Single-point intersection signal control parameter optimization method based on recommendation system

Abstract

The invention relates to a single-point intersection signal control parameter optimization method based on a recommendation system, and the method comprises the following steps: obtaining the real-time traffic data of a to-be-optimized single-point intersection, inputting the real-time traffic data into the recommendation system, recognizing a corresponding most similar historical scene, calculating a benefit-cost ratio, and diagnosing the signal control parameters of the to-be-adjusted single-point intersection according to the benefit-cost ratio. A corresponding optimization scheme is generated; the construction process of the recommendation system specifically comprises the following steps that S101, historical traffic data are obtained, an initial multi-dimensional matrix is constructed, and elements in the initial multi-dimensional matrix represent the maximum vehicle average delay value of a specific traffic scene under a specific signal control scheme; s102, complementing uncovered traffic scenes in the initial multi-dimensional matrix by adopting a BiasSVD (Singular Value Decomposition) algorithm to generate a dense matrix; and S103, constructing a recommendation system according to the dense matrix. Compared with the prior art, the intersection signal control parameters with problems can be automatically diagnosed, and corresponding modification guidance is provided.

Description

Technical Field

[0001] The invention belongs to the field of traffic signal control, and in particular relates to a single-point intersection signal control parameter optimization method based on a recommendation system. Background Art

[0002] Optimizing signal control at urban intersections is a key link in improving urban traffic efficiency. When the original signal control scheme is close to the optimal, the benefits of optimizing all parameters will gradually decrease and the cost will be high. Therefore, engineers usually only select specific parameters for adjustment, and the selection of parameters depends largely on the experience and expertise of engineers. In order to achieve the transition from manual control to intelligent transportation systems, a method that can diagnose the parameters that affect intersection signal control is urgently needed as a bridge between traffic state perception evaluation and signal control optimization schemes, so as to accurately identify and analyze problems in traffic flow and put forward targeted optimization suggestions.

[0003] At present, the research on signal control problem diagnosis can be roughly divided into two categories. Among them, the first category mainly focuses on evaluating the rationality of each single signal control parameter, such as green light duration, signal cycle, etc. This type of method usually only diagnoses one parameter in each scenario. Due to the limitation of specific scenarios, its portability is weak, and most of them use advanced data sources, such as trajectory data of networked vehicles, which lacks applicability. The second category focuses on obtaining a comprehensive overview of signal control, comprehensively considering all evaluation indicators, and thus drawing conclusions such as congestion bottlenecks. For example, the Chinese invention patent with publication number CN115620518B proposes a deep learning-based intersection traffic conflict discrimination method, which can extract traffic parameters and traffic conflict indicators, making traffic safety analysis more systematic and comprehensive. Compared with single evaluation indicators such as queue length and vehicle delay, the diagnostic conclusions obtained by this method are more comprehensive, but they cannot provide clear guidance for the specific optimization of signal control parameters. Therefore, it is urgent to design a new single-point intersection signal control parameter method to automatically diagnose problematic intersection signal control parameters and provide corresponding modification guidance, which is of great practical significance for realizing intelligent traffic management and control. Summary of the invention

[0004] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and provide a single-point intersection signal control parameter optimization method based on a recommendation system, which can automatically diagnose problematic intersection signal control parameters and provide corresponding modification guidance.

[0005] The purpose of the present invention can be achieved by the following technical solutions:

[0006] The present invention provides a single-point intersection signal control parameter optimization method based on a recommendation system, characterized in that it comprises the following steps:

[0007] The real-time traffic data of the single-point intersection to be optimized is obtained and input into the recommendation system to identify the most similar historical scenario corresponding to the single-point intersection, and the benefit-cost ratio is calculated according to the maximum average vehicle delay value of the most similar historical scenario, and the signal control parameters of the single-point intersection that need to be adjusted are diagnosed accordingly, and a corresponding optimization plan is generated;

[0008] The construction process of the recommendation system is as follows:

[0009] S101, acquiring historical traffic data and constructing an initial multidimensional matrix, wherein the historical traffic data includes traffic scene channelization and flow data and signal control scheme information, and in the initial multidimensional matrix, the element in the i-th row and j-th column represents the maximum average vehicle delay value of traffic scene i under signal control scheme j;

[0010] S102, using a BiasSVD algorithm to complete uncovered traffic scenes in the initial multidimensional matrix to generate a dense matrix;

[0011] S103: construct a recommendation system according to the dense matrix.

[0012] Furthermore, in step S101, the maximum average vehicle delay value is calculated as follows:

[0013]

[0014] Among them, D is the maximum average vehicle delay value, S is the saturated flow, x is the saturation, g is the effective green light time, c is the cycle length, and q is the average vehicle arrival rate.

[0015] Furthermore, in step S102, the expression of the uncovered traffic scene in the initial multi-dimensional matrix data is completed using the BiasSVD algorithm as follows:

[0016]

[0017] in, is the characteristic value of different signal control schemes for different traffic demands, μ is the global average score, b i is the bias of traffic scene i, b j is the bias of signal control scheme j, p i is the potential feature vector of traffic scene i, q j is the potential feature vector of signal control scheme j.

[0018] Furthermore, in the process of completing uncovered traffic scenes, the objective function of the BiasSVD algorithm is as follows:

[0019]

[0020] Among them, λ is a regular term to prevent overfitting.

[0021] Furthermore, the calculation formula of the benefit-cost ratio is as follows:

[0022] R=B′ / C′

[0023] Where R is the benefit-cost ratio, B′ and C′ are the normalized benefit value B and cost value C respectively; the cost value C includes the direct cost C d and indirect costs C i , the direct cost C d is the total operating time cost, the indirect cost C i Cost calculation based on changes in different signal control parameters; the benefit value B is the intersection operation benefit brought about by the reduction of delay time after the traffic flow is improved.

[0024] Furthermore, the indirect cost C i The calculation formula is as follows:

[0025]

[0026] Among them, ω 1 ,ω 2 ,ω 3 and ω 4 are the cost weights required to change the cycle duration, phase number, phase sequence and green-to-signal ratio, respectively. 周期时长 , C 相序 and C 绿信比 are the modification costs of the control parameters cycle duration, phase sequence and green-to-signal ratio, It is an indicator variable. It takes the value of 1 when the green-to-signal ratio changes and causes the number of phases to change, and 0 otherwise.

[0027] Furthermore, the analytic hierarchy process is used to determine the specific values ​​of each cost weight.

[0028] Furthermore, a binary decision framework is introduced to estimate the modification cost of each control parameter.

[0029] Furthermore, the calculation formula of the benefit value B is as follows:

[0030] B=ΔT

[0031] Among them, ΔT is the reduction of the average delay time of the intersection, which is obtained by weighted summation of traffic volume:

[0032]

[0033] Among them, i is the traffic direction index, t iis the average delay of vehicles in the ith direction, Q i is the traffic volume in the ith direction.

[0034] Furthermore, the process of diagnosing and optimizing the single-point intersection signal control parameters that need to be adjusted according to the benefit-cost ratio is as follows:

[0035] Calculate the action adjustment ratio of the signal control parameters of the u-th single-point intersection, including increasing the adjustment ratio f increase，u , reduce the adjustment ratio f decrease，u and unchanged adjustment ratio f no_change，u :

[0036]

[0037]

[0038] Among them, R v is the benefit-cost ratio of the vth sampling point, Δp uv is the change of the u-th signal control parameter at the v-th sampling point, [·] is a Boolean expression, which takes 1 when the condition is true, otherwise it takes 0;

[0039] The adjustment action adjusts the signal control parameters of single-point intersections whose proportion exceeds the set threshold.

[0040] Compared with the prior art, the present invention has the following beneficial effects:

[0041] 1. The present invention proposes a method for optimizing signal control parameters of a single-point intersection based on a recommendation system. First, the real-time traffic data of the single-point intersection to be optimized is input into the recommendation system. The recommendation system can identify the most similar historical scene corresponding to the single-point intersection, and calculate the benefit-cost ratio according to the maximum average vehicle delay value of the most similar historical scene, which is used as a parameter to measure the pros and cons of the scheme. Based on this, the signal control parameters of the single-point intersection that need to be adjusted are diagnosed, and the corresponding optimization scheme is generated; wherein the construction process of the recommendation system is specifically as follows: S101, historical traffic data is obtained, and an initial multidimensional matrix is ​​constructed. The historical traffic data includes traffic scene channelization and flow The elements in the initial multidimensional matrix represent the maximum average vehicle delay value of a specific traffic scenario under a specific signal control scheme; S102, the BiasSVD algorithm is used to complete the uncovered traffic scenarios in the initial multidimensional matrix, and a dense matrix of single-point intersections that can cover a variety of scenarios is generated, which can effectively deal with the data sparsity problem, enhance the robustness of the recommendation system, and significantly improve the efficiency and accuracy of subsequent diagnosis of problem parameters; S103, a recommendation system is constructed based on the dense matrix; through the above method, the problem parameters in the signal control scheme can be automatically identified and clear and reasonable modification guidance can be provided.

[0042] 2. In the present invention, the signal control parameters of a single-point intersection that need to be adjusted are diagnosed based on the benefit-cost ratio, and are further optimized. Specifically, the benefit-cost ratio is the ratio of the normalized benefit value to the normalized cost value, and the cost value includes direct cost and indirect cost. The direct cost is the overall operating time cost, and the indirect cost is calculated based on the cost of changing different signal control parameters such as cycle duration, phase sequence and green-to-signal ratio. The benefit value is the reduction in the average delay time of the entire intersection, which is obtained by weighted summation of traffic volume; the action adjustment ratio of the signal control parameters of the single-point intersection is calculated based on the benefit-cost ratio and the change in the signal control parameters, including increasing the adjustment ratio, decreasing the adjustment ratio and keeping the adjustment ratio unchanged. The signal control parameters of the single-point intersection whose action adjustment ratio exceeds the set threshold are the parameters that need to be adjusted. Parameter fine-tuning based on the above method can maximize the benefit at the minimum adjustment cost.

[0043] 3. The recommendation system framework of the present invention has great scalability. It is not only applicable to a single intersection, but can also be extended to urban trunk roads and complex urban network environments, covering a variety of traffic management scenarios and having wide applicability. The present invention can not only be used to optimize signal timing, but can also be expanded to diagnose the rationality of other traffic control related parameters, such as lane marking settings, etc. It has good parameter extensibility and can further improve the overall effect of traffic management. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 is a schematic diagram of the method of the present invention;

[0045] Figure 2 It is a schematic diagram of the intersection for demonstration and simulation verification in the present invention;

[0046] Figure 3 It is a schematic diagram of the comparison of the delay reduction (seconds) between the experimental group and the control group under different signal parameter adjustment step sizes in the simulation verification of the present invention;

[0047] Figure 4 It is a schematic diagram of the influence of the initial matrix sample size on the accuracy of the recommendation system in the simulation verification (sensitivity analysis) of the present invention;

[0048] Figure 5 Schematic diagram of the impact of data sparsity on the accuracy of the recommendation system in the simulation verification (sensitivity analysis) of the present invention. DETAILED DESCRIPTION

[0049] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process, but the protection scope of the present invention is not limited to the following embodiments.

[0050] Example:

[0051] This embodiment provides a single-point intersection signal control parameter optimization method based on a recommendation system, such as Figure 1 As shown, the following steps are included:

[0052] S1. Build a recommendation system. Specifically, it includes the following steps:

[0053] S101. Combining historical traffic data and sampling technology, construct an initial multidimensional matrix.

[0054] Historical traffic data includes traffic scene channelization and flow data, signal control scheme information, etc. In the initial multidimensional matrix constructed based on this, the element in the i-th row and j-th column represents the maximum average vehicle delay value of traffic scene i under signal control scheme i. The calculation formula of the maximum average vehicle delay value is as follows:

[0055]

[0056] Among them, D is the maximum average vehicle delay value (s), S is the saturated flow, x is the saturation, g is the effective green light time (s), c is the cycle length (s), and q is the average vehicle arrival rate.

[0057] S102. Use the BiasSVD algorithm to complete the uncovered traffic scenes in the initial multi-dimensional matrix to generate a dense matrix.

[0058] Assume that the bias value of each demand scenario for each signal control solution is generally higher or lower than that of other scenarios. i , the bias value of each signal control scheme for each traffic demand scenario is generally higher or lower than that of other schemes, which is b j , the expression of using BiasSVD algorithm to complete the uncovered traffic scenes in the initial multidimensional matrix data is as follows:

[0059]

[0060] in, is the characteristic value of different signal control schemes for different traffic demands, μ is the global average score, b i is the bias of traffic scene i, b j is the bias of signal control scheme j, p i is the potential feature vector of traffic scene i, q j is the potential feature vector of signal control scheme j.

[0061] The BiasSVD algorithm aims to minimize the loss function:

[0062]

[0063] Among them, λ is a regular term to prevent overfitting.

[0064] This embodiment uses the gradient descent method to update the parameters in the BiasSVD algorithm. The specific process is as follows:

[0065] First, calculate the prediction error e ij :

[0066]

[0067] Then, each bias term and latent feature vector is updated:

[0068] b i ←b i +γ(e ij -δb i )

[0069] b j ←b j +γ(e ij -δb j )

[0070] p i ←p i +γ(e ij q j -δp i )

[0071] q j ←q j +γ(e ij p i -δq j )

[0072] Among them, y is the learning rate and δ is the regularization parameter.

[0073] S103. Construct a recommendation system based on the dense matrix.

[0074] S2. Input the real-time traffic data of the single-point intersection to be optimized into the recommendation system, diagnose the signal control parameters of the single-point intersection that need to be adjusted, and generate a corresponding optimization plan.

[0075] S201. The recommendation system identifies the most similar historical scenario of the single-point intersection to be optimized based on the Euclidean distance.

[0076] S202. Calculate the benefit-cost ratio R based on the delay performance of the most similar historical scenario, that is, the maximum average vehicle delay value D, as a parameter for measuring the pros and cons of the solution:

[0077] R=B′ / C′

[0078] Among them, B′ and C′ are the benefit value B and cost value C normalized by the min-max normalization method.

[0079] The calculation formula of benefit value B is as follows:

[0080] B=ΔT

[0081] Among them, ΔT is the reduction of the average delay time of the entire intersection, which is obtained by weighted summation of traffic volume, and the unit is s. In this embodiment, the traffic flow direction is the 8 flow directions divided by the NEMA ring fence structure:

[0082]

[0083] Among them, i is the traffic direction index, t i is the average delay of vehicles in the ith direction, Q i is the traffic volume in the i-th direction.

[0084] The cost value C is calculated as follows:

[0085] C=C d +C i

[0086] Among them, C d is the direct cost, is the total running time cost; C i It is an indirect cost, which can be further subdivided into the cost of changing different signal control parameters. In the calculation process, the cost of changing different parameters is distinguished by adding weights, as shown in the following formula:

[0087] C i =ω 1 ·C 周期时长 +ω 2 ·C 相位数 +ω 3 ·C 相序 +ω 4 ·C 绿信比

[0088] Among them, ω 1 ,ω 2 ,ω 3 and ω 4 are the cost weights required to change the cycle duration, phase number, phase sequence and green-to-signal ratio, respectively. 周期时长 , C 相位数 , C 相序 and C 绿信比 They are the modification costs of cycle duration, number of phases, phase sequence and green-to-signal ratio. In actual signal control scheme adjustment, the NEMA method for adjusting the green-to-signal ratio may result in changes in the number of phases, so the formula is adjusted to:

[0089] It is an indicator variable, which takes the value of 1 when the change in the green ratio causes a change in the number of phases, and 0 otherwise.

[0090] In this embodiment, the analytic hierarchy process (AHP) is used to determine ω 1 , ω 2 , ω 3 and ω 4 The specific values of are obtained, and the normalized weights are as follows: cycle length ω 1 = 0.118, number of phases ω 2 = 0.565, phase sequence ω 3 = 0.262, green light duration ω 4 = 0.055. The consistency ratio (CR) is 0.035, far lower than the generally accepted standard of 0.1, indicating that the constructed judgment matrix has good consistency.

[0091] For the cost estimation of the modification of the corresponding control parameters, this embodiment introduces a binary decision framework, aiming to accurately capture the impacts associated with the changes in the signal control parameters, as shown in the following formula:

[0092]

[0093] Among them, the assignment of each variable reflects whether the corresponding signal control parameter deviates from the original plan. If a certain parameter is adjusted, the corresponding decision variable is assigned a value of 1, indicating that such a change incurs additional cost; otherwise, it is assigned a value of 0, indicating that the parameter remains unchanged and no additional cost is generated.

[0094] The finally calculated benefit-cost ratio R is divided into three cases: if R < 0, it means that no benefit is generated under this scheme; if 0 < R < 1, it means that a certain improvement benefit is generated under this scheme, but a relatively large cost is required; if R > 1, it means that the scheme obtains greater benefits at a smaller cost and is economically feasible.

[0095] S203. Diagnose the single-point intersection signal control parameters that need to be adjusted according to the benefit-cost ratio, and generate corresponding optimization schemes.

[0096] The recommendation system extracts the part where the benefit-cost ratio is greater than 1, counts the sample points in the matrix sampling where the parameter changes compared with the original signal control scheme, and finally calculates the direction and proportion of the adjustment of the signal control parameter, and completes the optimization on this basis:

[0097]

[0098] Among them, f increase,u , f decrease,u and f no_change,uare the adjustment ratios of the u-th single-point intersection signal control parameter increase, decrease, and unchanged, respectively, R v is the benefit-cost ratio of the vth sampling point, Δp uv is the change of the uth signal control parameter at the vth sampling point, [·] is a Boolean expression, which takes 1 when the condition is true, otherwise it takes 0. Parameters with higher adjustment ratios are the main causes of traffic problems. Parameter fine-tuning based on the diagnosis results can maximize benefits at the minimum adjustment cost.

[0099] The above method uses historical data and sampling technology, takes traffic flow as demand, signal control parameters as supply, and the average vehicle delay at the intersection as the evaluation indicator to construct a multidimensional matrix representing the evaluation indicators under various traffic conditions. The BiasSVD algorithm is used to complete the uncovered scenarios in the matrix, and the matching between the traffic scenarios to be diagnosed and the historical data is analyzed based on the recommendation system, the most similar historical scenarios are identified, and the probability of unreasonable timing parameters is calculated. Automatic diagnosis of signal timing parameters at single-point intersections can be realized, which has the advantages of reducing manual intervention costs, improving diagnostic efficiency and accuracy, and wide applicability.

[0100] This embodiment is based on Figure 2 The above method was verified by the typical intersection of Hanjiang Road and Tongjiang Middle Road in Changzhou City, Jiangsu Province shown in the figure.

[0101] In this embodiment, the electric police data actually collected from 7:00 to 9:00 on January 8, 2019 is used to calibrate the flow of each entrance lane of the intersection. The phase sequence is as follows: Figure 2 First, based on the recommendation system, the potential problems of the intersection signal control parameters are diagnosed; then, according to the diagnosis results, the experimental group and the control group are set to adjust the signal control parameters in steps. The experimental diagnosis results are shown in the following table:

[0102] Table 1 Probability of adjusting control parameters given by the recommendation system

[0103] parameter Increased probability (%) Probability of reduction (%) Probability of unchanged (%) Green Time Ratio 1 67.05％ 0.00％ 32.95％ Green Time Ratio 2 15.15％ 64.02％ 20.83％ Green Time Ratio 3 35.98％ 35.23％ 28.79％ Green Time Ratio 4 45.83％ 27.27％ 26.89％ Green Time Ratio 5 67.05％ 0.00％ 32.95％ Green Time Ratio 6 15.15％ 64.02％ 20.83％ Green Time Ratio 7 35.98％ 35.23％ 28.79％ Green Time Ratio 8 45.83％ 27.27％ 26.89％

[0104] The empirical verification results are as follows Figure 3 As shown. The recommendation system will provide the probability that each parameter needs to be adjusted. Parameters with higher probabilities are the main causes of signal control problems. In this experiment, the probability that the green-to-signal ratio of lane numbered 1 needs to be increased is the highest, while the probability that the green-to-signal ratio of lane numbered 2 needs to be reduced is the highest, and it is determined to be the parameter that needs to be adjusted most. At the same time, in order to ensure that the number of phases remains unchanged, lanes numbered 5 and 6 also need to change synchronously according to the adjustment results of lanes 1 and 2. The experimental group adjusted the green-to-signal ratios of lanes 1, 2, 5, and 6, while the control group adjusted the green-to-signal ratios of lanes 3, 4, 7, and 8.

[0105] The empirical verification results show that the delay reduction rate of the experimental group is significantly higher than that of the control group. At the peak, the delay of the experimental group was reduced by 24.76%, while the control group was only reduced by 3.40%. This shows that adjusting the high probability parameters according to the suggestions can effectively reduce intersection delays, indicating that the signal control problem diagnosis accuracy of the present invention is high.

[0106] This embodiment also performs sensitivity analysis on the initial matrix sample size and data sparsity. By increasing the sample size, the richness of the training data of the recommendation system can be improved, thereby enhancing the generalization ability of the model. By adjusting the proportion of the training set under a fixed matrix size, the impact of data sparsity on the performance of the recommendation system is studied. This study selects the root mean square error (RMSE), mean absolute error (MAE), mean square error (MSE) and ranking consistency property (FCP) as verification indicators, which can be calculated according to the following formula:

[0107]

[0108] (1) Verify the impact of the initial matrix sample size on the accuracy of the recommendation system.

[0109] Signal control sampling and traffic flow sampling were used to generate initial matrices of different sizes. The BiasSVD algorithm was used to cross-validate samples of different sizes. The results are shown in Figure 2. Figure 4 As shown. Based on the empirical validation data, the filling accuracy of the recommendation system increases significantly with the increase in the number of samples of the recommendation system. As the sample size increases from 250,000 to 450,000, the RMSE, MAE, and MSE gradually decrease, indicating that the error of the recommendation system in filling the delay matrix decreases and the prediction accuracy improves. At the same time, the FCP index shows an upward trend, indicating that the ability of the recommendation system to capture user preferences is enhanced. These results show that as the number of samples increases, the refinement of the delay matrix increases, and the recommendation system can use richer data for training, thereby improving the filling accuracy and recommendation effect.

[0110] (2) Verify the impact of data sparsity on the accuracy of the recommendation system.

[0111] First, the sample size of the initial matrix is ​​fixed. During the cross-validation process, the cross-validation (CV) value is changed to form samples with different training ratios. The results are as follows: Figure 5 As shown in the figure. From the results of empirical verification, the accuracy of delayed matrix cross-validation is positively correlated with the cross-experimental subset data, indicating that the accuracy will be significantly improved by predicting fewer missing values ​​with richer data information. The FCP increases first and then decreases with the change of CV value. When there are too many training samples, it may cause overfitting of the model, thereby reducing the accuracy of the diagnostic method.

[0112] If the above method is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0113] The above description of the embodiments is to facilitate the understanding and use of the invention by those skilled in the art. It is obvious that those skilled in the art can easily make various modifications to these embodiments and apply the general principles described herein to other embodiments without creative work. Therefore, the present invention is not limited to the above embodiments, and improvements and modifications made by those skilled in the art based on the disclosure of the present invention without departing from the scope of the present invention should be within the scope of protection of the present invention.

Claims

1. A single-point intersection signal control parameter optimization method based on a recommendation system, characterized in that: The following steps are involved: The real-time traffic data of the single-point intersection to be optimized is obtained and input into the recommendation system to identify the most similar historical scenario corresponding to the single-point intersection, and the benefit-cost ratio is calculated according to the maximum average vehicle delay value of the most similar historical scenario, and the signal control parameters of the single-point intersection that need to be adjusted are diagnosed accordingly, and a corresponding optimization plan is generated; The construction process of the recommendation system is as follows: S101, acquiring historical traffic data and constructing an initial multidimensional matrix, wherein the historical traffic data includes traffic scene channelization and flow data and signal control scheme information, and in the initial multidimensional matrix, the element in the i-th row and j-th column represents the maximum average vehicle delay value of traffic scene i under signal control scheme j; S102, using a BiasSVD algorithm to complete uncovered traffic scenes in the initial multidimensional matrix to generate a dense matrix; S103: construct a recommendation system according to the dense matrix.

2. The single-point intersection signal control parameter optimization method based on a recommendation system according to claim 1, characterized in that: In step S101, the maximum average vehicle delay value is calculated as follows: Among them, D is the maximum average vehicle delay value, S is the saturated flow, x is the saturation, g is the effective green light time, c is the cycle length, and q is the average vehicle arrival rate.

3. The single-point intersection signal control parameter optimization method based on a recommendation system according to claim 1, characterized in that: In step S102, the BiasSVD algorithm is used to complete the expression of the uncovered traffic scene in the initial multi-dimensional matrix data as follows: in, is the characteristic value of different signal control schemes for different traffic demands, μ is the global average score, b i is the bias of traffic scene i, b j is the bias of signal control scheme j, p i is the potential feature vector of traffic scene i, q j is the potential feature vector of signal control scheme j.

4. The single-point intersection signal control parameter optimization method based on a recommendation system according to claim 3 is characterized in that: In the process of completing uncovered traffic scenes, the objective function of the BiasSVD algorithm is as follows: Among them, λ is a regular term to prevent overfitting.

5. The single-point intersection signal control parameter optimization method based on a recommendation system according to claim 1, characterized in that: The calculation formula of the benefit-cost ratio is as follows: R=B ′ / C ′ Among them, R is the benefit-cost ratio, B ′ and C ′ are the normalized benefit value B and cost value C respectively; the cost value C includes the direct cost C d and indirect costs C i , the direct cost C d is the total operating time cost, the indirect cost C i Cost calculation based on changes in different signal control parameters; the benefit value B is the intersection operation benefit brought about by the reduction of delay time after the traffic flow is improved.

6. The single-point intersection signal control parameter optimization method based on a recommendation system according to claim 5, characterized in that: The indirect cost C i The calculation formula is as follows: Among them, ω1, ω2, ω3 and ω4 are the cost weights required to change the cycle length, number of phases, phase sequence and green-to-signal ratio respectively. 周期时长 , C 相序 and C 绿信比 are the modification costs of the control parameters cycle duration, phase sequence and green-to-signal ratio, It is an indicator variable. It takes the value of 1 when the green-to-signal ratio changes and causes the number of phases to change, and 0 otherwise.

7. The single-point intersection signal control parameter optimization method based on a recommendation system according to claim 6, characterized in that: The analytic hierarchy process is used to determine the specific value of each cost weight.

8. The single-point intersection signal control parameter optimization method based on a recommendation system according to claim 6, characterized in that: A binary decision framework is introduced to estimate the cost of modifying each control parameter.

9. The single-point intersection signal control parameter optimization method based on a recommendation system according to claim 5, characterized in that: The calculation formula of the benefit value B is as follows: B=ΔT Among them, ΔT is the reduction of the average delay time of the intersection, which is obtained by weighted summation of traffic volume: Among them, i is the traffic direction index, t i is the average delay of vehicles in the ith direction, Q i is the traffic volume in the ith direction.

10. The single-point intersection signal control parameter optimization method based on a recommendation system according to claim 1, characterized in that: The specific process of diagnosing and optimizing the single-point intersection signal control parameters that need to be adjusted according to the benefit-cost ratio is as follows: Calculate the action adjustment ratio of the signal control parameters of the u-th single-point intersection, including increasing the adjustment ratio f increase,u , reduce the adjustment ratio f decrease,u and unchanged adjustment ratio f no_change,u : Among them, R v is the benefit-cost ratio of the vth sampling point, Δp uv is the change of the u-th signal control parameter at the v-th sampling point, [·] is a Boolean expression, which takes 1 when the condition is true, otherwise it takes 0; The adjustment action adjusts the signal control parameters of single-point intersections whose proportion exceeds the set threshold.

Images