Intelligent traffic signal management regulation and control system based on machine learning

Abstract

The invention relates to the technical field of traffic signal management, in particular to an intelligent traffic signal management regulation and control system based on machine learning, which comprises a memory, a processor and a computer program stored in the memory and running on the processor, when the processor executes the computer program, the following steps are realized: aiming at any phase of a target intersection at a to-be-predicted moment, acquiring at least one front intersection of the target intersection at any phase, and according to a possibility index of the traffic flow of each front intersection passing through the target intersection at the to-be-predicted moment, calculating the traffic flow of each front intersection; obtaining a traffic flow prediction value of any phase at a to-be-predicted moment, and obtaining a total flow prediction value by combining the traffic flow prediction value of any phase at the to-be-predicted moment; the total flow prediction value of each phase of the target intersection at the to-be-predicted moment is obtained, the green light duration of each phase of the target intersection at the to-be-predicted moment is dynamically regulated and controlled, the timeliness of traffic signal management is ensured, and the traffic jam problem is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of traffic signal management, and in particular, to an intelligent traffic signal management and control system based on machine learning. Background Art

[0002] Traffic problems have always been an important part of public issues. Especially with the increase in private vehicles, the fixed traffic signal mode can no longer maintain road smoothness and ensure traffic safety. Therefore, traffic signal management systems have become more and more intelligent, capable of dynamically adjusting the signal light mode according to the current road congestion situation.

[0003] Existing intelligent traffic signal management systems usually extend or shorten the green light duration according to the currently monitored traffic flow. For example, when there are continuously vehicles passing through an intersection, the green light duration will be extended by a fixed unit until a threshold is reached. Although existing intelligent traffic signal management systems can solve the traffic congestion problem caused by insufficient green light duration to a certain extent, the adjustment accuracy is relatively low. There may be a situation where the traffic flow decreases after the duration is increased, resulting in a decrease in the utilization rate of the green light. And the adjustment is also made after identifying that the traffic flow is large, which may also lead to vehicle congestion due to untimely dredging.

[0004] Therefore, how to improve the dynamic adjustment accuracy of the signal light mode has become an urgent problem to be solved. Summary of the Invention

[0005] In view of this, an embodiment of the present invention provides an intelligent traffic signal management and control system based on machine learning to solve the problem of how to improve the dynamic adjustment accuracy of the signal light mode.

[0006] An intelligent traffic signal management and control system based on machine learning provided in an embodiment of the present invention includes a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, the following steps are implemented: For any phase of a target intersection at a to-be-predicted moment, obtain at least one pre-intersection of the target intersection in the any phase, where, in the vehicle driving route, the pre-intersection is passed through first and then the target intersection; For any pre-intersection, obtain at least one adjacent intersection of the any pre-intersection, where the adjacent intersection refers to an intersection where the traffic flow of the any pre-intersection is diverted. According to the historical traffic flow monitoring values and historical weather monitoring data of each adjacent intersection, obtain the possibility index of the traffic flow of the any pre-intersection passing through the target intersection at the to-be-predicted moment; Obtain the predicted traffic flow value of any phase at the to-be-predicted moment according to the possibility index of the traffic flow of each upstream intersection passing through the target intersection at the to-be-predicted moment. Obtain the predicted pedestrian flow value of any phase at the to-be-predicted moment according to the historical pedestrian flow monitoring values of any phase of the target intersection within a preset duration. Combine the predicted traffic flow value and the predicted pedestrian flow value to obtain the total flow predicted value of any phase of the target intersection at the to-be-predicted moment; Obtain the total flow predicted value of each phase of the target intersection at the to-be-predicted moment, and dynamically adjust the green light duration of each phase of the target intersection at the to-be-predicted moment according to the total flow predicted value of each phase of the target intersection at the to-be-predicted moment.

[0007] The beneficial effects of the embodiments of the present invention compared with the prior art are as follows: The present invention combines the influence characteristics of time and weather on traffic flow, analyzes the possibility of the traffic flow of each upstream intersection of the target intersection being diverted to different intersections, that is, the possibility index of the traffic flow of any upstream intersection passing through the target intersection at the to-be-predicted moment, fully considers the regular characteristics of people's daily travel, and then combines the traffic flow data of each upstream intersection to predict in advance the traffic flow data of the target intersection at the to-be-predicted moment, that is, the predicted traffic flow value of each phase of the target intersection at the to-be-predicted moment, improving the accuracy of traffic flow prediction. Since the traffic signal at the intersection not only needs to ensure the normal driving of vehicles but also consider the normal passage of pedestrians, therefore, further combine the predicted pedestrian flow value of the target intersection at the to-be-predicted moment to predict the overall flow of the target intersection, which is used to represent the total flow predicted value of each phase of the target intersection at the to-be-predicted moment, so as to adjust in advance the green light duration of each phase of the target intersection according to the prediction situation of the target intersection to ensure the timeliness of traffic signal management, and at the same time, reduce the occurrence of traffic congestion problems. Description of the Drawings

[0008] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for use in the embodiments or the description of the prior art. Obviously, the following drawings are only some embodiments of the present invention, and those of ordinary skill in the art can also obtain other drawings without creative efforts based on these drawings.

[0009] Figure 1 It is a method flow chart of an intelligent traffic signal management and control method based on machine learning provided in Embodiment 1 of the present invention. Detailed Embodiments

[0010] The following will describe in detail the embodiments of the present disclosure, and the examples of the embodiments are shown in the drawings. The embodiments described below with reference to the drawings are exemplary and are intended to explain the present disclosure and should not be construed as limiting the present disclosure.

[0011] It should be noted that the terms "first", "second", etc. in the description of the present disclosure and the above-mentioned drawings are used to distinguish similar objects, and do not necessarily describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. On the contrary, they are merely examples of devices and methods consistent with some aspects of the present disclosure.

[0012] In order to illustrate the technical solution of the present invention, specific embodiments will be used for illustration below.

[0013] The specific scenario targeted by the present invention is as follows: In the process of the existing traffic signal management system dynamically adjusting the signal light duration according to the traffic flow situation, when traffic congestion is recognized and then the green light duration is adjusted, it is often impossible to dredge vehicles and people in time, resulting in a poor solution to the traffic congestion problem after adjustment, and the control of the green light duration is not precise enough, and still relies on the assistance of traffic police.

[0014] The embodiment of the present invention provides an intelligent traffic signal management and control system based on machine learning, including a processor and a memory. The processor executes the computer program stored in the memory to implement an intelligent traffic signal management and control method based on machine learning, as Figure 1 shown, the intelligent traffic signal management and control method based on machine learning includes the following steps: Step S101, for any phase of the target intersection at the moment to be predicted, obtain at least one pre-intersection of the target intersection in any phase, where the vehicle driving route passes through the pre-intersection first and then the target intersection.

[0015] The existing traffic signal management system ensures the smooth driving of vehicles at intersections and the timely passage of pedestrians by adjusting the duration of green lights and red lights. Therefore, the embodiments of the present invention need to monitor the traffic flow and pedestrian flow of each intersection in real time to facilitate the subsequent dynamic regulation of signal lights. Also, because different lanes at intersections represent different directions of vehicle turning, that is, the places where the vehicles will continue to drive after passing through are different. In order to predict the traffic flow at intersections based on the traffic flow and the possible driving routes of vehicles, it is necessary to monitor the traffic flow of each lane at the intersection separately. Therefore, the embodiments of the present invention install geomagnetic sensors on each lane at the intersection, and count the number of passing vehicles every 10s, which is recorded as the traffic flow monitoring value. At the same time, thermal imagers are installed at the entrance positions of each crosswalk at the intersection, and the number of passing pedestrians is also counted every 10s, which is recorded as the pedestrian flow monitoring value.

[0016] In the embodiment of the present invention, any intersection is denoted as the target intersection, which is used as the analysis object for dynamically regulating its traffic lights. Since each phase of an intersection has a fixed direction, not all other intersections can reach any phase of the target intersection. However, different intersections are interconnected. When a vehicle passes through the previous intersection, it is very likely to pass through other intersections connected to it. Therefore, based on the traffic flow monitoring value of the previous intersection connected to the target intersection, the traffic flow situation of the target intersection can be analyzed. Then, the previous intersections adjacent to each phase of the target intersection on the vehicle driving route are respectively obtained, that is, the vehicle passes through the previous intersection first and then the target intersection on the vehicle driving route, and thus the traffic flow situation of the target intersection can be predicted according to the traffic flow situation of the previous intersection later.

[0017] It should be noted that a signal phase is a term in a traffic system. During a signal cycle, if one or several vehicle flows obtain exactly the same traffic signal color display at any moment, then the consecutive time sequence in which they obtain different light colors (green light, yellow light, full red) is called a signal phase.

[0018] Since there are at least two phases at each intersection, in the embodiment of the present invention, any phase of the target intersection is taken as an example for analysis, and the analysis process for other phases is the same. At the same time, in order to avoid the situation of adjusting the green light duration after traffic congestion is identified, in the embodiment of the present invention, the current moment is used as the time point for prediction for the moment to be predicted. Here, the current moment refers to the previous sampling moment of the moment to be predicted, that is, the current moment is t and the moment to be predicted is t + 1. Therefore, for any phase of the target intersection at the moment to be predicted, first, at least one previous intersection of the target intersection in any phase is obtained, which is used for subsequent analysis of the possibility that the vehicle passing through the previous intersection will pass through the target intersection.

[0019] Step S102, for any previous intersection, obtain at least one adjacent intersection of the previous intersection. The adjacent intersection refers to the intersection where the traffic flow of any previous intersection is diverted. According to the historical traffic flow monitoring values and historical weather monitoring data of each adjacent intersection, obtain the possibility index of the traffic flow of any previous intersection passing through the target intersection at the moment to be predicted.

[0020] Since the traffic flow of the previous intersection may be diverted to different intersections, the different intersections where the diversion occurs are used as the adjacent intersections of the previous intersection. Among them, the adjacent intersections of the previous intersection also include the target intersection. Therefore, it is necessary to evaluate the traffic flow situation of the traffic flow diverted from the previous intersection to each adjacent intersection according to the historical traffic flow situation of the adjacent intersections of the previous intersection, so as to predict the possibility that the traffic flow of each previous intersection passes through the target intersection at the moment to be predicted.

[0021] Since the vehicle driving route is highly correlated with the intersection location, time, and weather conditions, it is possible to analyze the likelihood of the traffic flow at each upstream intersection passing through the target intersection at the prediction moment by combining the time and weather conditions corresponding to the historical traffic flow monitoring values. In the embodiment of the present invention, taking any upstream intersection of the target intersection as an example, first, at least one adjacent intersection of any upstream intersection is obtained. Then, for any adjacent intersection, according to the historical traffic flow monitoring values and historical weather monitoring data within one month before the prediction moment, the historical average traffic flow of any adjacent intersection is obtained, which is used to characterize the traffic diversion feature of any upstream intersection to any adjacent intersection before the prediction moment. The greater the historical average traffic flow, the greater the traffic flow of any adjacent intersection at the prediction moment.

[0022] Considering that time is a key factor affecting traffic flow and vehicle driving routes, that is, the traffic flow conditions at different times of the day will vary significantly. For example, the traffic flow during commuting hours will increase significantly, while the traffic flow during noon and evening rest times will decrease greatly. Then, the closer the time point between the historical traffic flow monitoring value of any adjacent intersection within one month before the prediction moment and the prediction moment, the closer the historical traffic flow monitoring value is to the traffic flow situation at the prediction moment. In addition, since the traffic flow change rules at different times of the week are also different. Generally speaking, people's travel activities from Monday to Friday are related to work and school hours, while weekend activities do not conform to the daily work and school hours. Therefore, the more consistent the week number corresponding to the historical traffic flow monitoring value is with the week number corresponding to the prediction moment, the more accurately the historical traffic flow monitoring value can reflect the traffic flow situation at the prediction moment. Therefore, in the embodiment of the present invention, according to the historical traffic flow monitoring values at each sampling moment within the preset historical period (within one month before the prediction moment) of any adjacent intersection, the first similarity degree in time between each historical traffic flow monitoring value and the traffic flow at the prediction moment is obtained respectively.

[0023] Among them, the method for obtaining the first similarity degree in time between each historical traffic flow monitoring value of any adjacent intersection and the traffic flow at the prediction moment is as follows: For any historical traffic flow monitoring value of any adjacent intersection within the preset historical period, record the sampling moment corresponding to the any historical traffic flow monitoring value as the target moment, calculate the absolute value of the time difference between the target moment and the prediction moment, obtain the first ratio between the absolute value of the time difference and the maximum time, and take the difference between the constant 1 and the first ratio as the time proximity degree; Obtain the week number to which the target moment belongs and the week number to which the moment to be predicted belongs, calculate the absolute value of the difference in week numbers between the week number to which the target moment belongs and the week number to which the moment to be predicted belongs, obtain the second ratio between the absolute value of the difference in week numbers and the maximum week number, and use the difference between the constant 1 and the second ratio as the week number consistency degree; Use the product of the moment proximity degree and the week number consistency degree as the first similarity degree in time between any historical traffic flow monitoring value and the traffic flow at the moment to be predicted.

[0024] In one embodiment, taking the i-th historical traffic flow monitoring value of the c-th adjacent intersection as an example, the calculation formula for the first similarity degree in time between the i-th historical traffic flow monitoring value and the traffic flow at the moment to be predicted is: Wherein, represents the first similarity degree in time between the i-th historical traffic flow monitoring value of the c-th adjacent intersection and the traffic flow at the moment to be predicted, 1 represents a constant, represents the sampling moment of the i-th historical traffic flow monitoring value of the c-th adjacent intersection (i.e., what time of day, such as 9:10 am), represents the moment to be predicted, represents the maximum sampling moment ( ), | | represents the absolute value symbol, represents the week number to which the i-th historical traffic flow monitoring value of the c-th adjacent intersection belongs (such as Monday, Tuesday, etc.), represents the week number to which the moment to be predicted belongs, represents the maximum week number ( ).

[0025] It should be noted that, is used to characterize the proximity degree of the sampling moment of the i-th historical traffic flow monitoring value and the moment to be predicted at a specific time point, that is, the moment proximity degree, The smaller the value of, the closer the sampling moment of the i-th historical traffic flow monitoring value is to the moment to be predicted at the time point, and the greater the first similarity degree in time between the i-th historical traffic flow monitoring value and the traffic flow at the moment to be predicted; is used to characterize whether the week number of the i-th historical traffic flow monitoring value is consistent with the week number of the moment to be predicted, The smaller the value of, the more consistent the week number of the i-th historical traffic flow monitoring value is with the week number of the moment to be predicted, and the greater the first similarity degree in time between the i-th historical traffic flow monitoring value and the traffic flow at the moment to be predicted.

[0026] Considering that weather is another important factor affecting traffic flow, and rainfall is the main influencing factor, because rainfall will greatly affect people's travel modes and travel decisions. For example, in heavy rain weather compared to sunny days, people are more inclined to drive, and their willingness to engage in non-essential outdoor activities will be greatly reduced. Therefore, in the embodiments of the present invention, according to the historical rainfall at each sampling moment of any adjacent intersection within a preset historical period (within one month before the moment to be predicted), the second similarity degree in terms of weather between each historical traffic flow monitoring value and the traffic flow at the moment to be predicted is obtained respectively.

[0027] Among them, the method for obtaining the second similarity degree in terms of weather between any historical traffic flow monitoring value and the traffic flow at the moment to be predicted is as follows: According to the historical rainfall at each sampling moment of the any adjacent intersection within the preset historical period, obtain the rainfall range between the maximum historical rainfall and the minimum historical rainfall; calculate the absolute value of the rainfall difference between the historical rainfall at the target moment and the rainfall at the moment to be predicted, calculate the third ratio between the absolute value of the rainfall difference and the rainfall range, and take the difference between the constant 1 and the third ratio as the second similarity degree in terms of weather between the any historical traffic flow monitoring value and the traffic flow at the moment to be predicted, where the rainfall at the moment to be predicted refers to the rainfall at the previous sampling moment of the moment to be predicted.

[0028] In one embodiment, considering that the time difference between the moment to be predicted and the current moment is small, and generally the rainfall does not change significantly in a short time, the rainfall at the moment to be predicted is set as the rainfall at the current moment. Taking the i-th historical traffic flow monitoring value of the c-th adjacent intersection as an example, the calculation formula for the second similarity degree in terms of weather between the i-th historical traffic flow monitoring value and the traffic flow at the moment to be predicted is as follows: Among them, represents the second similarity degree in terms of weather between the i-th historical traffic flow monitoring value of the c-th adjacent intersection and the traffic flow at the moment to be predicted, 1 represents a constant, represents the historical rainfall at the sampling moment (i.e., the target moment) of the i-th historical traffic flow monitoring value of the c-th adjacent intersection, represents the rainfall at the moment to be predicted, | | represents the absolute value symbol, represents the maximum historical rainfall within one month, represents the minimum historical rainfall within one month.

[0029] It should be noted that The smaller the value, the more similar the sampling moment of the $i$-th historical traffic flow monitoring value is to the weather condition at the moment to be predicted, that is, the closer the rainfall is, and the greater the second similarity degree between the $i$-th historical traffic flow monitoring value and the traffic flow at the moment to be predicted in terms of weather.

[0030] Similarly, the first similarity degree in terms of time and the second similarity degree in terms of weather between each historical traffic flow monitoring value of any adjacent intersection and the traffic flow at the moment to be predicted can be obtained. Considering that the influence degrees of time and weather on traffic flow may not be exactly the same, therefore, it is also necessary to further calculate the influence weights of time and weather on traffic flow respectively. If the influence degree of a certain factor on traffic flow is large, it indicates that the correlation between the change of traffic flow and the change of this factor is strong. Furthermore, when the traffic flow is at a certain level, the corresponding characteristic value of this factor should also be similar. In the embodiment of the present invention, according to the first similarity degree and the second similarity degree corresponding to each historical traffic flow monitoring value of any adjacent intersection, the time influence weight and the weather influence weight are respectively obtained.

[0031] Among them, the methods for obtaining the time influence weight and the weather influence weight are as follows: Using the Min - Max normalization method, normalize all historical traffic flow monitoring values to obtain the normalized traffic flow. Taking the normalized traffic flow as the horizontal axis, the first similarity degree as the vertical axis, and the second similarity degree as the vertical axis, a three - dimensional space is constructed. Map all historical traffic flow monitoring values in the three - dimensional space. One historical traffic flow monitoring value corresponds to a data point in the three - dimensional space. Perform k - means clustering on all data points in the three - dimensional space to obtain a preset number of clustering clusters; where $k = 6$, that is, the preset number is 6. There is no limitation here and it can be set according to the implementation scenario.

[0032] For any clustering cluster, obtain the standard deviation of the first similarity degree and the standard deviation of the second similarity degree of all data points in the any clustering cluster, which are respectively denoted as the time standard deviation and the weather standard deviation; accumulate the time standard deviations of all clustering clusters to obtain the time standard deviation accumulation value, normalize the reciprocal of the time standard deviation accumulation value to obtain the time influence weight; accumulate the weather standard deviations of all clustering clusters to obtain the weather standard deviation accumulation value, normalize the reciprocal of the weather standard deviation accumulation value to obtain the weather influence weight.

[0033] In an embodiment, the calculation formula of the time influence weight is: Among them, represents the time influence weight of the $c$-th adjacent intersection, represents the normalization function, represents the number of clustering clusters, represents the standard deviation of the first similarity degree of all data points in the o-th cluster corresponding to the c-th adjacent intersection, that is, the time standard deviation, and 1 represents a constant.

[0034] It should be noted that if the influence degree of time on traffic flow is large, then the traffic flow data corresponding to the data points with similar first similarity degrees will also be relatively similar. Therefore, the data points will be more concentrated and easier to be clustered into the same category. If the influence degree of time on traffic flow is small, then the distribution of data points with similar time corresponding to the same traffic flow level will be relatively scattered, that is, they are more likely to be clustered into different categories. Therefore, the smaller the time standard deviation of each cluster, the closer the time corresponding to the data points in the cluster, the greater the influence degree of time on traffic flow, and the greater the corresponding time influence weight.

[0035] Similarly, the calculation formula for the weather influence weight is: where, represents the weather influence weight of the c-th adjacent intersection, represents the normalization function, represents the number of clusters, represents the standard deviation of the second similarity degree of all data points in the o-th cluster corresponding to the c-th adjacent intersection, that is, the weather standard deviation, and 1 represents a constant.

[0036] After determining the first similarity degree and the second similarity degree corresponding to each historical traffic flow monitoring value of any adjacent intersection, as well as the time influence weight and the weather influence weight of any adjacent intersection, analyze the conditional similarity between each historical traffic flow monitoring value and the traffic flow at the moment to be predicted. Specifically: for any historical traffic flow monitoring value, according to the time influence weight and the weather influence weight, perform a weighted sum of the first similarity degree and the second similarity degree of the any historical traffic flow monitoring value to obtain the conditional similarity between the any historical traffic flow monitoring value and the traffic flow at the moment to be predicted.

[0037] In an embodiment, the calculation formula for the conditional similarity between the i-th historical traffic flow monitoring value of the c-th adjacent intersection and the traffic flow at the moment to be predicted is: where, represents the conditional similarity between the i-th historical traffic flow monitoring value of the c-th adjacent intersection and the traffic flow at the moment to be predicted, represents the first similarity degree in time between the i-th historical traffic flow monitoring value of the c-th adjacent intersection and the traffic flow at the moment to be predicted, represents the second similarity in weather between the i-th historical traffic flow monitoring value of the c-th adjacent intersection and the traffic flow at the moment to be predicted. represents the time influence weight of the c-th adjacent intersection. represents the weather influence weight of the c-th adjacent intersection.

[0038] Similarly, the conditional similarity between each historical traffic flow monitoring value of any adjacent intersection and the traffic flow at the moment to be predicted can be obtained. Then, based on the conditional similarity between each historical traffic flow monitoring value and the traffic flow at the moment to be predicted, the historical average traffic flow of any adjacent intersection can be obtained. The specific method is: taking the average of the products of each historical traffic flow monitoring value and its corresponding conditional similarity to obtain the historical average traffic flow of the said any adjacent intersection.

[0039] In one embodiment, the calculation formula for the historical average traffic flow of the c-th adjacent intersection is: where represents the historical average traffic flow of the c-th adjacent intersection, represents the i-th historical traffic flow monitoring value of the c-th adjacent intersection, represents the conditional similarity between the i-th historical traffic flow monitoring value of the c-th adjacent intersection and the traffic flow at the moment to be predicted, represents the number of historical traffic flow monitoring values of the c-th adjacent intersection.

[0040] According to the above method for obtaining the historical average traffic flow of the c-th adjacent intersection, the historical average traffic flow of each adjacent intersection of any pre-intersection can be obtained. The more the historical average traffic flow of each intersection is, the more likely the vehicle is to pass through this intersection after passing through the pre-intersection. Thus, calculate the possibility that the vehicle passing through the pre-intersection reaches the target intersection. The specific calculation method is: accumulating the historical average traffic flows of all adjacent intersections to obtain the accumulated historical traffic flow value; obtaining the historical average traffic flow of the target intersection, and calculating the ratio of the historical average traffic flow of the target intersection to the accumulated historical traffic flow value as the possibility index that the traffic flow of any pre-intersection passes through the target intersection at the moment to be predicted.

[0041] In one embodiment, the calculation formula for the possibility index that the traffic flow of the a-th pre-intersection passes through the target intersection at the moment to be predicted is: where represents the possibility index that the traffic flow of the a-th pre-intersection passes through the target intersection at the moment to be predicted, represents the number of adjacent intersections of the a-th pre-intersection, It represents the historical average traffic flow of the c-th adjacent intersection. It represents the historical average traffic flow of the target intersection.

[0042] Similarly, the possibility index of the traffic flow of each upstream intersection on any phase of the target intersection passing through the target intersection at the moment to be predicted can be obtained.

[0043] Step S103: According to the possibility index of the traffic flow of each upstream intersection passing through the target intersection at the moment to be predicted, obtain the traffic flow prediction value of any phase at the moment to be predicted. According to the historical pedestrian flow monitoring value of any phase of the target intersection within a preset time period, obtain the pedestrian flow prediction value of any phase at the moment to be predicted. Combine the traffic flow prediction value and the pedestrian flow prediction value to obtain the total flow prediction value of any phase of the target intersection at the moment to be predicted.

[0044] The greater the possibility of the traffic flow of the upstream intersection reaching the target intersection, the more likely it is that more vehicles will reach the target intersection at the moment to be predicted. Therefore, combine the real-time traffic flow data of all upstream intersections on any phase of the target intersection at the current moment and the possibility index of reaching the target intersection to obtain the traffic flow prediction value of any phase of the target intersection at the moment to be predicted. Taking the u-th phase of the target intersection as an example, obtain the real-time traffic flow monitoring value of each upstream intersection of the u-th phase at the current moment, multiply the real-time traffic flow monitoring value corresponding to each upstream intersection by the possibility index to obtain the multiplication value corresponding to each upstream intersection, and accumulate all the multiplication values to obtain the traffic flow prediction value of the u-th phase at the moment to be predicted.

[0045] Among them, the calculation formula for the traffic flow prediction value of the u-th phase at the moment to be predicted is: Among them, represents the traffic flow prediction value of the u-th phase of the target intersection at the moment to be predicted, represents the number of upstream intersections of the u-th phase of the target intersection, represents the possibility index of the traffic flow of the a-th upstream intersection of the u-th phase passing through the target intersection at the moment to be predicted, represents the real-time traffic flow monitoring value of the a-th upstream intersection of the u-th phase at the current moment.

[0046] When there are more pedestrians at the intersection, it is also necessary to appropriately extend the duration of the street lights to ensure that pedestrians can pass normally within the green light time and reduce the gathering and congestion of pedestrians at the intersection. Therefore, on the basis of obtaining the predicted value of the traffic flow at the target intersection, it is also necessary to combine the pedestrian flow situation at the intersection to predict the total flow of each phase at the target intersection. Since the number of pedestrians at the intersection will not change significantly in a short period of time, the historical pedestrian flow monitoring values at each sampling moment within the preset duration before the prediction moment for the u-th phase of the target intersection are obtained, and the average value of all historical pedestrian flow monitoring values is calculated, which is recorded as the predicted value of the pedestrian flow at the u-th phase at the prediction moment. Furthermore, the sum value between the predicted value of the traffic flow and the predicted value of the pedestrian flow at the u-th phase of the target intersection at the prediction moment is recorded as the predicted value of the total flow at the u-th phase of the target intersection at the prediction moment. .

[0047] Similarly, the predicted value of the total flow at the u-th phase of the target intersection at the prediction moment is obtained, and then the predicted value of the total flow of each phase at the target intersection at the prediction moment can be obtained.

[0048] Step S104, obtain the predicted value of the total flow of each phase at the target intersection at the prediction moment, and dynamically adjust the green light duration of each phase at the target intersection at the prediction moment according to the predicted value of the total flow of each phase at the target intersection at the prediction moment.

[0049] Because when ensuring that the signal changes between different intersections are continuous, the signal cycle duration of each intersection is usually fixed, and the phase with more traffic flow requires a longer green light duration to reduce road congestion. Therefore, it is necessary to dynamically adjust the green light duration of each phase at the target intersection at the prediction moment according to the predicted value of the total flow of each phase at the target intersection at the prediction moment. It should be noted that the signal cycle duration refers to the time required for the traffic signal to complete a full cycle of light colors (such as red - green - yellow - red), and the signal cycle duration should be set to 30 seconds to 150 seconds in the non-saturated traffic state and should not exceed 180 seconds in the saturated state.

[0050] Among them, dynamically adjusting the green light duration of each phase at the target intersection at the prediction moment according to the predicted value of the total flow of each phase at the target intersection at the prediction moment includes: Accumulate the predicted total flow values of all phases of the target intersection at the moment to be predicted to obtain the accumulated total flow value. For any phase of the target intersection, calculate the ratio between the predicted total flow value of the any phase at the moment to be predicted and the accumulated total flow value as the duration adjustment coefficient. Obtain the signal light cycle duration of the target intersection, and take the product of the duration adjustment coefficient and the signal light cycle duration as the target green light duration required for the any phase at the moment to be predicted, and adjust the green light duration of the any phase of the target intersection at the moment to be predicted to the target green light duration.

[0051] In one embodiment, taking the u-th phase of the target intersection as an example, the calculation formula for the target green light duration required for the u-th phase at the moment to be predicted is: Wherein, represents the target green light duration required for the u-th phase of the target intersection at the moment to be predicted, represents the predicted total flow value of the u-th phase at the moment to be predicted, represents the predicted total flow value of the v-th phase at the moment to be predicted, represents the number of phases of the target intersection, represents the signal light cycle duration of the target intersection.

[0052] According to the above calculation formula of the target green light duration, the target green light duration required for each phase of the target intersection at the moment to be predicted can be obtained. Furthermore, according to the target green light duration required for each phase at the moment to be predicted, the green light duration of each phase of the target intersection at the moment to be predicted is adjusted in advance to the target green light duration, which improves the timeliness and accuracy of traffic signal management.

[0053] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. An intelligent traffic signal management and control system based on machine learning, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, the following steps are implemented: For any phase of the target intersection at the moment to be predicted, obtain at least one preceding intersection of the target intersection in the any phase, where the vehicle driving route passes through the preceding intersection first and then the target intersection; For any preceding intersection, obtain at least one adjacent intersection of the any preceding intersection, where the adjacent intersection refers to the intersection where the traffic flow of the any preceding intersection is diverted. According to the historical traffic flow monitoring values and historical weather monitoring data of each adjacent intersection, obtain the possibility index of the traffic flow of the any preceding intersection passing through the target intersection at the moment to be predicted; According to the possibility index of the traffic flow of each preceding intersection passing through the target intersection at the moment to be predicted, obtain the traffic flow prediction value of the any phase at the moment to be predicted. According to the historical pedestrian flow monitoring values of the any phase of the target intersection within a preset time period, obtain the pedestrian flow prediction value of the any phase at the moment to be predicted. Combine the traffic flow prediction value and the pedestrian flow prediction value to obtain the total flow prediction value of the any phase of the target intersection at the moment to be predicted; Obtain the total flow prediction value of each phase of the target intersection at the moment to be predicted. According to the total flow prediction value of each phase of the target intersection at the moment to be predicted, dynamically adjust the green light duration of each phase of the target intersection at the moment to be predicted.

2. The intelligent traffic signal management and control system based on machine learning according to claim 1, characterized in that, The obtaining the possibility index of the traffic flow of the any preceding intersection passing through the target intersection at the moment to be predicted according to the historical traffic flow monitoring values and historical weather monitoring data of each adjacent intersection includes: For any adjacent intersection, respectively obtain the first similarity degree in time and the second similarity degree in weather between each historical traffic flow monitoring value and the traffic flow at the moment to be predicted according to the historical traffic flow monitoring values and historical rainfall at each sampling moment within a preset historical period; According to the first similarity degree and the second similarity degree corresponding to each historical traffic flow monitoring value, respectively obtain the time influence weight and the weather influence weight; according to the first similarity degree, the second similarity degree, the time influence weight and the weather influence weight corresponding to each historical traffic flow monitoring value, obtain the historical average traffic flow of the any adjacent intersection; Obtain the historical average traffic flow of each adjacent intersection, accumulate the historical average traffic flows of all adjacent intersections to obtain the historical traffic flow accumulation value; obtain the historical average traffic flow of the target intersection, and calculate the ratio of the historical average traffic flow of the target intersection to the historical traffic flow accumulation value as the possibility index of the traffic flow of the any preceding intersection passing through the target intersection at the moment to be predicted.

3. The intelligent traffic signal management and control system based on machine learning according to claim 2, characterized in that, The respectively obtaining the first similarity degree in time and the second similarity degree in weather between each historical traffic flow monitoring value and the traffic flow at the moment to be predicted according to the historical traffic flow monitoring values and historical rainfall at each sampling moment within a preset historical period includes: For any historical traffic flow monitoring value at any adjacent intersection within a preset historical period, record the sampling moment corresponding to the any historical traffic flow monitoring value as the target moment, calculate the absolute value of the time difference between the target moment and the moment to be predicted, obtain the first ratio between the absolute value of the time difference and the maximum moment, and take the difference between the constant 1 and the first ratio as the time proximity degree; Obtain the week number to which the target moment belongs and the week number to which the moment to be predicted belongs, calculate the absolute value of the week number difference between the week number to which the target moment belongs and the week number to which the moment to be predicted belongs, obtain the second ratio between the absolute value of the week number difference and the maximum week number, and take the difference between the constant 1 and the second ratio as the week number consistency degree; Take the product of the time proximity degree and the week number consistency degree as the first similarity degree in time between the any historical traffic flow monitoring value and the traffic flow at the moment to be predicted.

4. The intelligent traffic signal management and control system based on machine learning according to claim 3, characterized in that, The obtaining of the first similarity degree in time and the second similarity degree in weather between each historical traffic flow monitoring value and the traffic flow at the moment to be predicted according to the historical traffic flow monitoring value and the historical rainfall at each sampling moment within the preset historical period at any adjacent intersection further includes: According to the historical rainfall at each sampling moment within the preset historical period at any adjacent intersection, obtain the rainfall range between the maximum historical rainfall and the minimum historical rainfall; calculate the absolute value of the rainfall difference between the historical rainfall at the target moment and the rainfall at the moment to be predicted, calculate the third ratio between the absolute value of the rainfall difference and the rainfall range, and take the difference between the constant 1 and the third ratio as the second similarity degree in weather between the any historical traffic flow monitoring value and the traffic flow at the moment to be predicted, where the rainfall at the moment to be predicted refers to the rainfall at the previous sampling moment of the moment to be predicted.

5. The intelligent traffic signal management and control system based on machine learning according to claim 2, characterized in that, The obtaining of the time influence weight and the weather influence weight according to the first similarity degree and the second similarity degree corresponding to each historical traffic flow monitoring value respectively includes: Perform normalization processing on all historical traffic flow monitoring values to obtain normalized traffic flows. Take the normalized traffic flows as the horizontal axis, the first similarity degree as the vertical axis, and the second similarity degree as the vertical axis to construct a three-dimensional space. Map all historical traffic flow monitoring values in the three-dimensional space. One historical traffic flow monitoring value corresponds to one data point in the three-dimensional space. Perform k-means clustering on all data points in the three-dimensional space to obtain a preset number of clustering clusters; For any clustering cluster, obtain the standard deviation of the first similarity degree and the standard deviation of the second similarity degree of all data points in the any clustering cluster, and denote them as the time standard deviation and the weather standard deviation respectively; accumulate the time standard deviations of all clustering clusters to obtain the accumulated time standard deviation value, perform normalization on the reciprocal of the accumulated time standard deviation value to obtain the time influence weight; accumulate the weather standard deviations of all clustering clusters to obtain the accumulated weather standard deviation value, perform normalization on the reciprocal of the accumulated weather standard deviation value to obtain the weather influence weight.

6. The intelligent traffic signal management and control system based on machine learning according to claim 2, characterized in that, Obtaining the historical average traffic flow of any adjacent intersection according to the first similarity degree, second similarity degree, time influence weight, and weather influence weight corresponding to each historical traffic flow monitoring value includes: For any historical traffic flow monitoring value, perform a weighted sum of the first similarity degree and the second similarity degree of the any historical traffic flow monitoring value according to the time influence weight and the weather influence weight to obtain the conditional similarity between the any historical traffic flow monitoring value and the traffic flow at the to-be-predicted moment; Obtain the conditional similarity between each historical traffic flow monitoring value and the traffic flow at the to-be-predicted moment, and perform an averaging operation on the product of each historical traffic flow monitoring value and its corresponding conditional similarity to obtain the historical average traffic flow of any adjacent intersection.

7. The intelligent traffic signal management and control system based on machine learning according to claim 1, characterized in that, Obtaining the traffic flow prediction value of any phase at the to-be-predicted moment according to the possibility index of the traffic flow of each upstream intersection passing through the target intersection at the to-be-predicted moment includes: Denote the previous sampling moment of the to-be-predicted moment as the current moment, obtain the real-time traffic flow monitoring value of each upstream intersection at the current moment, multiply the real-time traffic flow monitoring value corresponding to each upstream intersection by the possibility index to obtain the multiplication value corresponding to each upstream intersection, and accumulate all the multiplication values to obtain the traffic flow prediction value of any phase at the to-be-predicted moment.

8. The intelligent traffic signal management and control system based on machine learning according to claim 1, characterized in that, Obtaining the pedestrian flow prediction value of any phase at the to-be-predicted moment according to the historical pedestrian flow monitoring values of any phase of the target intersection within a preset duration includes: Obtain the historical pedestrian flow monitoring values of each sampling moment within the preset duration before the to-be-predicted moment of any phase of the target intersection, calculate the average value of all the historical pedestrian flow monitoring values, and denote it as the pedestrian flow prediction value of any phase at the to-be-predicted moment.

9. The intelligent traffic signal management and control system based on machine learning according to claim 1, characterized in that, Combining the traffic flow prediction value and the pedestrian flow prediction value to obtain the total flow prediction value of any phase of the target intersection at the to-be-predicted moment includes: Denote the added value between the traffic flow prediction value and the pedestrian flow prediction value as the total flow prediction value of any phase of the target intersection at the to-be-predicted moment.

10. The intelligent traffic signal management and control system based on machine learning according to claim 1, characterized in that, Dynamically adjusting the green light duration of each phase of the target intersection at the to-be-predicted moment according to the total flow prediction value of each phase of the target intersection at the to-be-predicted moment includes: Accumulate the total flow prediction values of all phases of the target intersection at the to-be-predicted moment to obtain the total flow accumulation value. For any phase of the target intersection, calculate the ratio between the total flow prediction value of the any phase at the to-be-predicted moment and the total flow accumulation value as the duration adjustment coefficient, obtain the signal light cycle duration of the target intersection, take the product of the duration adjustment coefficient and the signal light cycle duration as the target green light duration required for the any phase at the to-be-predicted moment, and adjust the green light duration of the any phase of the target intersection at the to-be-predicted moment to the target green light duration.

Images