Depth division method and system for intersection traffic period based on time and space state

Abstract

The present invention discloses a method and system for deep division of intersection traffic time periods based on spatiotemporal states. First, the time periods are divided once at a macro level by an ordered clustering algorithm based on Fisher optimal segmentation. Then, in order to compensate for the problem of dense segmentation points in time periods with low saturation due to data volatility, the time periods with a larger macroscopic time division range are deeply divided, and the critical points of the flow state in space are obtained by the k-means algorithm, and the segmentation points in the same state are merged. After the above three steps, the division of time periods and the flow state corresponding to each time period can be obtained, and corresponding timing plans are configured for different flow states. Through Vissim simulation and field test verification, the time period division method has greatly improved the queue length during peak hours and the empty situation during flat peak hours compared with fixed timing.

Description

Technical Field

[0001] The present invention belongs to the technical fields of traffic time period division, ordered clustering, traffic state recognition, intersection signal timing and traffic simulation, and in particular to a method and system for deeply dividing intersection traffic time periods based on spatiotemporal states. Background Art

[0002] In the process of traffic signal control, commonly used methods include timing control, adaptive control and real-time control. Adaptive control and real-time control have high requirements for the real-time nature of data. At present, there are not many intersections covered by radar and real-time detectors in cities, so these two schemes cannot be effectively implemented. Timing control is currently the most commonly used and easiest to implement signal control method. Commonly used timing control methods are divided into fixed timing and time-of-day timing control (TOD). Time-of-day timing control can more effectively consider the traffic status of different time periods and adjust the timing plan in time, so it is currently more widely used.

[0003] In the existing time period division scheme, the traffic control department relies more on the experience of signal control engineers. Experienced engineers can locate the key positions of time periods more accurately, but they are unable to achieve accurate division of multiple intersections. In the current research on automatic time period division, the document “Data–Driven Methodology for Signal Timing Plan Development: A Computational Approach” uses the K-means algorithm to divide time periods. In order to address the defect that the K-means algorithm needs to input the number of categories in advance, the document “Subtractive Clustering-Based K-means Technique for Determining Optimum Time-of-Day Breakpoints” introduces subtractive clustering and uses Synochro simulation to obtain the optimal number of clusters. However, clustering algorithms such as k-means will produce many outliers. Therefore, the document “Using K-means clustering to identify time-of-day break points for traffic signal timing plans” uses non-hierarchical clustering and sorts out the principles of manual adjustment to address the problem of outliers. In addition, there are artificial immune algorithms such as the document “Automatic traffic time period division method based on artificial immune algorithm”, NJW algorithm and improved NJW algorithm such as the document “Traffic control time period division based on improved NJW algorithm”. However, the above algorithms lack consideration of time characteristics, so they have certain deficiencies in time continuity and accuracy. To address this problem, ordered clustering (such as the literature "Research on intersection signal control model and algorithm based on dynamic time division", "Signal control time division method using recursive ordered clustering", etc.) has been applied to time division and has achieved certain results. However, due to the volatility of traffic data in different time periods, this time division method will cause the segmentation points to be concentrated in the time period with low saturation, resulting in the time division being not detailed enough at the micro level, and the results obtained are often not ideal. Moreover, the result of the division is only the segmentation state of the traffic in the time period, and it is impossible to identify the traffic state of the time period. Summary of the invention

[0004] The purpose of the present invention is to provide a method and system for deeply dividing the traffic time periods of an intersection based on the time and space state in order to solve the problems existing in the above-mentioned prior art.

[0005] The technical solution to achieve the purpose of the present invention is: a method for depth division of intersection traffic time periods based on time and space states, the method comprising the following steps:

[0006] Step 1: Arrange the traffic data of the intersection to be optimized in the order of the time when the data was collected, construct a relationship curve between the time series and the traffic, and then segment the time period;

[0007] Step 2, select the longest period in the first segmentation result, and then perform a second depth segmentation;

[0008] Step 3, perform one-dimensional segmentation on the flow data to obtain segmentation points of each flow state, and the interval between two segmentation points is a state interval, and the flow in it is in the same state;

[0009] Step 4, merging the time periods divided in steps 1 and 2 according to the flow status value in step 3;

[0010] Step 5: Use the webster algorithm and the improved HCM2000 delay model to time each time period, and use the same timing scheme for time periods with the same state value.

[0011] Furthermore, the time period is segmented once in step 1 and the second depth segmentation is performed in step 2, specifically using the Fisher optimal segmentation algorithm based on dynamic programming.

[0012] Furthermore, step 3 performs one-dimensional segmentation on the traffic data, specifically using a k-means algorithm to obtain k traffic segmentation points.

[0013] Furthermore, in step 4, the time periods divided in steps 1 and 2 are merged according to the flow state value in step 3, and the specific process includes:

[0014] Step 4-1, determine the flow state of each time period: obtain the state intervals included in the time period, then obtain the number of flow data points in each state interval, and take the state corresponding to the state interval with the largest number of flow data points as the flow state of the time period;

[0015] Step 4-2, for each time period division point after segmentation in steps 1 and 2, compare the flow states of the two adjacent time periods at the division point. If the two flow states are the same, remove the time period division point and merge the two adjacent time periods.

[0016] Furthermore, in step 5, the webster algorithm and the improved HCM2000 delay model are used to schedule each time period, and the specific process includes:

[0017] Step 5-1, calculate the optimal signal cycle duration:

[0018] Webster's formula:

[0019] If the k-means state value is less than or equal to the saturation division point

[0020] HCM2000 improved estimation formula:

[0021] C=1.23L·e (2.46-0.02L)Y , if the k-means state value is greater than the saturation division point

[0022] in,

[0023] L=nl+AR

[0024]

[0025] Where, C is the optimal signal cycle duration, L is the total loss duration of the intersection, l is the loss time of each phase; n is the number of phases, y is the critical flow ratio of the i-th phase, Y is the sum of the critical flow ratios of all phases, and AR is the full red duration;

[0026] Step 5-2, calculate the green light duration based on the optimal signal cycle duration:

[0027]

[0028] In the formula, g i is the green light duration of the i-th phase, g min is the shortest green light duration of the i-th phase, q i is the critical lane flow rate in the i-th phase, S i is the saturation flow of the i-th phase.

[0029] Furthermore, the method further comprises:

[0030] Step 6: Use Vissim simulation software to simulate the traffic and timing schemes in different time periods, and select average delay and average queue length as evaluation indicators to evaluate the simulation results.

[0031] A system for deeply dividing intersection traffic periods based on time and space states, the system comprising:

[0032] The first time period division module is used to arrange the traffic data of the intersection to be optimized according to the time sequence of data collection, construct a relationship curve between time series and traffic, and then divide the time period into segments;

[0033] The second time segmentation module is used to select the longest time period in the first segmentation result and then perform a second depth segmentation;

[0034] The flow state division module is used to perform one-dimensional segmentation on the flow data to obtain segmentation points of each flow state. The interval between two segmentation points is a state interval, and the flows in it are in the same state.

[0035] A time period merging module, used for merging the time periods segmented by the first time period dividing module and the second time period dividing module according to the flow state value;

[0036] The timing module is used to use the webster algorithm and the improved HCM2000 delay model to time each time period, and use the same timing scheme for time periods with the same state value.

[0037] Furthermore, the system further comprises:

[0038] The evaluation module is used to simulate the traffic and timing schemes in different time periods using the Vissim simulation software, and to evaluate the simulation results by selecting the average delay and average queue length as evaluation indicators.

[0039] Compared with the prior art, the present invention has the following significant advantages: 1) It solves the problem that most of the time period division points fall near the time periods with lower saturation, and the division effect on the time periods with higher saturation is poor; 2) By using the k-means algorithm to perform one-dimensional segmentation on the flow data, the flow state segmentation points can be obtained, and the flow state corresponding to each time period can be obtained; and the high and low limits of saturation are defined accordingly, and different delay models are used for time periods with different saturations to optimize the timing, which has better results; 3) According to the flow state points obtained by the k-means algorithm, the time periods are merged, which effectively solves the problem of too many time period divisions caused by secondary micro-ordered clustering.

[0040] The present invention is further described in detail below in conjunction with the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 The present invention is a flow chart of a method for deeply dividing intersection traffic time periods based on time and space states.

[0042] Figure 2 FIG. 4 is a diagram of an intersection channelization in one embodiment.

[0043] Figure 3 This is a result diagram obtained by performing macro-ordered clustering in an embodiment.

[0044] Figure 4 This is a result diagram obtained by performing micro-ordered clustering in an embodiment.

[0045] Figure 5 This is a diagram showing the result of k-means traffic state segmentation in one embodiment.

[0046] Figure 6 This is a result diagram of merging time periods according to k-means state values ​​in one embodiment.

[0047] Figure 7This is a diagram showing the result of time period division in one embodiment.

[0048] Figure 8 It is a Vissim simulation diagram of the intersection channelization diagram in one embodiment.

[0049] Fig. 9 The figure is a comparison result of average delay in Vissim simulation in one embodiment.

[0050] Fig.10 The figure is a comparison result of average queue lengths simulated in Vissim in one embodiment. DETAILED DESCRIPTION

[0051] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0052] In one embodiment, in combination Figure 1 , provides a method for deep division of intersection traffic periods based on time-space state, the method comprising the following steps:

[0053] Step 1: Arrange the traffic data of the intersection to be optimized in the order of the time when the data was collected, construct a relationship curve between the time series and the traffic, and then segment the time period;

[0054] Here, preferably, in one embodiment, the time period is segmented once using the Fisher optimal segmentation algorithm based on dynamic programming. For the Fisher optimal segmentation algorithm:

[0055] For a set of ordered samples, the observation values ​​are x1, x2, ..., xn. The method of clustering them without destroying the order between samples is called ordered clustering algorithm. To divide n samples into k segments, according to the total number of permutations and combinations, This segmentation method needs to traverse all possibilities. When the amount of data is large, for example, the data source collected is the traffic with a frequency of 5 minutes, 288 data need to be divided into 8 segments, and there are about 2.9 trillion possibilities in total, so it takes a long time.

[0056] Here we use dynamic programming combined with recursion to improve the calculation speed. First, define the following parameters:

[0057] 1.1 Variance of Classes

[0058] Suppose a class G contains samples {x(i), x(i+1), …, x(j)} (j>i), denoted by G = {i, i+1, …, j}. The mean vector of this class is for

[0059]

[0060] Use D(i,j) to represent this type of variance. Commonly used variances are:

[0061]

[0062] 1.2 The loss function of the classification method is:

[0063]

[0064] First, calculate the variance table between the data and record the data in a two-dimensional array, expressed as {D(0,1), D(0,2), D(0,3)…D(0,n), D(1,2), D(1,3}…D(1,n), D(2,3)…D(2,n),…, D(n-1,n)}. After obtaining the above variance table, the addition is the loss function of the two segments.

[0065] According to the idea of ​​Fisher optimal segmentation, to calculate the optimal k-segment segmentation, first divide the data into two parts, the first part is the k-th segment, and then find the optimal k-1 segment of the second part, so there are n-k+1 possibilities for k-segment optimization. Then calculate the optimal k-1 segment segmentation of the second part in the previous step, or divide the data into two parts, the first part is the k-1 segment, and find the optimal k-2 segment segmentation of the second part, and so on, until the 2-segment optimal segmentation of the last segment. The recursive formula is as follows:

[0066]

[0067]

[0068] Through the above steps, the division result of traffic data in time periods can be obtained.

[0069] For example, for Figure 2 The schematic diagram of an intersection is shown in Figure 1. The division result of a certain period of time is as follows: Figure 3 shown.

[0070] Step 2: Select the longest period in the first segmentation result, and then perform a second depth segmentation. The segmentation result is as follows: Figure 4 shown.

[0071] Step 3, perform one-dimensional segmentation on the flow data to obtain segmentation points of each flow state, and the interval between two segmentation points is a state interval, and the flow in it is in the same state;

[0072] Step 4, merging the time periods divided in steps 1 and 2 according to the flow status value in step 3;

[0073] Step 5, use the webster algorithm and the improved HCM2000 delay model to time each time period, and use the same time arrangement scheme for time periods with the same state value. In terms of timing, the webster algorithm is the most commonly used algorithm for calculating the cycle. This method calculates the optimal cycle length with the minimum vehicle delay time as the goal, but the algorithm is only applicable to situations with low saturation. Therefore, when the saturation is high, the present invention uses an improved estimation formula based on the US HCM2000 to calculate the cycle.

[0074] Further preferably, in one of the embodiments, step 3 performs one-dimensional segmentation on the traffic data, specifically using the k-means algorithm to obtain k traffic segmentation points. K-means is one of the ten classic data mining algorithms. The basic idea of ​​the k-means algorithm is to cluster with k random points in space as the center and classify the objects closest to them. Then update the value of each cluster center, reclassify, and iterate until the best clustering result is obtained.

[0075] The k-means clustering algorithm can be used to cluster traffic data in the spatial dimension to obtain k traffic segmentation points. The results of this part play an important role in the subsequent time period merging operation.

[0076] For example, the state values ​​and corresponding flow segmentation points obtained after state division are shown in Table 1 below. Figure 5 shown.

[0077] Table 1 Comparison table of state values ​​and flow segmentation points of k-means flow state segmentation

[0078] K-means state value Traffic segmentation point 1 17 2 52 3 81 4 108 5 148 6 172 7 209 8 242

[0079] Further, in one embodiment, the step 4 combines the time periods divided in step 1 and step 2 according to the flow state value in step 3, and the specific process includes:

[0080] Step 4-1, determine the flow state of each time period: obtain the state intervals included in the time period, then obtain the number of flow data points in each state interval, and take the state corresponding to the state interval with the largest number of flow data points as the flow state of the time period;

[0081] Step 4-2, for each time segmentation point after segmentation in steps 1 and 2, compare the flow status of the two adjacent time segments at the segmentation point. If the two flow statuses are the same, remove the time segmentation point and merge the two adjacent time segments. The result after merging is as follows: Figure 6 and 7After the above time period division, the corresponding table of time series and time periods shown in Table 2 is obtained, and the k-means state value corresponding to each time period is obtained. Different k-means state values ​​correspond to different signal control schemes.

[0082] Table 2 Time series, time periods and timing scheme numbers obtained after time period division

[0083] Time Series Time k-means state value 0-39 0:00-3:15 2 39-78 3:15-6:30 1 78-89 6:30-7:25 3 89-104 7:25-8:40 8 104-129 8:40-10:45 6 129-143 10:45-11:55 4 143-167 11:55-13:55 5 167-195 13:55-16:15 6 195-212 16:15-17:40 7 212-268 17:40-22:20 3 268-287 22:20-24:00 2

[0084] Further, in one embodiment, according to Figure 2 The intersection is channelized as shown, and the lanes are phased to design a signal control scheme. The design results are shown in Table 3 below.

[0085] Table 3 Phase number and phase release lane number

[0086] Phase number Lane number A 1，2，3，4，5，6，16 B 13，14，15，20 C 7，8，9，10，11，12，17，18，19，20 D 7，8，9，10，11，12,17，18，19

[0087] Step 5 uses the webster algorithm and the improved HCM2000 delay model to schedule each time period. The specific process includes:

[0088] Step 5-1, calculate the optimal signal cycle duration:

[0089] Webster's formula:

[0090] If the k-means state value is less than or equal to the saturation division point

[0091] HCM2000 improved estimation formula:

[0092] C=1.23L·e (2.46-0.02L)Y , if the k-means state value is greater than the saturation division point

[0093] in,

[0094] L=nl+AR

[0095]

[0096] Where, C is the optimal signal cycle duration, L is the total loss duration of the intersection, l is the loss time of each phase; n is the number of phases, y is the critical flow ratio of the i-th phase, Y is the sum of the critical flow ratios of all phases, and AR is the full red duration;

[0097] Step 5-2, calculate the green light duration based on the optimal signal cycle duration:

[0098]

[0099] In the formula, g iis the green light duration of the i-th phase, g min is the shortest green light duration of the i-th phase, q i is the critical lane flow rate in the i-th phase, S i is the saturation flow of the i-th phase.

[0100] For example, the timing scheme is shown in Table 4 below:

[0101] Table 4 The timing scheme of the present invention and the timing scheme obtained by an ordered clustering

[0102] Time Green light duration (ABCD) Time Green light duration (ABCD) 0:00-3:15 26-22-30-30(120) 0:00-0:50 24-23-30-30(119) 3:15-6:30 24-24-28-28(116) 0:50-1:55 23-23-29-29(116) 6:30-7:25 28-25-32-32(129) 1:55-3:00 24-21-29-29(115) 7:25-8:40 28-26-40-40(146) 3:00-6:30 23-23-29-29(116) 8:40-10:45 27-25-32-32(128) 6:30-7:15 24-24-30-30(120) 10:45-11:55 26-26-33-33(130) 7:15-17:40 28-26-41-41(148) 11:55-13:55 27-26-32-32(129) 17:40-18:50 27-25-32-32(128) 13:55-16:15 27-25-32-32(128) 18:50-22:00 28-24-32-32(128) 16:15-17:40 27-25-37-37(138) 22:00-24:00 26-24-30-30(122) 17:40-22:20 28-25-32-32(129) —— —— 22:20-24:00 26-22-30-30(120) —— ——

[0103] Furthermore, in one embodiment, the data used in the present invention does not have saturation-related content, so according to the data source and the result of k-means clustering, 80% of the highest state value of the k-means result is set as a high saturation situation, so the saturation division point in step 5-1 = the highest flow state value × 80%. For example, if the highest state value is 8, the saturation division point is 6.4.

[0104] Furthermore, in one embodiment, the method further comprises:

[0105] Step 6: Use Vissim simulation software to simulate the flow and timing schemes at different time periods. The simulation channelization diagram is as follows: Figure 8 As shown in Figure 2, the average delay and average queue length are selected as evaluation indicators to evaluate the simulation results. The evaluation results of the average delay are shown in Figure 2. Fig. 9 The evaluation results of the average queue length are shown in Fig.10 shown.

[0106] By comparing parameters such as average delay and average queue length, it can be found that the average queue length during the day (7:15-17:40) is optimized by 9%, the average delay during the day is optimized by 12.8%, the average queue length throughout the day is optimized by 14%, and the average delay throughout the day is optimized by 9.5%. Since the algorithm of the present invention can divide and identify the traffic status throughout the day and shorten the cycle length of each period during the day, reducing the phenomenon of empty parking, the average delay and average queue length are reduced. The queue length shows that the method proposed by the present invention is more accurate in identifying the traffic status and more detailed in the result of time period division.

[0107] In one embodiment, a system for depth division of intersection traffic periods based on spatiotemporal states is provided, the system comprising:

[0108] The first time period division module is used to arrange the traffic data of the intersection to be optimized according to the time sequence of data collection, construct a relationship curve between time series and traffic, and then divide the time period into segments;

[0109] The second time segmentation module is used to select the longest time period in the first segmentation result and then perform a second depth segmentation;

[0110] The flow state division module is used to perform one-dimensional segmentation on the flow data to obtain segmentation points of each flow state. The interval between two segmentation points is a state interval, and the flows in it are in the same state.

[0111] A time period merging module, used for merging the time periods segmented by the first time period dividing module and the second time period dividing module according to the flow state value;

[0112] The timing module is used to use the webster algorithm and the improved HCM2000 delay model to time each time period, and use the same timing scheme for time periods with the same state value.

[0113] Furthermore, in one embodiment, the system further comprises:

[0114] The evaluation module is used to simulate the traffic and timing schemes in different time periods using the Vissim simulation software, and to evaluate the simulation results by selecting the average delay and average queue length as evaluation indicators.

[0115] For the specific definition of the depth division system of the intersection traffic time period based on the time-space state, please refer to the definition of the depth division method of the intersection traffic time period based on the time-space state in the above text, which will not be repeated here. Each module in the above-mentioned depth division system of the intersection traffic time period based on the time-space state can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0116] The above shows and describes the basic principles, main features and advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments. The above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A method for deep division of intersection traffic periods based on spatiotemporal states, characterized in that: The method comprises the following steps: Step 1: Arrange the traffic data of the intersection to be optimized in the order of the time when the data was collected, construct a relationship curve between the time series and the traffic, and then segment the time period; Step 2, select the longest period in the first segmentation result, and then perform a second depth segmentation; Step 3, perform one-dimensional segmentation on the flow data to obtain segmentation points of each flow state, and the interval between two segmentation points is a state interval, and the flow in it is in the same state; Step 4, merging the time periods divided in steps 1 and 2 according to the flow status value in step 3; Step 5, use the webster algorithm and the improved HCM2000 delay model to time each time period, and use the same time schedule for time periods with the same state value; In step 1, the time period is segmented once, and in step 2, the second depth segmentation is performed, specifically using the Fisher optimal segmentation algorithm based on dynamic programming; Step 3 performs one-dimensional segmentation on the traffic data, specifically using the k-means algorithm to obtain k traffic segmentation points; Step 4 combines the time periods divided in steps 1 and 2 according to the flow status value in step 3. The specific process includes: Step 4-1, determine the flow state of each time period: obtain the state intervals included in the time period, then obtain the number of flow data points in each state interval, and take the state corresponding to the state interval with the largest number of flow data points as the flow state of the time period; Step 4-2, for each time period division point after segmentation in steps 1 and 2, compare the flow states of the two adjacent time periods at the division point. If the two flow states are the same, remove the time period division point and merge the two adjacent time periods.

2. The method for deep division of intersection traffic time periods based on time-space state according to claim 1 is characterized in that: Step 5 uses the webster algorithm and the improved HCM2000 delay model to schedule each time period. The specific process includes: Step 5-1, calculate the optimal signal cycle duration: Webster's formula: If the k-means state value is less than or equal to the saturation division point HCM2000 improved estimation formula: C=1.23L·e (2.46-0.02L)Y , if the k-means state value is greater than the saturation division point in, L=nl+AR Where, C is the optimal signal cycle duration, L is the total loss duration of the intersection, l is the loss time of each phase; n is the number of phases, y is the critical flow ratio of the i-th phase, Y is the sum of the critical flow ratios of all phases, and AR is the full red duration; Step 5-2, calculate the green light duration based on the optimal signal cycle duration: In the formula, g i is the green light duration of the i-th phase, g min is the shortest green light duration of the i-th phase, q i is the critical lane flow rate in the i-th phase, S i is the saturation flow of the i-th phase.

3. The method for deep division of intersection traffic time periods based on time-space state according to claim 2 is characterized in that: In step 5-1, the saturation dividing point = the maximum value of the flow state × 80%.

4. The method for deep division of intersection traffic time periods based on time-space state according to claim 3 is characterized in that: The method further comprises: Step 6: Use Vissim simulation software to simulate the traffic and timing schemes in different time periods, and select average delay and average queue length as evaluation indicators to evaluate the simulation results.

5. A system for deeply dividing intersection traffic time periods based on time-space states to implement the method described in any one of claims 1 to 4, characterized in that: The system comprises: The first time period division module is used to arrange the traffic data of the intersection to be optimized according to the time sequence of data collection, construct a relationship curve between time series and traffic, and then divide the time period into segments; The second time segmentation module is used to select the longest time period in the first segmentation result and then perform a second depth segmentation; The flow state division module is used to perform one-dimensional segmentation on the flow data to obtain segmentation points of each flow state. The interval between two segmentation points is a state interval, and the flows in it are in the same state. A time period merging module, used for merging the time periods segmented by the first time period dividing module and the second time period dividing module according to the flow state value; The timing module is used to use the webster algorithm and the improved HCM2000 delay model to time each time period, and use the same timing scheme for time periods with the same state value.

6. The in-depth division system of intersection traffic time periods based on time-space state according to claim 5 is characterized in that: The system further comprises: The evaluation module is used to simulate the traffic and timing schemes in different time periods using the Vissim simulation software, and to evaluate the simulation results by selecting the average delay and average queue length as evaluation indicators.

Images