Traffic linkage control method, system and equipment based on road condition environment perception

Abstract

This application relates to the field of traffic control systems, and specifically to a method, system, and device for traffic linkage control based on road condition perception. The method comprises: collecting multi-source road condition data at intersections; analyzing various types of road condition data at different intersections using network analysis and entropy weight methods to adjust the input weight matrix of an LSTM model; and training the adjusted LSTM model and DQN model using a preset data set, and adjusting the phase of the corresponding traffic lights based on the model output. This application aims to consider the impact of multi-source road condition data other than the number of vehicles at a traffic intersection on road conditions, improve the accuracy of the LSTM model's road condition prediction, and thus enhance the effectiveness of subsequent traffic linkage control.

Description

Technical Field

[0001] The present application relates to the field of traffic control systems, and specifically to a traffic linkage control method, system, and equipment based on road condition and environment perception. Background Art

[0002] Intersections are crucial nodes in urban transportation systems, and their operational performance directly impacts the smoothness of traffic throughout the entire road network. Traffic coordination control is an efficient and sustainable method for traffic management and control, playing a vital role in improving road efficiency, alleviating congestion, reducing energy consumption and emissions, and ensuring safety. Traffic signal phase control is a crucial component of traffic coordination control. By adjusting the timing of different traffic signal phases in real time, it manages traffic flow and alleviates congestion, achieving efficient traffic coordination control.

[0003] Chinese patent CN114613168A discloses a traffic signal control method based on deep reinforcement learning using a memory network. This patent uses an LSTM to predict road conditions based on the number of vehicles entering and exiting an intersection, and further implements traffic light control based on a DQN algorithm. However, this patent fails to consider the impact of other multi-source road environment data on road conditions, resulting in a decrease in the accuracy of the LSTM model's road condition predictions, which in turn affects the effectiveness of subsequent traffic control. Summary of the Invention

[0004] In order to solve the above technical problems, the purpose of this application is to provide a traffic linkage control method, system and device based on road environment perception. The technical solutions adopted are as follows:

[0005] In a first aspect, an embodiment of the present application provides a traffic linkage control method based on road condition perception, the method comprising the following steps:

[0006] Step 1: Collect multi-source road environment data at the intersection;

[0007] Step 2: Use the network analysis method to obtain the stable weights of various road environment data at different intersections; use the entropy weight method to obtain the entropy weights of various road environment data at all intersections at the current moment; forward fuse the stable weights and entropy weights to determine the adjustment coefficients of various road environment data at different intersections at the current collection moment to adjust the input weight matrix of the LSTM model;

[0008] Step 3: Use the preset dataset to train the adjusted LSTM model to predict the saturated flow and phase flow ratio at the next acquisition time at different intersections. The predicted results are used to update the corresponding input data of the DQN model in real time. The DQN model is trained using the preset dataset to obtain the phase time series of the current control cycle at different intersections to adjust the phase of the corresponding traffic lights.

[0009] Preferably, in step 1, the road environment data includes saturated flow, phase flow ratio, maximum design flow, temperature, humidity and illumination.

[0010] Preferably, saturation flow refers to the highest traffic flow passing through the stop line of the intersection within a unit time period, phase flow ratio refers to the ratio of the actual flow at the intersection at different phases of the traffic light to the saturation flow, and maximum design flow refers to the maximum value of the design traffic volume of all roads at the intersection.

[0011] Preferably, in step 2, the input of the network analysis method is the absolute value of the Pearson correlation coefficient between different road condition environment data at different intersections.

[0012] Preferably, in step 2, the input of the entropy weight method is various types of road environment data at different intersections at the current moment.

[0013] Preferably, in step 2, before adjusting the input weight matrix of the LSTM model using the adjustment coefficient, each column of the input weight matrix corresponds to the initial input weight of different types of road environment data at different intersections, which is obtained by adopting Xavier initialization, and the number of rows of the input weight matrix is determined by the number of hidden layer neurons of the LSTM model.

[0014] Preferably, in step 2, each adjustment coefficient is multiplied by each corresponding column of data in the input weight matrix to make adjustments.

[0015] Preferably, in step three, the preset data set is: a sequence consisting of multi-source road condition environment data, a phase time series, multi-source road condition environment data at the next collection moment, and a timestamp of the current data collection moment at each collection moment within the preset data collection time range at different intersections; wherein the phase time series is the time length that the traffic light control device controls different phases of the traffic light at the collection moment.

[0016] In a second aspect, embodiments of the present application provide a traffic linkage control system based on road condition and environment perception, which implements any of the above-described traffic linkage control methods based on road condition and environment perception, and the system includes:

[0017] The road condition environment perception module is used to collect multi-source road condition environment data through various data acquisition devices installed at the intersection and send the data to the road condition data processing module;

[0018] The road condition data processing module is used to analyze the multi-source road condition data obtained from the road condition environment perception module based on the network analysis method and entropy weight method, and use the LSTM model and DQN model to train the traffic linkage control model;

[0019] The traffic linkage control module is used to adjust the phase time according to the phase time series predicted by the traffic linkage control model to achieve phase control of traffic lights.

[0020] In the third aspect, an embodiment of the present application also provides a traffic linkage control device based on road condition and environment perception, the device including a memory, a processor, and a computer program stored in the memory and running on the processor, and when the processor executes the computer program, it implements any one of the steps of the above-mentioned traffic linkage control method based on road condition and environment perception.

[0021] As can be seen from the above embodiments, the traffic linkage control method, system, and device based on road condition perception provided by the embodiments of the present application have at least the following beneficial effects:

[0022] This application acquires multi-source road condition and environment data by building a road condition and environment perception module, and analyzes and acquires the influence of the interaction between different road condition and environment factors on the road condition state, thereby reducing the training error introduced by the redundant information in the multi-source road condition and environment data during the subsequent prediction model training, and improving the generalization ability and prediction accuracy of the prediction model. This application adjusts the calculation method of the LSTM model input weights and the training process of the prediction model according to the degree of influence of the interaction between different road condition and environment factors on the road condition state, enhances the learning ability of the hidden information of different road condition and environment factors during the model training process, improves the accuracy and reliability of the prediction of the road condition state, and thus improves the overall effectiveness of the traffic linkage control. In summary, this application aims to consider the influence of other multi-source road condition and environment data on the road condition state in addition to the number of vehicles at the traffic intersection, improve the accuracy of the LSTM model in predicting the road condition state, and thus improve the effectiveness of the subsequent traffic linkage control. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present application or the prior art, the following is a brief introduction to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0024] Figure 1A flowchart of the steps of a traffic linkage control method based on road condition perception provided in one embodiment of the present application;

[0025] Figure 2 A schematic structural diagram of a traffic linkage control system based on road condition and environment perception provided in one embodiment of the present application. DETAILED DESCRIPTION

[0026] In order to further illustrate the technical means and effects adopted by this application to achieve the predetermined invention purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, describes in detail the specific implementation methods, structures, features and effects of the traffic linkage control method, system and device based on road environment perception proposed in this application. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics of one or more embodiments may be combined in any suitable form.

[0027] Unless otherwise specified and limited, terms such as "comprises", "includes" or any other variants thereof are intended to cover non-exclusive inclusion, so that a circuit structure, article or device comprising a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the presence of other identical elements in the article or device comprising the element. In addition, the term "and\or" used herein includes any and all combinations of one or more related listed items. All technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs.

[0028] The specific solutions of the traffic linkage control method, system and equipment based on road condition and environment perception provided by this application are described in detail below with reference to the accompanying drawings.

[0029] See also Figure 1 , which shows a flow chart of the steps of a traffic linkage control method based on road condition environment perception provided by an embodiment of the present application, the method comprising the following steps:

[0030] Step 1: Collect multi-source road environment data at the intersection.

[0031] In traffic coordination control, accurate road condition prediction directly impacts the effectiveness of subsequent traffic light phase control. Because road conditions are affected by a variety of environmental factors, accurate road condition prediction requires the support of multi-source road condition data.

[0032] Considering the importance of multi-source road condition data in predicting road conditions in traffic linkage control, the road condition perception module of this application collects and acquires multi-source road condition data by setting up multiple data collection devices at intersections.

[0033] Specifically, this application provides an implementation method for building data acquisition equipment and collecting multi-source road condition environment data as follows. The implementer can appropriately adjust the collected multi-source road condition environment data and the type of data acquisition equipment that collects the corresponding data according to the specific traffic conditions of the location.

[0034] This embodiment performs traffic linkage control on a fixed area in a city, with a control cycle of 15 minutes. The number of intersections in the fixed area is 20, and two-phase traffic lights are installed at all intersections.

[0035] At each intersection in a fixed area, high-definition surveillance cameras, temperature sensors, humidity sensors, and illumination sensors are installed to capture images of vehicles at the intersection, as well as temperature, humidity, and illumination. Temperature, humidity, and illumination reflect the weather and light conditions at the intersection, which together influence the speed at which vehicles pass through the intersection and, in turn, the prediction of road conditions. Therefore, these three factors serve as the road environment data in this application.

[0036] This application installs wireless communication modules in different devices and uses LTE-Advanced Pro technology to achieve wireless data transmission between data acquisition equipment, a host computer, and traffic light control equipment. Specifically, this embodiment uses the Huawei B618-22d wireless communication module. Implementers can use other communication technologies and communication equipment to achieve data transmission between different devices.

[0037] To accurately reflect the traffic conditions at intersections and accurately control traffic lights based on these conditions, this application uses the intersection's saturated flow rate and phase flow rate ratio as road environment data for evaluating the intersection's traffic conditions and as a reference for subsequent traffic light phase timing control. Saturated flow rate refers to the maximum traffic flow passing through the intersection's stop line within a unit time period, and phase flow rate ratio refers to the ratio of the intersection's actual flow rate at different traffic light phases to the saturated flow rate. The specific methods for obtaining both are as follows:

[0038] Because the viewing angle of high-definition surveillance cameras at intersections remains constant, the stop lines are manually annotated in the intersection vehicle images. Using the intersection vehicle images as input, a trained YOLOv5n model is used to output vehicle bounding boxes. The number of vehicles whose vehicle bounding boxes have crossed the stop line for the first time within the previous 15 minutes is calculated and recorded as the saturation flow rate for that intersection.

[0039] At the same time, the time information of different signal light phases in the signal light control device is synchronized to the host computer through the wireless communication module to obtain the actual traffic flow at the intersection at different signal light phases in the previous 15 minutes, and the phase flow ratio of different phases is calculated using the above method.

[0040] Furthermore, the road design traffic volume reflects the peak number of vehicles allowed to pass through the road. Therefore, the maximum of all road design traffic volumes at an intersection is recorded as the intersection's maximum design flow rate, reflecting the maximum traffic flow the intersection can withstand. The road design traffic volume can be obtained by querying statistical files from the transportation department.

[0041] At this point, the saturated flow, phase flow ratio, maximum design flow, temperature, humidity, and illumination at different intersections in fixed areas within the city are obtained. The above six types of data are numbered in sequence and recorded as the road condition environment data of the intersection, which serves as a reference for subsequent prediction of road conditions.

[0042] Considering that both the road condition prediction model and the traffic linkage control model need to be trained through a dataset, this application provides the following implementation method for establishing a dataset:

[0043] Using the aforementioned road condition data acquisition method, all road condition data is collected every 15 minutes at 20 intersections in a fixed area within the city. This data sequence is constructed in the order of saturated flow rate, phase flow ratio, maximum design flow rate, temperature, humidity, and illumination. The number of intersections and the duration of data collection can be customized by the implementer.

[0044] At the same time, at each data collection moment, the time lengths that the traffic light control device controls different phases of the traffic light at the collection moment are obtained, and a phase time series is formed.

[0045] In addition, in order to realize the supervised learning of the road condition prediction model and the traffic linkage control model, the road condition environment data sequence at the next data collection moment is used as the evaluation of the above two models, where the interval between two adjacent data collection moments is 15 minutes.

[0046] The road condition data sequence for a single intersection at each data collection moment, along with the phase time sequence, the multi-source road condition data for the next collection moment, and the timestamp of the current data collection moment, are arranged in sequence to form the intersection data sequence for that intersection. Furthermore, the different intersections are numbered, and the data sequences of 20 intersections at the same data collection moment are combined in a fixed order to form a training data sequence for the dataset.

[0047] The above method is used to continuously collect training data sequences of different intersections in a fixed area of the city throughout the year as a data set for training the road condition prediction model and the traffic linkage control model. The number of training data sequences in the data set is 35,040. In this embodiment, the preset data collection time range is one year, and 20 intersections are collected each time. In other embodiments, the data collection time range and the number of collected intersections can be set by the implementer.

[0048] Step 2: Use the network analysis method to obtain the stable weights of various road environment data at different intersections; use the entropy weight method to obtain the entropy weights of various road environment data at all intersections at the current moment; forward fuse the stable weights and entropy weights to determine the adjustment coefficients of various road environment data at different intersections at the current collection moment to adjust the input weight matrix of the LSTM model.

[0049] When controlling traffic light phases, solely considering traffic conditions at a specific intersection can alleviate local congestion while exacerbating pressure at adjacent intersections. For example, a prolonged green light at one intersection can cause a large backlog of vehicles at downstream intersections, ultimately reducing traffic efficiency across the entire city.

[0050] To achieve coordinated traffic control within an urban area, a specific intersection is designated as the source intersection, and the K-1 nearest intersections are designated as its neighboring intersections. The source intersection and its neighboring intersections are collectively designated as reference intersections for subsequent road condition prediction, where K is set to 5 in this example. The reference intersections are then numbered from smallest to largest based on their distance from the source intersection, with the source intersection being numbered 1.

[0051] The impact of road condition data on road conditions varies between different reference intersections. This application uses the Analytical Network Process (ANP) to analyze and calculate the stability weights between different road condition data to reflect the degree of influence between intersection road condition data. The specific implementation method is as follows:

[0052] For each type of road condition data at the reference intersection, take the T types of road condition data collected previously, and construct its time series in chronological order to reflect the changes in the road condition data. In this embodiment, the value of T is 5, which can be set by the implementer.

[0053] This application uses the Pearson correlation coefficient to measure the degree of influence between different road environment data at different reference intersections. Since the Pearson correlation coefficient ranges from [-1, 1], where 1 indicates a perfect positive correlation, -1 indicates a perfect negative correlation, and 0 indicates no correlation, the larger the absolute value of the Pearson correlation coefficient, the greater the degree of mutual influence between the two types of data.

[0054] Furthermore, the Pearson correlation coefficient of the time series of different road environment data between different reference intersections is calculated, and its absolute value is recorded as the influence weight of the two.

[0055] As an extended decision analysis method, network analysis method has strong analysis capabilities for the mutual influence of intersection road environment data and the interdependence of different intersections.

[0056] This application takes the influence weights of various types of road condition and environmental data between all reference intersections as input, constructs a super matrix in the network analysis method, and further constructs a weighted super matrix using the network analysis method, and obtains the stable weights of various types of road condition and environmental data in different reference intersections, reflecting the degree to which the road condition and environmental data in the reference intersection are affected by the source intersection and its adjacent intersections.

[0057] On the one hand, the absolute value of the Pearson correlation coefficient of different road environment data at different reference intersections is used as input to improve the interpretability of the network analysis method for the degree of influence between road environment data at each moment, avoiding the analytical errors introduced by artificially setting fixed weights in traditional network analysis methods.

[0058] On the other hand, the stability weights output by the network analysis method reflect the mutual influence of the remaining road environment data at different reference intersections. The larger the stability weight, the stronger the correlation between the road environment data at the reference intersection and the surrounding intersections, and the greater its role in subsequent road condition prediction. The process of setting up a typical network structure for the network analysis method and calculating the stability weights is well known.

[0059] Among them, the network analysis method is a well-known technology and will not be described in detail in this application.

[0060] Considering that the distribution of road conditions at intersections at different times of the day is different, this application uses the entropy weight method to obtain the relative impact of various types of road condition environment data on the road conditions at different times.

[0061] Specifically, the data for all types of road conditions at all intersections with the same timestamp in the dataset is used as input. The entropy weighting method is then used to output the entropy weights of each type of road condition at that time, reflecting their relative impact on the current road condition. The larger the entropy weight, the greater the predictive information provided by the corresponding road condition data, and the greater its usefulness in training the road condition prediction model.

[0062] Furthermore, the stability weight and entropy weight are forward fused to determine the adjustment coefficients of various road environment data at different intersections at the current moment, so as to adjust the input weight matrix of the LSTM model.

[0063] It can be understood that forward fusion is a fusion method such as addition and multiplication between data. The specific forward fusion method is determined by the implementer according to the actual situation, and this application does not impose any special restrictions.

[0064] In this embodiment, the adjustment coefficient of each type of road condition data at the source intersection at the current moment is obtained based on the entropy weight and the stability weight:

[0065]

[0066] in, Indicates the entropy weight of the nth type of road condition data at the current data collection moment; It represents the stable weight of the nth type of road condition data in the source intersection, reflecting the degree to which the road condition data in the source intersection is affected by the other road condition data of the source intersection and its adjacent intersections; norm() represents the normalization function, the purpose of which is to limit the calculation range of the adjustment coefficient; It represents the adjustment coefficient of the nth type of road condition environment data in the source intersection. It is used in the subsequent training of the road condition prediction model to adjust the input weight of the training data, improve the correlation between important data, and reduce the prediction noise of redundant data.

[0067] The larger the adjustment coefficient of the intersection's road condition data, on the one hand, reflects the greater the correlation between the intersection and its surrounding intersections' road conditions; on the other hand, it also reflects that the road condition data contains more information on road condition prediction, and the stronger its effect on predicting road conditions.

[0068] This application uses an LSTM model as the basis for training a traffic condition prediction model. The input of the LSTM model used in this application is the traffic environment data of the source intersection and all adjacent intersections. The output is the saturation flow rate and phase flow rate ratio at the next data collection time of the source intersection, which serves as a prediction of the traffic state of the source intersection. In this embodiment, the number of hidden layer neurons of the LSTM model is set to 10.

[0069] The input weight matrix of the LSTM model determines the degree of influence of the input features on the hidden state. By adjusting this matrix, the role of important features can be enhanced and the interference of redundant features can be reduced.

[0070] The number of rows in the input weight matrix is equal to the number of hidden layer neurons in the LSTM model, which is 10. The number of columns is set to 30, that is, the six types of road environment data at the five intersections correspond to the weight values in each column of the input weight matrix, and the weight values correspond to the road environment data of the source intersection and all adjacent intersections in turn. In this application, the initial input weight matrix is initialized using Xavier.

[0071] Furthermore, according to the above-calculated adjustment coefficient, each corresponding column of data in the input weight matrix is multiplied to make an adjustment, so as to obtain an adjusted input weight matrix.

[0072] The adjustment coefficient is obtained by calculating the entropy weights and stable weights of various road environment data, and the input weight matrix after adjustment of the LSTM model is further obtained.

[0073] On the one hand, the entropy weight of the road environment data reflects the relative amount of information it provides to the model prediction, thereby improving the utilization efficiency of road environment data containing more information and enhancing the learning ability of the prediction model. On the other hand, the stable weights obtained through network analysis methods incorporate the degree of mutual influence between different road environment data at each intersection, thereby reflecting the importance of different road environment data to the learning of prediction results. This enhances the ability to learn hidden information of different road environment factors during model training and improves the accuracy and reliability of road condition predictions. The above calculation method is used to obtain the input weight matrix of the LSTM model, which can improve the accuracy of road condition predictions.

[0074] Step 3: Use the preset dataset to train the adjusted LSTM model to predict the saturated flow and phase flow ratio at the next acquisition time at different intersections. The predicted results are used to update the corresponding input data of the DQN model in real time. The DQN model is trained using the preset dataset to obtain the phase time series of the current control cycle at different intersections to adjust the phase of the corresponding traffic lights.

[0075] Based on the adjusted LSTM model, the data set constructed in step 1 is used to complete the training of the road condition prediction model to predict the saturated flow and phase flow ratio at the next collection time of different intersections.

[0076] The multi-source road condition environment data of the source intersection and all its adjacent intersections obtained at the current collection moment are used as input, and the trained road condition prediction model is used to output the saturated flow and phase flow ratio predicted at the next data collection moment of the source intersection.

[0077] Furthermore, this application uses the DQN model as a foundation and the dataset from step 1 to train a traffic linkage control model. This model is used to obtain the phase time series of the current control cycle at different intersections to adjust the phase of the corresponding traffic lights. The current control cycle is the time period between the current acquisition time and the next acquisition time.

[0078] The multi-source road condition data and phase time series of the source intersection and its adjacent intersections at the current data collection time, as well as the saturated flow and phase flow ratio predicted for the source intersection and its adjacent intersections at the next data collection time are used as input. The trained traffic linkage control model is used to output the adjusted phase time series of the source intersection in this control cycle. The phase control of traffic lights is completed through the signal light control device based on the phase times of the signal lights in the phase time series.

[0079] See also Figure 2 , Figure 2 This is a schematic diagram of a traffic linkage control system based on road condition perception provided by an embodiment of the present application. The system includes: a road condition perception module, a road condition data processing module, and a traffic linkage control module.

[0080] The road condition environment perception module is used to collect multi-source road condition environment data through various data acquisition devices installed at the intersection and send the data to the road condition data processing module;

[0081] The road condition data processing module is used to analyze the multi-source road condition data obtained from the road condition environment perception module based on the network analysis method and entropy weight method, and use the LSTM model and DQN model to train the traffic linkage control model;

[0082] The traffic linkage control module is used to adjust the phase time according to the phase time series predicted by the traffic linkage control model to achieve phase control of traffic lights.

[0083] Based on the same inventive concept as the above method, another embodiment of the present application also provides a traffic linkage control device based on road condition environment perception, which 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, it implements the steps of any one of the above-mentioned traffic linkage control methods based on road condition environment perception.

[0084] The various embodiments in this application are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments.

[0085] It should be noted that, unless otherwise specified and limited, terms such as "include", "comprising" or any other variations thereof are intended to cover non-exclusive inclusion, so that a circuit structure, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such article or device. In the absence of further restrictions, the phrase "including a ..." defines an element, does not exclude the presence of other identical elements in the article or device including the element. In addition, the term "and\or" used herein includes any and all combinations of one or more related listed items.

[0086] Those skilled in the art will readily appreciate other embodiments of the present application after considering the specification and practicing the invention herein. This application is intended to cover any variations, uses, or adaptations of the present application that follow the general principles of the present application and include common knowledge or customary techniques in the art not invented herein.

[0087] It will be understood that the present application is not limited to the exact construction that has been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof.

Claims

1. A traffic linkage control method based on road condition and environment perception, characterized in that: The method comprises the following steps: Step 1: Collect multi-source road environment data at the intersection; Step 2: Use the network analysis method to obtain the stable weights of various road condition data at different intersections; use the entropy weight method to obtain the entropy weights of various road condition data at all intersections at the current moment; forward fuse the stable weights and entropy weights to determine the adjustment coefficients of various road condition data at different intersections at the current collection moment to adjust the input weight matrix of the LSTM model; wherein the input of the network analysis method is the absolute value of the Pearson correlation coefficient between different road condition data at different intersections; the larger the stable weight, the stronger the correlation between the road condition data at the intersection and the surrounding intersections; the input of the entropy weight method is various road condition data at different intersections at the current moment; the larger the entropy weight, the greater the amount of prediction information provided by the corresponding road condition data; Step 3: Use the preset dataset to train the adjusted LSTM model to predict the saturated flow and phase flow ratio at the next acquisition time at different intersections. The predicted results are used to update the corresponding input data of the DQN model in real time. The DQN model is trained using the preset dataset to obtain the phase time series of the current control cycle at different intersections to adjust the phase of the corresponding traffic lights.

2. The traffic linkage control method based on road condition perception according to claim 1, characterized in that: In step 1, the road environment data includes saturated flow, phase flow ratio, maximum design flow, temperature, humidity and illumination.

3. The traffic linkage control method based on road condition perception as claimed in claim 2, characterized in that: Saturation flow refers to the maximum traffic flow passing through the stop line of an intersection within a unit time period. Phase flow ratio refers to the ratio of the actual flow at the intersection at different phases of the traffic lights to the saturation flow. Maximum design flow refers to the maximum value of the design traffic volume of all roads at the intersection.

4. The traffic linkage control method based on road condition perception according to claim 1, characterized in that: In step 2, before adjusting the input weight matrix of the LSTM model using the adjustment coefficient, each column of the input weight matrix corresponds to the initial input weight of different types of road environment data at different intersections, which is obtained by using Xavier initialization. The number of rows of the input weight matrix is determined by the number of hidden layer neurons of the LSTM model.

5. The traffic linkage control method based on road condition perception as claimed in claim 4, characterized in that: In step 2, each adjustment coefficient is multiplied by each corresponding column of data in the input weight matrix to make adjustments.

6. The traffic linkage control method based on road condition perception according to claim 1, characterized in that: In step three, the preset data set is: a sequence consisting of multi-source road condition environment data, a phase time series, multi-source road condition environment data at the next collection time, and a timestamp of the current data collection time at each collection time of different intersections within the preset data collection time range; wherein the phase time series is the time length of the traffic light control device controlling different phases of the traffic light at the collection time.

7. Traffic linkage control system based on road condition perception, characterized by: To implement the traffic linkage control method based on road condition environment perception according to any one of claims 1 to 6, the system comprises: The road condition environment perception module is used to collect multi-source road condition environment data through various data acquisition devices installed at the intersection and send the data to the road condition data processing module; The road condition data processing module is used to analyze the multi-source road condition data obtained from the road condition environment perception module based on the network analysis method and entropy weight method, and use the LSTM model and DQN model to train the traffic linkage control model; The traffic linkage control module is used to adjust the phase time according to the phase time series predicted by the traffic linkage control model to achieve phase control of traffic lights.

8. A traffic linkage control device based on road condition and environment perception, 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 traffic linkage control method based on road condition environment perception as described in any one of claims 1 to 6 is implemented.

Images