Traffic signal optimization method, system and equipment based on semi-intelligent prediction-optimization

Abstract

The invention provides a traffic signal optimization method, system and equipment based on semi-intelligent prediction-optimization. The method comprises the following steps: collecting traffic flow data of intersections in different time periods; a plurality of signal control schemes are designed based on the traffic flow data, a signal scheme set is formed, and each signal control scheme comprises different phase sequences, signal periods and green signal ratios; constructing a signal scheme optimization model and a model loss function; training the signal scheme optimization model based on the traffic flow data and the model loss function to obtain a signal scheme selection optimization model; and selecting an optimal signal control scheme in the signal scheme set by using the signal scheme selection optimization model, and generating an optimal signal control scheme sequence of multiple periods of the intersection. According to the method, the accuracy of multi-period signal scheme selection can be remarkably improved, so that the vehicle delay at the intersection is reduced, and the vehicle passing efficiency is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of urban road traffic and relates to a traffic signal optimization method, system and device based on semi-intelligent prediction-optimization. Background Art

[0002] Urban traffic control systems are one of the important factors affecting traffic operation and also play a huge role in promoting the development of cities. The single intersection signal control methods mainly include three categories: signal-controlled intersection multi-period (Time-of-day, TOD) timing control, inductive control and adaptive control. In terms of multi-period timing control, in current practical applications, most intersections in China are mainly controlled by multi-period timing control. Multi-period timing control mainly includes two aspects: multi-period division and signal timing. In terms of signal timing optimization, most current researchers construct optimization models with single or comprehensive consideration of multiple performance indicators (intersection delay, intersection capacity, number of stops, queue length, fuel consumption, etc.) as optimization objectives, and combine constraint conditions to optimize parameters such as cycle length, green light duration and phase difference of intersections.

[0003] The above multi-period timing control has a prerequisite, that is, it is assumed that the traffic flow in the controlled period can be obtained (usually the historical average value or obtained through prediction methods). In such a case, the problem to be solved in each period of multi-period timing control can be regarded as a two-stage problem of first predicting the traffic flow and then formulating the optimal signal plan for the intersection according to the traffic flow arrival situation. In actual situations, the process of formulating the optimal signal plan for this two-stage problem usually selects the most suitable signal plan for the current traffic state from a limited set of plans (for example, the signal plan that minimizes the intersection delay). For this two-stage problem, previous studies were almost all carried out separately, that is, traffic prediction was carried out first and then plan selection was carried out, while the present invention integrates these two processes to provide a traffic signal optimization method, system and device based on semi-intelligent prediction-optimization. Summary of the Invention

[0004] Aiming at the deficiencies of the prior art, the present invention provides a traffic signal optimization method, system and device based on semi-intelligent prediction-optimization.

[0005] The technical solution adopted by the present invention is as follows:

[0006] A traffic signal optimization method based on semi-intelligent prediction-optimization includes the following steps:

[0007] S1. Collect traffic flow data of intersections at different times;

[0008] S2. Design several signal control schemes based on traffic flow data to form a signal scheme set. Each signal control scheme includes different phase sequences, signal cycles, and green ratios.

[0009] S3. Construct a signal scheme optimization model and a model loss function.

[0010] S4. Train the signal scheme optimization model based on traffic flow data and the model loss function to obtain an optimized signal scheme selection model.

[0011] S5. Use the optimized signal scheme selection model to select the optimal signal control scheme from the signal scheme set and generate an optimal signal control scheme sequence for multiple time periods at the intersection.

[0012] Further, the specific steps for forming the signal scheme set are as follows:

[0013] Consider the arrival characteristics of traffic flows in different control time periods at the intersection to determine an optional subset of phase sequences. The phase sequence determines the organizational structure and release order of each direction at the intersection.

[0014] Take values around the optimal cycle in a specific traffic flow scenario to determine an optional subset of signal cycles. The optimal cycle is calculated according to the HCM signal timing method.

[0015] Consider the actual situation to determine an optional subset of green ratios and delete green ratio schemes that do not conform to the actual traffic situation. The green ratio determines the release time weight of different phases.

[0016] After determining all feasible subsets of phase sequences, signal cycles, and green ratios, combine the three to obtain the final signal scheme set.

[0017] Further, the signal control scheme satisfies the double-loop structure signal timing method. Specifically, the green light time lengths of each phase should satisfy the following limiting conditions:

[0018]

[0019]

[0020]

[0021]

[0022]

[0023]

[0024]

[0025] Among them, Denotes the green light time of the i-th phase, where i = 1, 2, 3, ……, 8; Denotes the minimum green light time of the i-th phase. The minimum green light time takes into account the pedestrian crossing time and the minimum time required to clear the queued vehicles in this phase. The timing is determined by judging whether the green light time displayed for each straight-ahead phase is greater than the pedestrian crossing time. If the green light time displayed for each straight-ahead phase is greater than the pedestrian crossing time, then the timing does not need to be modified; otherwise, the timing needs to be modified. Denotes the maximum green light time, which takes into account the maximum waiting time of pedestrians.

[0026] Furthermore, the signal plan optimization model is specifically as follows:

[0027] To achieve the best signal control effect throughout the day, the sum of the total vehicle delays generated by the signal plans adopted in each time period should be as small as possible. Therefore, for a single control period in a day, the signal plan that best matches the arrival of vehicle flows during this period should be selected as the signal control plan for this period. Based on this, a binary decision variable is defined. If signal plan m is selected as the signal control plan for control period h, then the decision variable is set to 1; otherwise, it is 0. The signal plan selection model can be represented by the mathematical model M1:

[0028] Optimization objective M1:

[0029]

[0030] Constraint conditions:

[0031]

[0032]

[0033] Among them, h represents a certain time period in a day, and H represents the total number of time periods in a day; m represents a certain signal plan, and M represents the total number of plans in the phase sequence plan set; Denotes the total vehicle delay generated at the intersection when phase sequence plan m is adopted in time period h; is a function of Denotes the traffic flow vector of each approach at the intersection in time period h.

[0034] The optimization objective M1 minimizes the total vehicle delay at the intersection, that is, minimizes the sum of the total vehicle delays generated by the phase sequence plans selected for all time periods. The first constraint ensures that one phase sequence plan is assigned to each time period, and the second constraint ensures the domain of the decision variable.

[0035] In the optimization objective M1, the coefficient of the decision variable is , is obtained by first predicting the traffic flow , and then through the delay formula. Considering that for a specific signal plan, traffic flow is the most critical factor affecting the total vehicle delay, and the accurate value of vehicle delay under different time periods and different signal plans can be obtained through historical traffic flow and the delay formula. Therefore, the vehicle delay of each time period and each signal plan for the next day can be directly predicted based on the historical delays of each signal plan .

[0036]

[0037] Among them, is the initial loss function, is the number of signal plans, is the actual vehicle delay under the i-th signal plan, is the predicted vehicle delay under the i-th signal plan.

[0038] Furthermore, while minimizing the mean square error, the model loss function considers reducing the prediction error between every two signal control plans to directly predict the total intersection vehicle delay under each signal optimization plan. The model loss function is:

[0039]

[0040] Among them, represents the sum of the squares of the differences between the predicted value and the actual value of the total intersection vehicle delay, represents the sum of the squares of the differences between the predicted value and the actual value of the total vehicle delay under every two plans and then subtracting them; and respectively represent the actual value and the predicted value of the total intersection vehicle delay of the i-th plan. The first term represents minimizing the sum of the squares of the differences between the predicted value and the actual value of the total intersection vehicle delay, and the second term represents minimizing the sum of the squares of the differences between the predicted value and the actual value of the total vehicle delay under every two plans and then subtracting them. That is, the optimization objective of

[0041] While minimizing the difference between the predicted value and the actual value, it also considers reducing the prediction error between all plans.

[0042]

[0043] Constraint conditions:

[0044]

[0045]

[0046] Among them, h represents a certain period of a day, and H represents the total number of periods in a day; m represents a certain signal plan, and M represents the total number of plans in the phase sequence plan set. It represents the total vehicle delay generated at the intersection when using the phase sequence plan m in the period h. is a function of It represents the traffic flow vector of each approach of the intersection in the period h.

[0047] Furthermore, for a single control period in a day, the signal plan selection optimization model should select the signal control plan that minimizes the total vehicle delay at the intersection during this period as the optimal signal control plan.

[0048] A traffic signal optimization system based on semi-intelligent prediction-optimization, including:

[0049] Data acquisition module: used to collect traffic flow data of intersections at different times;

[0050] Scheme design module: used to design several signal control schemes based on traffic flow data to form a signal plan set, and each signal control scheme includes different phase sequences, signal cycles, and green signal ratios;

[0051] Model construction module: used to construct a signal plan optimization model and a model loss function;

[0052] Model training module: used to train the signal plan optimization model based on traffic flow data and the model loss function to obtain a signal plan selection optimization model;

[0053] Scheme selection module: used to use the signal plan selection optimization model to select the optimal signal control plan in the signal plan set and generate an optimal signal control plan sequence for multiple periods of the intersection.

[0054] A computer device, the computer device includes:

[0055] One or more processors;

[0056] A memory for storing one or more programs;

[0057] When the one or more programs are executed by the one or more processors, the one or more processors implement the above-mentioned traffic signal optimization method based on semi-intelligent prediction-optimization.

[0058] A computer-readable storage medium storing computer instructions that, when executed by one or more processors, cause the one or more processors to perform the steps in the above method.

[0059] Compared with the prior art, the present invention has the following advantages:

[0060] 1. The method of the present invention constructs a loss function driven by an optimization objective function and reconstructs the optimization objective based on this loss function, with excellent effects, superior to the existing conventional method of first predicting traffic flow and then optimizing the signal plan.

[0061] 2. The present invention can be applied to the actual situation of traffic arrival changes, and the greater the traffic arrival changes, the more obvious the advantages of the proposed model become.

[0062] 3. The framework proposed by the present invention can be easily migrated to any two-stage traffic problem of first predicting and then optimizing, with strong method migration ability. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 It is a schematic diagram of an intersection in an embodiment of the present invention.

[0064] Figure 2 It is a schematic diagram of a traditional prediction and optimization framework.

[0065] Figure 3 It is a schematic diagram of the SPO prediction and optimization framework in an embodiment of the present invention.

[0066] Figure 4 It is a phase sequence plan diagram in the signal plan set in an embodiment of the present invention.

[0067] Figure 5 It is a schematic diagram of the data structure of each step of SPO prediction and optimization in an embodiment of the present invention.

[0068] Figure 6 It is a diagram of the optimization results of the signal plan at each time period under different methods in an embodiment of the present invention.

[0069] Figure 7 It is a diagram of the absolute error between the scheme obtained by different methods and the optimal scheme in an embodiment of the present invention.

[0070] Figure 8 It is a diagram of the delay error between two schemes in the schemes obtained by different methods in an embodiment of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0071] The technical solutions of the present invention will be further clearly and detailedly described below with reference to the drawings and specific examples.

[0072] Embodiment

[0073] A traffic signal optimization method based on semi-intelligent prediction-optimization, comprising the following steps:

[0074] S1. Collect traffic flow data at intersections during different time periods;

[0075] S2. Design a number of signal control schemes based on the traffic flow data to form a signal scheme set, and each signal control scheme includes different phase sequences, signal cycles, and green signal ratios;

[0076] S3. Construct a signal scheme optimization model and a model loss function;

[0077] S4. Train the signal scheme optimization model based on the traffic flow data and the model loss function to obtain a signal scheme selection optimization model;

[0078] S5. Use the signal scheme selection optimization model to select the optimal signal control scheme from the signal scheme set, and generate an optimal signal control scheme sequence for multiple time periods at the intersection.

[0079] Further, the specific steps for forming the signal scheme set are as follows:

[0080] Consider the arrival characteristics of traffic flows during different control time periods at the intersection, and determine an optional subset of phase sequences, where the phase sequence determines the organizational structure and release order of each direction at the intersection;

[0081] Take values around the optimal cycle in a specific traffic flow scenario to determine an optional subset of signal cycles; the optimal cycle is calculated according to the HCM signal timing method;

[0082] Consider the actual situation to determine an optional subset of green signal ratios, and delete green signal ratio schemes that do not conform to the actual traffic situation, where the green signal ratio determines the release time weight of different phases;

[0083] After determining all feasible subsets of phase sequences, signal cycles, and green signal ratios, combine the three to obtain the final signal scheme set.

[0084] Further, the signal control scheme satisfies the double-loop structure signal timing method. Specifically, the green light time lengths of each phase should satisfy the following limiting conditions:

[0085]

[0086]

[0087]

[0088]

[0089]

[0090]

[0091]

[0092] Among them, represents the green light time of the i-th phase, where i = 1, 2, 3, ……, 8; represents the minimum green light time of the i-th phase. The minimum green light time is the minimum time considering the pedestrian crossing time and the clearance of queued vehicles in this phase. Whether to modify the timing is determined by judging whether the green light time displayed in each straight-through phase is greater than the pedestrian crossing time. When the green light time displayed in each straight-through phase is greater than the pedestrian crossing time, the timing does not need to be changed; otherwise, the timing needs to be changed; represents the maximum green light time, and the maximum green light time considers the maximum waiting time of pedestrians.

[0093] Furthermore, the signal plan optimization model is specifically as follows:

[0094] To achieve the best signal control effect throughout the day, the sum of the total vehicle delays generated by the signal plan adopted in each time period should be as small as possible. Therefore, for a single control period in a day, the signal plan that best matches the traffic flow arrival in this period should be selected as the signal control plan for this period. Based on this, the binary decision variable is defined. If signal plan m is selected as the signal control plan for control period h, the decision variable is set to 1; otherwise, it is 0. The signal plan selection model can be represented by the mathematical model M1:

[0095] Optimization objective M1:

[0096]

[0097] Constraint conditions:

[0098]

[0099]

[0100] Among them, h represents a certain period in a day, and H represents the total number of periods in a day; m represents a certain signal plan, and M represents the total number of plans in the phase sequence plan set; represents the total vehicle delay generated at the intersection when phase sequence plan m is adopted in period h; is a function of represents the traffic flow vector of each approach at the intersection in period h.

[0101] The optimization objective M1 minimizes the total vehicle delay at intersections, that is, minimizes the sum of the total vehicle delays generated by the phase sequence schemes selected for all time periods. The first constraint ensures that one phase sequence scheme is assigned to each time period, and the second constraint ensures the domain of the decision variables.

[0102] In the optimization objective M1, the coefficient of the decision variable is , which is obtained by first predicting the traffic flow , and then obtained through the delay formula. Considering that for a specific signal scheme, traffic flow is the most critical factor affecting the total vehicle delay, and the accurate values of vehicle delays under different time periods and different signal schemes can be obtained through historical traffic flow and the delay formula. Therefore, the vehicle delays of each time period and each signal scheme for the next day can be directly predicted based on the historical delays of each signal scheme. .

[0103]

[0104] Among them, is the initial loss function, is the number of signal schemes, is the actual vehicle delay under the i-th signal scheme, is the predicted vehicle delay under the i-th signal scheme.

[0105] Furthermore, while minimizing the mean squared error (MSE), the model loss function considers reducing the prediction error between every two signal control schemes to directly predict the total vehicle delay at intersections for all signal optimization schemes. The model loss function is:

[0106]

[0107] Among them, represents the sum of the squares of the differences between the predicted value and the actual value of the total vehicle delay at intersections, represents the sum of the squares of the differences obtained by subtracting the predicted value and the actual value of the total vehicle delay under every two schemes; and respectively represent the actual value and the predicted value of the total vehicle delay at intersections of the i-th scheme. The first term represents minimizing the sum of the squares of the differences between the predicted value and the actual value of the total vehicle delay at intersections, and the second term represents minimizing the sum of the squares of the differences obtained by subtracting the predicted value and the actual value of the total vehicle delay under every two schemes. That is, the optimization objective of Refer to the delay model given in HCM2010. This model is applicable to intersections where the lane group saturation X is less than 1. When calculating the delay using the HCM method, the lane group is divided into lanes. The delay calculation model based on the lane group is as follows:

[0108]

[0109]

[0110]

[0111] The total delay d of approach A within one cycle and the total delay D of the intersection are respectively:

[0112]

[0113]

[0114] In the formula, d is the uniform control delay; g is the effective green time of the lane group; C is the signal cycle; X is the lane group saturation, X = qi / s, where qi is the sum of the actual flow rates of the lane group and s is the saturation flow rate of lane group j; d is the incremental delay; T is the duration of the analysis period; k is the incremental delay parameter, related to the control setting; P is the correction coefficient screened or measured upstream of the intersection; c is the capacity of the lane group; d j is the average control delay of lane group j; PF is the signal linkage coefficient. Instructions for selecting PF under different arrival types are given in HCM2010, and generally 1.00 is taken; is the total delay of the intersection.

[0115] Furthermore, based on the traffic flow data and the model loss function, the signal plan optimization model is trained to obtain the signal plan selection optimization model M2. The signal plan selection optimization model M2 can be expressed as:

[0116]

[0117] Constraints:

[0118]

[0119]

[0120] Among them, h represents a certain time period in a day, and H represents the total number of time periods in a day; m represents a certain signal plan, and M represents the total number of plans in the phase sequence plan set; represents the total vehicle delay generated at the intersection when using the phase sequence plan m during time period h. is a function of Denote the traffic flow vector of each approach at the intersection during period h. In the optimization model M2, the value to be predicted is , usually, the prediction goal is to minimize the sum of the squared errors (MSE) between the predicted value and the actual value. By observing the optimization model M2, it can be found that if the same value is overestimated or underestimated for , the choice of the phase plan is not affected ( The predicted value is used to guide the selection of the signal plan, but is not used as a parameter of the signal plan), so the loss function is derived from the structure of the optimization model.

[0121] Furthermore, for a single control period in a day, the signal plan selection optimization model should select the signal control plan that minimizes the total vehicle delay at the intersection during this period as the optimal signal control plan. Since the signal plan selection optimization model is integrated into the loss function of the prediction process and directly predicts the delays of different signal plans, the optimization result can directly generate the signal plans for different periods. The present invention can significantly improve the accuracy of multi-period signal plan selection, thereby reducing the vehicle delay at the intersection and improving the vehicle passing efficiency.

[0122] To verify the application effect of the proposed scheme in practice, two intersections are selected for case analysis. Both of these two intersections are typical four-phase signal control intersections, as Figure 1 shown. The lane-level traffic flows with a 3-minute granularity are obtained through video detection from August 3, 2020 to August 31, 2020. When processing the data, the traffic flow data on Saturdays and Sundays are deleted, and the other data are reconstituted into time series data with a time length of 21 days. The data of the first 20 days are used as historical prediction data, and the data of the 21st day are used as verification data.

[0123] The signal plans of the two intersections are optimized under five methods respectively below, and the selection of the plans for 24 periods under each method is specifically compared to analyze its consistency with the theoretical optimal plan; the absolute error value of the total vehicle delay between it and the theoretical optimal plan is specifically compared to analyze the delay difference between it and the theoretical optimal plan. Figure 2 And Figure 3 are respectively the schematic diagrams of the traditional prediction and optimization framework and the SPO prediction and optimization framework in the embodiment of the present invention; Figure 4 is the phase sequence plan diagram in the signal plan set in the embodiment of the present invention; Figure 5 is the schematic diagram of the data structure of each step of SPO prediction and optimization in the embodiment of the present invention.

[0124] Figure 6It details the scenario of the third group of data in terms of the scheme selection under three proposed schemes and two control schemes. Among them, FS is the fixed scheme, that is, the first scheme in the scheme set (a total of 12 schemes) is selected for all 24 control periods in a day; BS is the theoretically optimal scheme. Figure 7 It is the absolute value of the error between the five schemes and the theoretically optimal scheme. It can be seen that the schemes for all 24 periods generated by FS are not the optimal schemes, and the error between it and the theoretically optimal scheme is significantly greater than that of other schemes; for R1 and R2, the optimal schemes are selected for 15 control periods, but the total vehicle delay error of R2 is less than that of R1; for R3, the optimal schemes are selected for 21 periods, and compared with FS, R1, and R2, its total vehicle delay error is the smallest.

[0125] Since the objective of the loss function in the prediction process of method R3 in the article is to minimize the error between the schemes in the scheme set, the differences between the schemes of the optimization results of R1, R2, and R3 will be specifically compared below, so as to show the differences between scheme R3 and R1, R2 from the perspective of the optimization results.

[0126] Figure 8 It is the result of the scheme comparison of R1, R2, and R3. The elements in each subgraph (all symmetric matrices) represent the differences between every two schemes. By calculating, the average differences between the schemes of R1, R2, and R3 are 0.133, 0.204, and 0.122 respectively, and R3 is the smallest. Thus, it can be seen that the optimization result of R3 is consistent with the expectation of the loss function in its prediction process. It should be noted that due to the randomness of the LSTM method itself, the difference between every two schemes of R3 is not less than that of R1 and R2, but on average, the difference between the schemes of R3 is the smallest.

[0127] Embodiment 2

[0128] A traffic signal optimization system based on semi-intelligent prediction-optimization, comprising:

[0129] A data collection module: used to collect traffic flow data at intersections in different periods;

[0130] A scheme design module: used to design a number of signal control schemes based on traffic flow data to form a signal scheme set, and each signal control scheme includes different phase sequences, signal cycles, and green signal ratios;

[0131] A model construction module: used to construct a signal scheme optimization model and a model loss function;

[0132] A model training module: used to train the signal scheme optimization model based on traffic flow data and the model loss function to obtain a signal scheme selection optimization model;

[0133] Scheme selection module: used to select the optimal signal control scheme from the signal scheme set by using the signal scheme selection optimization model, and generate the optimal signal control scheme sequence for multiple time periods at the intersection.

[0134] Those skilled in the art should understand that the embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0135] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to the embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram can be implemented by computer program instructions, and the combination of the processes and / or blocks in the flowchart and / or block diagram can also be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate a device for implementing the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0136] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured article including an instruction device, and the instruction device implements the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0137] These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process. Therefore, the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0138] The above is only the preferred embodiment of the present invention. Although the present invention has been disclosed above with preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make many possible changes and modifications to the technical solution of the present invention, or modify it into an equivalent embodiment with equivalent changes, without departing from the scope of the technical solution of the present invention. Therefore, any simple modification, equivalent change and modification made to the above embodiments according to the technical essence of the present invention without departing from the content of the technical solution of the present invention still fall within the scope of protection of the technical solution of the present invention.

Claims

1. A traffic signal optimization method based on semi-intelligent prediction-optimization, characterized in that It includes the following steps: S1. Collect traffic flow data at intersections at different time periods; S2. Design several signal control schemes based on the traffic flow data to form a signal scheme set, and each signal control scheme includes different phase sequences, signal cycles, and green ratios; S3. Construct a signal scheme optimization model and a model loss function; S4. Train the signal scheme optimization model based on the traffic flow data and the model loss function to obtain a signal scheme selection optimization model; S5. Use the signal scheme selection optimization model to select the optimal signal control scheme from the signal scheme set to generate an optimal signal control scheme sequence for the intersection at multiple time periods.

2. The traffic signal optimization method based on semi-intelligent prediction-optimization according to claim 1, characterized in that, The specific steps for forming the signal scheme set are as follows: Consider the arrival characteristics of the traffic flow at different control time periods of the intersection to determine an optional subset of phase sequences; Take values around the optimal cycle in a specific traffic flow scenario to determine an optional subset of signal cycles; Consider the actual situation to determine an optional subset of green ratios and delete the green ratio schemes that do not conform to the actual traffic situation; After determining all feasible subsets of phase sequences, signal cycles, and green ratios, combine the three to obtain the final signal scheme set.

3. The traffic signal optimization method based on semi-intelligent prediction-optimization according to claim 2, characterized in that The optimal cycle is calculated according to the HCM signal timing method.

4. The traffic signal optimization method based on semi-intelligent prediction-optimization according to claim 1, characterized in that, The signal control scheme satisfies the double-loop structure signal timing method.

5. The traffic signal optimization method based on semi-intelligent prediction-optimization according to claim 1, characterized in that The model loss function considers reducing the prediction error between every two signal control schemes while minimizing the mean square error to directly predict the total vehicle delay at the intersection under each signal optimization scheme.

6. The traffic signal optimization method based on semi-intelligent prediction-optimization according to claim 1, characterized in that, The model loss function is as follows: ， Among them, is the number of signal plans, represents the sum of squares of the differences between the predicted values and the true values of the total vehicle delay at the intersection, represents the sum of squares of the differences obtained by subtracting the difference between the predicted values and the true values of the total vehicle delay under every two plans from each other; and respectively represent the true value and the predicted value of the total vehicle delay at the intersection of the i-th plan.

7. The traffic signal optimization method based on semi-intelligent prediction-optimization according to claim 1, characterized in that For a single control time period in a day, the signal scheme selection optimization model should select the signal control scheme that minimizes the total vehicle delay at the intersection during this time period as the optimal signal control scheme.

8. A traffic signal optimization system based on semi-intelligent prediction-optimization, characterized in that, It includes: A data acquisition module: used to collect traffic flow data at intersections at different time periods; A scheme design module: used to design several signal control schemes based on the traffic flow data to form a signal scheme set, and each signal control scheme includes different phase sequences, signal cycles, and green ratios; A model construction module: used to construct a signal scheme optimization model and a model loss function; A model training module: used to train the signal scheme optimization model based on the traffic flow data and the model loss function to obtain a signal scheme selection optimization model; A scheme selection module: used to use the signal scheme selection optimization model to select the optimal signal control scheme from the signal scheme set to generate an optimal signal control scheme sequence for the intersection at multiple time periods.

9. A computer device, characterized in that, The computer device includes: One or more processors; A memory for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the semi-intelligent prediction-optimization-based traffic signal optimization method as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing computer instructions, characterized in that, When the computer instructions are executed by one or more processors, the one or more processors execute the steps in the method described in any one of claims 1 to 7.

Images