the j-th control scheme. The expression for calculating the benefit-cost ratio is:

[0182]

[0183] For each combination of scenarios, the benefit-cost ratio V N,i There are three possible values ​​for V: N,i <0: A negative benefit-cost ratio indicates that the control benefit of this signal control sampling scheme is lower than that of the original scheme; 0≤V N,i ≤1: A benefit-cost ratio between 0 and 1 indicates that the control efficiency of the signal control sampling scheme is improved compared to the original scheme, but the improvement in efficiency is insufficient to offset the increase in cost; that is, the efficiency has not kept pace with the cost. N,i >1: A benefit-cost ratio greater than 1 indicates that the control benefit of the signal control sampling scheme is improved compared to the original scheme, and is still significant when considering the cost.

[0184] The initial feature value matrix of the recommender system may contain some missing values ​​due to the sampling mechanism of each control parameter. Furthermore, since the feature values ​​in the benefit index matrix are based on relative values, the benefit index values ​​of various sampling point control schemes may be generally low when the original signal control scheme for an urban arterial road scenario is close to optimal; conversely, if the original scheme itself is unreasonable, the benefit index values ​​of the sampling point control schemes may be generally high. Secondly, the sampling process of the signal control scheme may include some unreasonable sampling point control schemes, and the benefit index values ​​of these points may not accurately reflect the actual situation. Therefore, it is necessary to introduce a bias term to complete and correct the feature value matrix of the recommender system.

[0185] The process of completing and correcting the feature matrix of a recommendation system based on the biased singular value decomposition method (BiasSVD algorithm) includes:

[0186] Set the bias values ​​of traffic demand scenario to signal control scenario and the bias values ​​of signal control scenario to traffic demand scenario, and construct an objective function that takes into account the bias values.

[0187] The characteristic values ​​of different traffic demands for different signal control schemes can be expressed in the following form:

[0188]

[0189] μ represents the system's average mean, b i Let b represent the bias term for the i-th traffic demand scenario. j This represents the bias term for the j-th signal-controlled scenario. p represents the latent vector of the j-th signal-controlled scenario. i Represents the latent vector of the i-th traffic demand scenario;

[0190] Similar to the objective functions of common supervised learning algorithms, the objective function of the biased singular value decomposition (SVD) method comprises two parts: a prediction error term and a regularization term. The prediction error term reflects the error between the predicted and true values ​​in the matrix, and the goal is to minimize this error, typically expressed as a least-squares loss function. The regularization term is used to prevent overfitting. The objective function considering the bias is expressed as follows:

[0191]

[0192] Where λ is the regularization coefficient, m ij Let represent the true characteristic values ​​of the i-th traffic demand scenario and the j-th signal control scenario, that is, the benefit-cost ratio of the i-th traffic demand scenario and the j-th signal control scenario. p represents the latent vector of the j-th signal-controlled scenario. iRepresents the latent vector of the i-th traffic demand scenario;

[0193] The objective function is solved iteratively using the gradient descent method. When the objective function value reaches its minimum, the resulting submatrix is ​​the optimal matrix decomposition result. The optimal latent vector and bias term are obtained, and the feature value matrix of the recommendation system is completed and corrected.

[0194] This invention discloses a solution method based on gradient descent, the main steps of which are as follows:

[0195] (1) Randomly initialize parameter p i ,q j ,b i ,b j It is usually initialized with a small random number or with a value of 0;

[0196] (2) Calculate the predicted value corresponding to each observation. The calculation can be performed based on the following formula:

[0197]

[0198] (3) The loss function value L under the current parameters can be calculated based on the following formula:

[0199]

[0200] (4) Calculate the gradient of the loss function value with respect to each parameter, as shown in the following formula;

[0201]

[0202] (5) Update each parameter as shown in the following formula;

[0203]

[0204] Where α is the learning rate, used to control the step size of each update.

[0205] (6) If the gradient descent method has converged, stop the calculation and obtain the final parameter values; otherwise, repeat steps (2) to (5). During this process, various convergence determination methods can be set. This invention selects the gradient norm as the determination criterion. When the norm of the gradient (i.e., the length of the gradient vector) is less than a preset threshold ∈, it is considered to be close to a local minimum, i.e.

[0206] Step S6: Quantify the similarity between the scenario to be diagnosed and each typical traffic demand scenario in the feature value matrix of the recommendation system, and calculate the reference value weight of the typical traffic demand scenario to the scenario to be diagnosed based on the similarity ratio.

[0207] The complete feature matrix V of the recommender systemMN This reflects the characteristic values ​​of various typical traffic demand scenarios under different signal control schemes at each sampling point; higher characteristic values ​​indicate better scheme performance. However, for any given new traffic demand scenario, due to the wide variety of path flow matrices, it is difficult to obtain the characteristic values ​​from the V matrix. MN Find the exact matching scenario. If the new requirement scenario is directly added as a new row to the feature matrix, a "cold start" problem may occur, meaning that the feature matrix may have no feature values, causing this method to fail.

[0208] Therefore, this method assumes that the traffic demand scenarios in the feature matrix can provide useful information for the scenarios that need to be diagnosed, and its reliability can be measured by the similarity between scenarios, and recommendations can be made by setting weights.

[0209] The expression for calculating the similarity between the diagnostic scenario and each typical traffic demand scenario in the feature matrix of the recommender system is as follows:

[0210]

[0211] Where sim(i,k) represents the similarity function between the scenario i to be diagnosed and the k-th typical traffic demand scenario. express, express;

[0212] The formula for calculating the reference value weight of typical traffic demand scenarios to the scenarios requiring diagnosis is as follows:

[0213]

[0214] Among them, w i,k Z represents the weight of scenario i to be diagnosed in relation to the k-th typical traffic demand scenario. i This represents the nearest neighbor set of scenario i that needs to be diagnosed, and its size is the number of typical traffic demand scenarios.

[0215] Step S7: Based on the reference value weight, calculate the probability of unreasonable settings of various signal control parameters in the scenario to be diagnosed, and realize the diagnosis of signal control problems according to the probability of unreasonable settings.

[0216] The formula for calculating the probability of unreasonable settings of various signal control parameters in the diagnostic scenario is as follows:

[0217]

[0218] X∈{C,g,sq,δ,α}

[0219] Where X represents the control parameter being analyzed, C, g, sq, δ, and α represent the common cycle length, green light ratio, phase sequence, coordinated phase difference, and variable lane function, respectively, and R XRepresents the potential causal probability of parameter X, card(·) is a function to calculate the number of elements in a finite set, I X This represents the number of control schemes for sampling points where parameter X in the signal control sampling scheme set has been changed. k,j This represents the value of the signal control scheme at the j-th sampling point in the k-th typical traffic demand scenario for parameter X.

[0220] For example, when X = C, R C I represents the potential causal probability of the common period length. C This indicates the number of sampling point control schemes where the common period length among the sampling points of the signal control scheme has changed.

[0221] Example 2

[0222] This embodiment, based on Embodiment 1 above, discloses a specific simulation verification process for a signal control problem diagnosis method for urban arterial roads. A simulation model of a certain arterial road in a city is established using VISSIM to verify the reliability of the problem diagnosis conclusions.

[0223] This embodiment selects two consecutive intersections on a main road in a certain city as the verification scenario. A schematic diagram of the main road area is shown below. Figure 6 As shown. Within the main line area, the sections are numbered clockwise, resulting in 6 outer sections and 2 inner sections, forming 56 paths. Both intersections use coordinated control, both with a traditional four-phase structure and a global phase difference of 39 seconds. Their respective signal timing schemes are as follows: Figure 7 As shown in Table 1. Furthermore, based on the actual collected trajectory data, this embodiment corrects the traffic flow based on penetration rate to obtain the final path traffic flow. On this basis, a new traffic demand scenario is randomly generated as the scenario requiring diagnosis, as shown in Table 2.

[0224] Table 1 Final Path Traffic Table

[0225]

[0226] Table 2: Path Traffic Table for Scenarios Requiring Diagnosis

[0227]

[0228] Since the true value of the problem diagnosis cannot be directly observed, this invention provides a verification method for the reliability of the problem diagnosis conclusion, such as... Figure 8As shown. First, the causal probability of the control parameters is calculated based on this invention. Then, using a method similar to the control variable method, a verification scheme for each type of optimization parameter is generated in the same trunk line coordination optimization model. For example, for a causal factor of unreasonable common cycle length, only the common cycle length is optimized, while other parameters remain unchanged, and corresponding constraints are set. Subsequently, these schemes are run in the same simulation model, and the benefit-cost ratio eigenvalues ​​compared to the original scenario are calculated.

[0229] Therefore, the validity of the verification conclusions can be judged from two aspects: First, the optimal parameter with the largest potential cause should correspond to the largest benefit-cost ratio; for example, if the cause probability of an unreasonable common period length is the highest, then the scheme that optimizes the common period length should show the largest benefit-cost ratio. Second, the cause probability of each signal control parameter and the benefit-cost ratio of the corresponding scheme should have a certain correlation. If the verification results conform to the above trends, the validity of the cause analysis probability can be indirectly verified.

[0230] Under this verification approach, the probability of unreasonable causes of each control parameter and its corresponding benefit-cost ratio are shown in Table 3.

[0231] Table 3 Problem Diagnosis Results

[0232]

[0233] As shown in Table 1, the problem diagnosis method constructed in this invention effectively locates potential optimization parameters. In the urban arterial road scenario studied, the probability of causing an unreasonable common cycle is 59.21%, significantly higher than the other four types of signal control parameters, with a benefit-cost ratio of 0.1137, the most significant eigenvalue. In contrast, the probability of causing an unreasonable green ratio, unreasonable coordinated phase difference, and unreasonable phase sequence are all lower in terms of both probability and benefit-cost ratio. This indicates that, in the verification results, the optimization parameter with the highest probability of causing an issue also has the highest benefit-cost ratio, meeting the first verification criterion. Further analysis of the four scenarios—common cycle length, green ratio, coordinated phase difference, and phase sequence—shows a Pearson correlation coefficient of 0.9079 between the benefit-cost ratio and the probability of causing an issue, indicating a strong correlation, meeting the second verification criterion.

[0234] Example 3

[0235] Based on Embodiment 1, this embodiment provides an electronic device, including: one or more processors and a memory, wherein the memory stores one or more programs, the one or more programs including instructions for executing the aforementioned diagnostic method for signal control problems of urban arterial roads.

[0236] At the hardware level, the electronic device includes a processor, internal bus, network interface, memory, and non-volatile memory, and may also include other hardware required for the business. The processor reads the corresponding computer program from the non-volatile memory into memory and then runs it to implement the aforementioned method for diagnosing signal control problems on urban arterial roads. Of course, in addition to software implementation, this invention does not exclude other implementation methods, such as logic devices or a combination of hardware and software, etc. That is to say, the execution entity of the following processing flow is not limited to individual logic units, but can also be hardware or logic devices.

[0237] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0238] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0239] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for diagnosing signal control problems on urban arterial roads, characterized in that, The method includes: Obtain the current diagnostic scenario that requires diagnosis; Considering the potential traffic demand and control parameters of urban arterial roads, typical traffic demand scenarios and signal control scenarios are generated. The typical traffic demand scenarios and signal control scenarios are then arbitrarily combined to obtain combined scenarios. Using the average vehicle delay on urban arterial roads as an evaluation indicator, the average vehicle delay value of combined scenarios is generated through numerical calculation or simulation. By comparing the average vehicle delay of each combined scenario with the average vehicle delay of the scenario to be diagnosed, the benefit matrix is ​​obtained. Based on the diagnostic scenario, a cost function is defined, and the signal control cost of each combined scenario is calculated using the cost function to obtain the cost matrix. Based on the benefit matrix and cost matrix, a recommendation system feature matrix is ​​generated, and the feature matrix is ​​completed and corrected using the matrix singular value decomposition method that considers bias. The similarity between the scenario requiring diagnosis and each typical traffic demand scenario in the feature value matrix of the recommendation system is quantified, and the reference value weight of the typical traffic demand scenario to the scenario requiring diagnosis is calculated based on the similarity ratio. Based on the aforementioned reference value weights, the probability of unreasonable settings of various signal control parameters in the scenario to be diagnosed is calculated, and signal control problem diagnosis is achieved based on the probability of unreasonable settings.

2. The method for diagnosing signal control problems on urban arterial roads according to claim 1, characterized in that, The traffic demand scenario is represented by a path flow matrix, which is composed of the flow of each path on the urban arterial road; the signal control scenario is formed by a combination of different signal control parameters, which include common cycle length, green ratio, coordinated phase difference, phase sequence and variable lane function.

3. The method for diagnosing signal control problems on urban arterial roads according to claim 1, characterized in that, The process of generating the typical traffic demand scenario includes: Each boundary point of the urban arterial road is regarded as a traffic origination point and a traffic attraction point, and the traffic generation volume of each traffic origination point and the traffic attraction volume of each traffic attraction point are set. Considering the different states of low and high traffic volume on urban arterial roads, a percentage fluctuation of traffic generation or attraction is set, the range of variation of traffic generation at each traffic generation point and traffic attraction at each traffic attraction point is obtained, and several sets of boundary generation and attraction combinations are constructed. Based on the combination of attraction quantities generated at the aforementioned boundaries, the path flow matrix of urban arterial roads is obtained using a traffic distribution method. Based on the combination of the attraction quantities at the boundary and their corresponding path flow matrices, a path flow tensor is constructed, and a preset number of typical traffic demand scenarios are obtained through a clustering algorithm. The process of generating the signal control scenario includes: Based on existing traffic standards, public cycle length parameters are set, including maximum cycle length and minimum cycle length; The green light ratio is set based on the NEMA double-ring structure, and the green light ratio includes the green light duration and the red light duration; Using the current main intersections of urban arterial roads as a reference, calculate the coordinated phase difference of other intersections. The minimum phase difference is 0 seconds, and the maximum is the common cycle length. For the current urban arterial road scenario, all reversible lane functions are enumerated, all possible values ​​are traversed, and the parameters of the reversible lane function are obtained. Based on the common cycle length parameter, green light ratio phase duration, intersection coordination phase difference, and variable lane function parameters, various signal control scenarios are generated.

4. The method for diagnosing signal control problems on urban arterial roads according to claim 1, characterized in that, The process of obtaining the benefit matrix includes: Using the average vehicle delay per vehicle on urban arterial roads as the evaluation index, the average vehicle delay value for each combined scenario is generated based on numerical calculation or simulation. The control effectiveness index matrix X is obtained by combining the average vehicle delay values ​​of each combined scenario. M·N ; Let the vehicle-average control delay in the scenario requiring diagnosis be x. O Based on the control benefit index matrix X M·N The benefit matrix is ​​calculated using the following expression: E ij =X ij -x O Among them, E ij X represents the value in the i-th row and j-th column of the benefit matrix, that is, the improvement in control benefit for the i-th traffic demand scenario under the j-th signal control scenario. ij This represents the value in the i-th row and j-th column of the benefit index matrix, that is, the average vehicle delay under the j-th signal control scenario for the i-th traffic demand scenario.

5. The method for diagnosing signal control problems on urban arterial roads according to claim 1, characterized in that, The expression for the cost function is: COST ij =c1·δ(C ij ≠C O )+c2·δ(g ij ≠g O )+c3·δ(sq ij ≠sq O )+c4·δ(δ ij ≠d O )+c5·δ(α ij ≠a O ) Among them, C ij ,g ij ,sq ij ,δ ij ,α ij Let C represent the common cycle length, green ratio, phase sequence, coordinated phase difference, and variable lane function parameters for the corresponding signal control scenario in the i-th row and j-th column, respectively. Let c1, c2, c3, c4, and c5 represent the costs of optimizing the common cycle length, green ratio, phase sequence, coordinated phase difference, and variable lane function, respectively. O ,g o ,sq o ,δ o and α o These represent the common cycle length, green light ratio, phase sequence, coordinated phase difference, and variable lane function, respectively, for reference in the diagnostic scenario. COST ij This represents the value in the i-th row and j-th column of the cost matrix, which is the cost of the i-th traffic demand scenario under the j-th signal control scenario.

6. The method for diagnosing signal control problems on urban arterial roads according to claim 1, characterized in that, The process of generating the recommendation system's feature value matrix based on the benefit matrix and cost matrix includes: The benefit matrix and cost matrix are normalized respectively, and the benefit-cost ratio is calculated based on the normalized benefit matrix and cost matrix. The benefit-cost ratio is used as the feature value to construct the recommendation system feature value matrix.

7. The method for diagnosing signal control problems on urban arterial roads according to claim 1, characterized in that, The process of completing and correcting the feature value matrix of the recommendation system includes: Set the bias values ​​of traffic demand scenario to signal control scenario and the bias values ​​of signal control scenario to traffic demand scenario, and construct an objective function that takes into account the bias values. The objective function is solved iteratively using the gradient descent method to obtain the optimal latent vector and bias term, and the feature value matrix of the recommendation system is then completed and corrected.

8. The method for diagnosing signal control problems on urban arterial roads according to claim 7, characterized in that, The objective function considering the deviation value is: Where μ represents the average mean of the system, b i Let b represent the bias term for the i-th traffic demand scenario. j Let λ represent the bias term for the j-th signal control scenario, where λ is the regularization coefficient, and m ij Let represent the true characteristic values ​​of the i-th traffic demand scenario and the j-th signal control scenario, that is, the benefit-cost ratio of the i-th traffic demand scenario and the j-th signal control scenario. p represents the latent vector of the j-th signal-controlled scenario. i Let represent the latent vector of the i-th traffic demand scenario.

9. The method for diagnosing signal control problems on urban arterial roads according to claim 1, characterized in that, The expression for calculating the similarity between the diagnostic scenario and each typical traffic demand scenario in the feature value matrix of the recommendation system is as follows: Where sim(i,k) represents the similarity function between the scenario i to be diagnosed and the k-th typical traffic demand scenario. express, express; The formula for calculating the reference value weight of the typical traffic demand scenario for the scenario requiring diagnosis is as follows: Among them, w i,k Z represents the weight of scenario i to be diagnosed in relation to the k-th typical traffic demand scenario. i This represents the nearest neighbor set of scenario i that needs to be diagnosed, and its size is the number of typical traffic demand scenarios.

10. A method for diagnosing signal control problems on urban arterial roads according to claim 1, characterized in that, The calculation expression for the probability of unreasonable settings of various signal control parameters in the scenario requiring diagnosis is as follows: Where X represents the control parameter being analyzed, C, g, sq, δ, and α represent the common cycle length, green light ratio, phase sequence, coordinated phase difference, and variable lane function, respectively, and R X Represents the potential causal probability of parameter X, card(·) is a function to calculate the number of elements in a finite set, I X This represents the number of control schemes for sampling points where parameter X in the signal control sampling scheme set has been changed. k,j This represents the value of the signal control scheme at the j-th sampling point in the k-th typical traffic demand scenario for parameter X.

Images