A method for signal control period division considering the frequent occurrence periods of extreme scenarios

Abstract

The present invention discloses a signal control period division method considering the frequent occurrence periods of extreme scenarios, which relates to the technical field of signal control period division, and includes: S1. Collecting historical traffic flow data of intersections; S2. Adopting the fisher optimal segmentation method; S3. Using the total flow of intersections to correct the basic signal control period, so as to solve the problem of too long division periods for morning and evening rush hours in the division results of the fisher optimal segmentation method; S4. Based on the historical traffic flow data, identifying the frequent occurrence periods of typical problems of extreme scenarios such as spillover at the exit lane, queue imbalance, and green light idling under three date states; S5. Fusing the frequent occurrence periods of typical problems of extreme scenarios and the corrected signal control periods. By analyzing the time-varying characteristics of historical traffic flow data and the frequent occurrence characteristics of typical problems, an innovative signal control period division method considering the frequent occurrence periods of extreme scenarios is proposed, and multiple signal control periods are divided for three date states of weekdays, non-working days, and holidays.

Description

Technical Field

[0001] The present invention relates to the technical field of signal control time period division, and specifically to a signal control time period division method considering the frequently occurring time periods of extreme scenarios. Background Art

[0002] The rapid growth of the motor vehicle ownership has made the urban traffic congestion, especially the traffic congestion at intersections, more and more serious. With the development of communication technology and signal control technology in recent years, the time - segmented signal control is expected to become an effective traffic control method to improve the intersection efficiency and relieve the traffic congestion problem. And the quality of the time period division result is particularly important for the effect of the time - segmented signal control.

[0003] At present, most of the signal control time period division methods use the traffic flow at the intersection approach level as the feature, and use the fisher optimal segmentation method or other machine learning methods for time period division. Some methods consider indicators such as the platoon wave speed and queue length at the approach. These methods are applied to unsaturated intersections with reasonable phase - sequence and green - time schemes, and can obtain relatively high - quality time period division results. However, for intersections prone to extreme scenarios such as spillover at the exit lane, unbalanced queuing, and approach empty - running, since the division features cannot identify these problems, the time period division results cannot reflect the frequently occurring time periods of typical problems in extreme scenarios. Therefore, the present solution proposes a signal control time period division method considering the frequently occurring time periods of extreme scenarios. Summary of the Invention

[0004] The present invention provides a signal control time period division method considering the frequently occurring time periods of extreme scenarios, which promotes the solution of the problems mentioned in the above - mentioned background art.

[0005] The present application provides a signal control time period division method considering the frequently occurring time periods of extreme scenarios, and adopts the following technical solutions: A signal control time period division method considering the frequently occurring time periods of extreme scenarios includes the following steps:

[0006] S1. Collect historical traffic flow data of the intersection;

[0007] Among them, the historical traffic flow data covers three date states: weekdays, non - weekdays, and holidays;

[0008] S2. Adopt the fisher optimal segmentation method to conduct basic time period division for the three date states based on the historical traffic flow data, and delimit multiple basic signal control time periods for each date state;

[0009] S3. Use the total intersection traffic flow to correct the basic signal control time periods to solve the problem of too long divided time periods for the morning and evening rush hours in the division results of the fisher optimal segmentation method;

[0010] S4. Based on historical traffic flow data, identify the frequently occurring periods of typical problems such as exit ramp spillover, queuing imbalance, and green light idle running in extreme scenarios under three date statuses;

[0011] S5. Integrate the frequently occurring periods of typical problems in extreme scenarios and the corrected signal control periods as the signal control periods considering the identification of frequently occurring periods in extreme scenarios.

[0012] By adopting the above technical solutions, the frequently occurring periods of deeper - level typical problems in the basic periods with similar traffic flow characteristics are identified, the division results of the basic periods based on traffic flow are corrected, and the similarity of traffic flow characteristics within each period is effectively improved.

[0013] Further, the fisher optimal segmentation method in step S2 is specifically as follows:

[0014] Among them, the phase - level traffic flow with a 5 - minute granularity is used as the period division feature;

[0015] Suppose there are N samples arranged in order {x1, x2, …, x N}, for the k - th group of samples {x i , x i+1 , …, x j}, the within - group sample difference is expressed by variance as:

[0016]

[0017] In the formula, d ij is the within - group variance of the k - th group of samples, is the feature vector of the k s - th sample, is the mean of the k - th group of samples;

[0018] Use b(N, K) to represent a method of dividing N ordered samples into K groups;

[0019] Define the loss function as:

[0020]

[0021] The classification method that minimizes the loss function is the optimal segmentation, denoted as P(N, K).

[0022] Further, for the optimal segmentation P(N, K), the diameter of the right - most category K is D(1, N), and at this time the loss function of the optimal segmentation is:

[0023]

[0024] Among them, L[P(j - 1, K - 1)] represents the optimal segmentation of the previous j - 1 samples divided into K - 1 groups. This loss function is recursively deduced forward to determine the optimal segmentation of each category;

[0025] Then, N samples are divided into one group. At this time, the loss function when K = 1 is: L[P(N, 1)] = L[b(N, 1)] = D(1, N);

[0026] After obtaining the loss functions for different numbers of samples N divided into different numbers of groups K, starting from the number of groups K = 1 and increasing. If the decrease in the loss function when increasing the number of groups is too low, the number of groups will no longer increase, and the grouping result corresponding to the current number of groups will be output as the final time period division result.

[0027] By adopting the above technical solution, using the fisher optimal segmentation method, based on historical traffic flow data, the basic time periods are divided for three date states, and multiple basic signal control time periods are defined for each date state.

[0028] Further, the specific steps for correcting the basic signal control time period using the total intersection flow in S3 are as follows:

[0029] a. Calculate the total intersection flow {V1, V2, …, V N} of each time unit using the traffic flow data set;

[0030] b. For a certain basic signal control time period i, extract the total intersection flow V i,s at the starting time unit of the time period, the maximum flow V i,max during the time period and its corresponding time unit n i,max ;

[0031] c. Starting from the time unit n i,max , traverse backward the flow V i,j of each time unit n i,j within the basic signal control time period i. If V i,j decreases to the flow V i,s at the starting time unit, then take the time unit n i,j as the early break time unit of the basic signal control time period i;

[0032] d. If no early break time unit n s,j is found until the end of the basic signal control time period i, then this time period does not have an early break;

[0033] e. Query the early break time units of each basic signal control time period and correct each basic signal control time period based on the early break time units;

[0034] f. If the corrected basic signal control time period is less than 15 minutes, merge it with the adjacent time period.

[0035] Further, the total traffic flow at the intersection of each time unit is the phase-level traffic flow with a granularity of 5 minutes.

[0036] By adopting the above technical solution, the correction is made based on the division result of the basic time period of the traffic flow, effectively improving the similarity of traffic flow characteristics in each time period.

[0037] Further, when identifying the frequent occurrence time periods of typical problems in the three date states, such as spillover at the exit lane, queue imbalance, and extreme scenarios of green light idling:

[0038] If the time occupancy rate of the exit lane is higher than 60% and the average vehicle speed is lower than 8 km / h, it is identified as spillover at the exit lane;

[0039] Let the maximum and minimum queue lengths of each approach be q max and q min , respectively. If q max > 100 m and q max - q min > 50 m, it is identified as queue imbalance;

[0040] If the green light utilization rate of the approach is lower than 10%, it is identified as low green light utilization rate of the approach;

[0041] Count the occurrence times x m,n of each typical problem m in each time unit n;

[0042] If ∑ n x m,n / 20 > 0.5, it is considered that the typical problem m frequently occurs in the time unit n, and record γ m,n = 1, otherwise γ m,n = 0;

[0043] Traverse the time unit n. If γ m,n = γ m,n+1 = … = γ m,n+j = 1, j ≥ 2 and γ m,n+j+1 = 0, then record the time unit sequence {n, n + 1, …, n + j} as a frequent occurrence time period of the typical problem m Count the frequent occurrence time periods of each typical problem;

[0044] If the frequent occurrence time periods of multiple typical problems overlap, then only retain the overlapping frequent occurrence time periods of the most important problem according to the problem importance level, and delete the overlapping frequent occurrence time periods of other problems. At this time, the problem importance level is: spillover at the exit lane > queue imbalance > green light idling;

[0045] Count the time unit sequence of the frequent occurrence time periods of each typical problem and its corresponding problem type.

[0046] Further, in S5, the fusion of the frequently occurring time periods of typical problems and the correction signal control time periods specifically includes the following steps:

[0047] Record the start time unit of each time period in the correction signal control time period as

[0048] Record the start and end time units of each time period in the frequently occurring time periods of typical problems as

[0049] If the start time unit of the i-th correction signal control time period is within the j-th frequently occurring time period of typical problems, that is then delete

[0050] Arrange all the start time units of the correction signal control time periods in sequence and the start and end time units of each time period in the frequently occurring time periods of typical problems and merge the time periods shorter than 15 minutes forward;

[0051] If the start time unit of the i-th correction signal control time period is flanked by frequently occurring time periods of typical problems, that is, there exists 0 ≤ j ≤ P - 1 such that and If the performance problems of the j-th and the (j + 1)-th frequently occurring time periods of typical problems are the same, then consider as an invalid division point and delete

[0052] According to the remaining start time units of the correction signal control time periods and the start and end time units of each time period in the frequently occurring time periods of typical problems Output the time period division result.

[0053] The present invention has the following beneficial effects:

[0054] 1. By analyzing the time-varying characteristics of historical traffic flow data and the frequently occurring characteristics of typical problems, the present invention innovatively proposes a signal control time period division method considering the frequently occurring time periods of extreme scenarios, and divides multiple signal control time periods for three date states: weekdays, non-weekdays, and holidays.

[0055] 2. The signal control time period division method considering the frequently occurring time periods of extreme scenarios combines traffic flow indicators such as green light utilization rate, time occupancy rate, queue length, and vehicle driving speed when dividing signal control time periods, identifies the frequently occurring time periods of typical problems at a deeper level within the basic time periods with similar flow characteristics, corrects the division result of the basic time periods based on flow, effectively improves the similarity of traffic flow characteristics within each time period, and is more conducive to designing the background scheme of signal control based on the time period division result. Brief Description of the Drawings

[0056] Figure 1 It is a schematic flow chart of the method of the present invention;

[0057] Figure 2 It is a channelization diagram of the intersection of G106 National Highway - Fanghua Road;

[0058] Figure 3 It is a schematic diagram of the division result of the basic signal control time period by the fisher optimal segmentation method. Different broken lines represent the total lane flow of different phases at the intersection, and the dotted line represents the time period demarcation point;

[0059] Figure 4 It is a schematic diagram of the division result of the signal control time period corrected by the total flow of the intersection. Different broken lines represent the total lane flow of different phases at the intersection, and the dotted line represents the time period demarcation point;

[0060] Figure 5 It is a schematic diagram of the division result of the signal control time period considering the frequently occurring time periods of extreme scenarios. Different broken lines represent the total lane flow of different phases at the intersection, and the dotted line represents the time period demarcation point. Detailed Implementation Manner

[0061] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0062] Embodiment 1

[0063] Referring to Figure 1 , a signal control time period division method considering the frequently occurring time periods of extreme scenarios includes the following steps:

[0064] S1. Collect historical traffic flow data of the intersection;

[0065] Among them, the historical traffic flow data covers three date states: weekdays, non - weekdays, and holidays;

[0066] S2. Adopt the fisher optimal segmentation method to conduct basic time period division on the three date states based on the historical traffic flow data, and demarcate multiple basic signal control time periods for each date state;

[0067] S3. Use the total flow of the intersection to correct the basic signal control time period to solve the problem that the divided time periods of the morning and evening rush hours are too long in the division result of the fisher optimal segmentation method;

[0068] S4. Identify the frequently occurring periods of typical problems of exit ramp spillover, queue imbalance, and green light idling in extreme scenarios under three date states based on historical traffic flow data;

[0069] S5. Integrate the frequently occurring periods of typical problems in extreme scenarios and the corrected signal control periods as the signal control periods considering the identification of frequently occurring periods in extreme scenarios.

[0070] The frequently occurring periods of deeper typical problems in the basic periods with similar traffic flow characteristics are identified, and the division results of the basic periods based on traffic flow are corrected, effectively improving the similarity of traffic flow characteristics within each period.

[0071] The specific fisher optimal segmentation method in S2 is as follows:

[0072] Among them, the phase-level traffic flow with a 5-minute granularity is used as the period division feature;

[0073] Suppose there are N samples arranged in order {x1, x2, …, x N}, for the kth group of samples {x i , x i+1 , …, x j}, the within-group sample difference is expressed by variance as:

[0074]

[0075] In the formula, d ij is the within-group variance of the kth group of samples, is the feature vector of the k s th sample, is the mean of the kth group of samples;

[0076] Let b(N, K) represent a method of dividing N ordered samples into K groups;

[0077] Define the loss function as:

[0078]

[0079] The classification method that minimizes the loss function is the optimal segmentation, denoted as P(N, K).

[0080] For the optimal segmentation P(N, K), the diameter of the rightmost category K is D(1, N), and the loss function of the optimal segmentation at this time is:

[0081]

[0082] Among them, L[P(j - 1, K - 1)] represents the optimal segmentation of the previous j - 1 samples divided into K - 1 groups, and this loss function is recursively deduced forward to determine the optimal segmentation of each category;

[0083] Then N samples are divided into a group. At this time, the loss function when K = 1 is: L[P(N,1)] = L[b(N,1)] = D(1,N);

[0084] After obtaining the loss functions for different numbers of samples N divided into different numbers of groups K, starting from the number of groups K = 1 and increasing it, if the decrease in the loss function when adding a group is too small, the number of groups will no longer increase, and the grouping result corresponding to the current number of groups will be output as the final time period division result.

[0085] Using the fisher optimal segmentation method, based on historical traffic flow data, basic time periods are divided for three date states, and multiple basic signal control time periods are defined for each date state.

[0086] The specific steps for correcting the basic signal control time period using the total intersection flow in S3 are as follows:

[0087] a. Calculate the total intersection flow {V1, V2, …, V N} of each time unit using the traffic flow data set;

[0088] b. For a certain basic signal control time period i, extract the total intersection flow V i,s at the starting time unit of the time period, the maximum flow V i,max during the time period and its corresponding time unit n i,max ;

[0089] c. Starting from the time unit n i,max , traverse backward the flow V i,j of each time unit n i,j within the basic signal control time period i. If V i,j decreases to the flow V i,s at the starting time unit, then the time unit n i,j is used as the early break time unit of the basic signal control time period i;

[0090] d. If no early break time unit n i,j is found until the end of the basic signal control time period i, then this time period does not have an early break;

[0091] e. Query the early break time units of each basic signal control time period and correct each basic signal control time period based on the early break time units;

[0092] f. If the corrected basic signal control time period is less than 15 minutes, merge it with the adjacent time period.

[0093] The total intersection flow of each time unit is the phase - level flow with a 5 - minute granularity.

[0094] Based on the correction of the basic time period division results according to traffic flow, the similarity of traffic flow characteristics within each time period is effectively improved.

[0095] When identifying the frequent occurrence time periods of typical problems such as exit ramp spillover, queue imbalance, and green light idling in extreme scenarios under three date states:

[0096] If the time occupancy rate of the exit ramp is higher than 60% and the average vehicle speed is lower than 8 km / h, it is identified as an exit ramp spillover;

[0097] Let the maximum and minimum queue lengths of each approach be q max and q min , if q max > 100 m and q max - q min > 50 m, it is identified as a queue imbalance;

[0098] If the green light utilization rate of the approach is lower than 10%, it is identified as a low green light utilization rate of the approach;

[0099] Count the occurrence times x m,n of each typical problem m of each extreme scenario in each time unit n;

[0100] If ∑ n x m,n / 20 > 0.5, it is considered that the typical problem m frequently occurs in the time unit n, and record γ m,n = 1, otherwise γ m,n = 0;

[0101] Traverse the time unit n. If γ m,n = γ m,n+1 = … = γ m,n+j = 1, j ≥ 2 and γ m,n+j+1 = 0, then record the time unit sequence {n, n + 1, …, n + j} as a frequent occurrence time period of the typical problem m Count the frequent occurrence time periods of each typical problem;

[0102] If the frequent occurrence time periods of multiple typical problems overlap, then according to the importance of the problems, only retain the overlapping frequent occurrence time periods of the most important problem and delete the overlapping frequent occurrence time periods of other problems. At this time, the importance of the problems is: exit ramp spillover > queue imbalance > green light idling;

[0103] Count the time unit sequence of the frequent occurrence time periods of each typical problem and its corresponding problem type.

[0104] In S5, fusing the frequent occurrence time periods of typical problems and the corrected signal control time periods specifically includes the following steps:

[0105] Record the start time unit of each time period in the corrected signal control time period as

[0106] Record the start and end time units of each time period during the frequently occurring periods of typical problems as

[0107] If the start time unit of the i-th correction signal control period is within the j-th frequently occurring period of typical problems, that is then delete

[0108] Arrange all the start time units of the correction signal control periods in sequence and the start and end time units of each time period in the frequently occurring periods of typical problems and merge the periods shorter than 15 minutes forward;

[0109] If the start time unit of the i-th correction signal control period is flanked by frequently occurring periods of typical problems, that is, there exists 0 ≤ j ≤ P - 1 such that and If the performance problems of the j-th and the (j + 1)-th frequently occurring periods of typical problems are the same, then it is considered that is an invalid division point and delete

[0110] According to the remaining start time units of the correction signal control periods and the start and end time units of each time period in the frequently occurring periods of typical problems Output the time period division result.

[0111] Embodiment 2

[0112] Refer to Figures 1 - 5 , a method for dividing signal control time periods considering frequently occurring periods of extreme scenarios, comprising the following steps:

[0113] S1. Collect historical traffic flow data at intersections, covering multiple days of data in three date states: weekdays, non-weekdays, and holidays;

[0114] In this example, taking the G106 National Highway - Fanghua Road in Baiyun District, Guangzhou as an example, historical traffic flow data with a time granularity of 5 minutes is collected through radar-vision perception equipment.

[0115] The channelization of the G106 National Highway - Fanghua Road intersection, and the signal phase numbers of each lane are as shown in the appendix Figure 2 as follows.

[0116] S2. Adopt the fisher optimal segmentation method to conduct basic time period division for the three date states based on historical traffic flow data, and delimit multiple basic signal control time periods for each date state;

[0117] In this example, the Fisher optimal segmentation method is used to delimit six application time periods for weekdays based on the historical traffic flow data collected in Step 1;

[0118] They are: 07:20 - 11:10, 11:10 - 12:20, 12:20 - 17:35, 17:35 - 20:35, 20:35 - 22:15, and 22:15 - 07:20, as Figure 3 shown.

[0119] Among them, 07:20 - 11:10 and 17:35 - 20:35 are the morning peak and evening peak respectively, but the end times of these two time periods are too late.

[0120] S3. Use the total intersection flow to correct the basic signal control time period to solve the problem of too long divided time periods for the morning and evening peaks in the division result of the Fisher optimal segmentation method;

[0121] For the time period of 07:20 - 11:10, its maximum flow appears at the 08:55 time unit. Traversing backward from the 08:55 time unit, 10:20 is the time unit when the flow decreases to the flow at 07:20, and it is used as the early cut-off time unit;

[0122] For the time period of 10:20 - 12:20, its maximum flow appears at the 10:20 time unit. Traversing backward from the 10:20 time unit, no early cut-off time unit is found;

[0123] For the time period of 12:20 - 17:35, its maximum flow appears at the 17:35 time unit. Traversing backward from the 17:35 time unit, no early cut-off time unit is found;

[0124] For the time period of 17:35 - 20:35, its maximum flow appears at the 18:50 time unit. Traversing backward from the 18:50 time unit, 19:20 is the time unit when the flow decreases to the flow at 17:35, and it is used as the early cut-off time unit;

[0125] For the time period of 19:20 - 22:15, its maximum flow appears at the 19:20 time unit. Traversing backward from the 19:20 time unit, no early cut-off time unit is found;

[0126] For the time period of 22:15 - 07:20, its maximum flow appears at the 07:20 time unit. Traversing backward from the 07:20 time unit, no early cut-off time unit is found;

[0127] The six corrected application time periods for weekdays are: 07:20 - 10:20, 10:20 - 12:20, 12:20 - 17:35, 17:35 - 19:20, 19:20 - 22:15, and 22:15 - 07:20, asFigure 4 as shown

[0128] Among them, 07:20 - 10:20 and 17:35 - 18:50 are the morning peak and the evening peak respectively. Compared with the morning and evening peaks obtained in step 2, the time lengths of the periods are reduced by 50 minutes and 75 minutes respectively.

[0129] S4. Based on historical traffic flow data, identify the frequently occurring periods of typical problems such as exit lane spillover, queuing imbalance, and green light idling under three date states;

[0130] No queuing imbalance and import lane idling problems were identified in each time unit. The occurrence times of exit lane spillover in each time unit are shown in the following table:

[0131]

[0132] According to the statistics in the above table, the occurrence frequencies of time units 08:00, 08:05, 17:40, and 18:45 are too low, so they are non - frequently occurring time units for exit lane spillover, and the rest are frequently occurring time units for exit lane spillover;

[0133] The periods of 08:20 - 08:25 and 17:50 are too short, and the periods of 08:55 - 09:10 and 18:10 - 18:40 are frequently occurring periods for exit lane spillover;

[0134] S5. Integrate the frequently occurring periods of typical problems in extreme scenarios and the corrected signal control periods as the signal control periods considering the identification of frequently occurring periods in extreme scenarios;

[0135] The starting time units of each period in the corrected signal control period are {07:20, 10:20, 12:20, 17:35, 19:20, 22:15}, and the starting and ending time units of each period in the frequently occurring periods of typical problems are {08:55, 09:10, 18:10, 18:40};

[0136] 07:20, 10:20, 12:20, 17:35, 19:20, 22:15 are not in the frequently occurring periods of typical problems and are not deleted;

[0137] Arrange the starting time units of each period in the corrected signal control period and the starting and ending time units of each period in the frequently occurring periods of typical problems in sequence to get: {07:20, 08:55, 09:10, 10:20, 12:20, 17:35, 18:10, 18:40, 19:20, 22:15}. There are no too - short periods and no merging is required;

[0138] The fusion results include a total of 10 time periods, namely 07:20 - 08:55, 08:55 - 09:10 (exit ramp overflow), 09:10 - 10:20, 10:20 - 12:20, 12:20 - 17:35, 17:35 - 18:10, 18:10 - 18:40 (exit ramp overflow), 18:40 - 19:20, 19:20 - 22:15, 22:15 - 07:20, as Figure 5 shown, where different broken lines represent the total lane flow of different phases at the intersection, and the dashed lines represent the time period demarcation points.

[0139] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device.

[0140] The above is only the preferred embodiment of the present invention. It should be pointed out that for those of ordinary skill in the art, without departing from the technical principle of the present invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present invention.

Claims

1. A method for signal control period division considering the frequent occurrence periods of extreme scenarios, characterized in that, It includes the following steps: S1. Collect historical traffic flow data at intersections; Among them, the historical traffic flow data covers three date states: weekdays, non - weekdays, and holidays; S2. Adopt the fisher optimal segmentation method to conduct basic time period division for the three date states based on the historical traffic flow data, and delimit multiple basic signal control time periods for each date state; S3. Use the total intersection flow to correct the basic signal control time periods to solve the problem of too long divided time periods for the morning and evening rush hours in the division results of the fisher optimal segmentation method; S4. Based on the historical traffic flow data, identify the frequently occurring time periods of typical problems such as exit - lane spillover, queue imbalance, and green - light idling in the three date states; S5. Integrate the frequently occurring time periods of typical extreme - scenario problems and the corrected signal control time periods as the signal control time periods considering the identification of frequently occurring time periods of extreme scenarios; The integration of the frequently occurring time periods of typical problems and the corrected signal control time periods specifically includes the following steps: Denote the start time unit of each time period in the correction signal control period as Denote the start and end time units of each time period during the frequently-occurring periods of typical problems as If the start time unit of the i-th correction signal control period is within the j-th typical problem frequent occurrence period, that is then delete Arrange all the start time units of the correction signal control periods in sequence and the start and end time units of each period in the periods with frequent occurrence of typical problems and merge the periods shorter than 15 minutes forward; If the start time unit of the i-th correction signal control period Both sides are periods with frequent occurrence of typical problems, that is, there exists 0 ≤ j ≤ P - 1 such that And If the performance problems in the j-th and the (j + 1)-th periods with frequent occurrence of typical problems are the same, it is considered that Is an invalid division point and is deleted Control the start time unit of the remaining correction signal control period and the start and end time units of each period in the typical problem frequent occurrence period Output the period division result.

2. The signal control period division method considering the frequently occurring periods of extreme scenarios according to claim 1, wherein The fisher optimal segmentation method in S2 is specifically as follows: Among them, the phase - level flow with a 5 - minute granularity is used as the time - period division feature; There are N samples arranged in order {x1, x2, …, x N}. For the k-th group of samples {x i , x i+1 , …, x j}, the within-group sample difference is expressed using variance as: where d ij is the within-group variance of the k-th group of samples, is the eigenvector of the k s -th sample, is the mean of the k-th group of samples; Use b(N,K) to represent a method of dividing N ordered samples into K groups; Define the loss function as: The classification method that minimizes the loss function is the optimal segmentation, denoted as P(N,K).

3. The signal control period division method considering the frequent occurrence periods of extreme scenarios according to claim 2, wherein For the optimal segmentation P(N,K), the diameter of the right - most category K is D(1,N), and at this time the loss function of the optimal segmentation is: Among them, L[P(j - 1,K - 1)] represents the optimal segmentation of the previous j - 1 samples divided into K - 1 groups, and this loss function is recursively deduced forward to determine the optimal segmentation of each category; Then the N samples are divided into one group, and at this time the loss function when K = 1 is: L[P(N,1)] = L[b(N,1)] = D(1,N); After obtaining the loss functions of different sample numbers N divided into different numbers of groups K, starting from the number of groups K = 1 and increasing, if the decrease amplitude of the loss function is too low when adding one more number of groups, the number of groups will no longer increase, and the grouping result corresponding to the current number of groups is output as the final time - period division result.

4. The signal control period division method considering the frequently occurring periods of extreme scenarios according to claim 3, wherein The use of the total intersection flow to correct the basic signal control time periods in S3 specifically includes the following steps: a. Calculate the total intersection flow rate {V1, V2, …, V N} for each time unit using the traffic flow dataset; b. For a certain basic signal control period i, extract the total intersection flow V at the starting time unit of the period i,s , the maximum flow V i,max during the period and its corresponding time unit n i,max ; c. Starting from time unit n i,max Traverse backward the traffic volume V of each time unit n within the basic signal control period i i,j from the start i,j If V i,j decreases to the traffic volume V of the starting time unit i,s , then take time unit n i,j as the early-break time unit of the basic signal control period i; d. If the early break time unit n has not been queried until the end of the basic signal control period i i,j , then there is no early break in this period; e. Query the early - break time units of each basic signal control time period and correct each basic signal control time period based on the early - break time units; f. If the corrected basic signal control time period is less than 15 minutes, merge it into the adjacent time period.

5. The signal control period division method considering the frequently occurring periods of extreme scenarios according to claim 4, characterized in that, The total intersection flow of each time unit is the phase - level flow with a 5 - minute granularity.

6. The method for dividing signal control time periods considering frequently occurring time periods of extreme scenarios according to claim 5, wherein In S4, when identifying the frequently occurring time periods of typical problems such as exit - lane spillover, queue imbalance, and green - light idling in the three date states: If the time occupancy rate of the exit lane is higher than 60% and the average vehicle speed is lower than 8 km / h, it is identified as an exit - lane spillover; Let the maximum and minimum queue lengths of each approach be denoted as q max and q min , respectively. If q max > 100 m and q max - q min > 50 m, it is identified as queue imbalance; If the green - light utilization rate of the entrance lane is lower than 10%, it is identified as a low green - light utilization rate of the entrance lane; Count the occurrence times x of the typical problem m in each extreme scenario in each time unit n m,n ; If ∑ n x m,n / 20 > 0.5, it is considered that the typical problem m occurs frequently within the time unit n, and record γ m,n = 1, otherwise γ m,n = 0; Traverse the time unit n. If γ m,n = γ m,n+1 = … = γ m,n+j = 1, j ≥ 2 and γ m,n+j+1 = 0, then record the time unit sequence {n, n + 1, …, n + j} as a frequent occurrence period of the typical problem m Count the frequent occurrence periods of each typical problem; If the frequently occurring time periods of multiple typical problems overlap, only the overlapping frequently occurring time periods of the most important problem are retained according to the importance of the problems, and the overlapping frequently occurring time periods of other problems are deleted. At this time, the importance of the problems is: exit - lane spillover > queue imbalance > green - light idling; Statistically analyze the time unit sequences of the frequently occurring periods of each typical problem and their corresponding problem types.

Images