Flow rate prediction method, device and equipment

Abstract

The invention provides a flow rate prediction method, device and equipment. The method comprises the following steps: acquiring flow data of a target intersection in a plurality of historical signal periods before a current signal period; the flow data comprises the flow of each lane and the signal period duration of the historical signal period; the signal period durations of different historical signal periods are the same or different; for each historical signal period, determining a historical flow rate corresponding to each lane in the historical signal period based on the flow data of the historical signal period; wherein the historical flow rate corresponding to each lane in the historical signal period is determined based on the flow of the lane and the duration of the signal period; and predicting a target flow rate corresponding to each lane in a target signal period based on the historical flow rates corresponding to each lane in the plurality of historical signal periods, wherein the target signal period is located behind the current signal period. Through the technical scheme of the invention, the flow rate in a period of time in the future can be predicted, and the prediction accuracy is improved.

Description

Technical Field

[0001] The present application relates to the field of intelligent transportation technology, and in particular to a flow rate prediction method, device and equipment. Background Art

[0002] Accurate and real-time traffic forecasting is an important part of the intelligent transportation system and is of great significance to urban traffic planning, traffic management and traffic control. For example, in the intelligent transportation system, it is often necessary to fine-tune the signal cycle and phase in combination with indicators such as traffic flow in the future to better meet actual traffic needs. Therefore, it is necessary to predict the traffic flow in the future.

[0003] However, there is no reasonable way to predict the traffic in the future in the relevant technology, and there are problems such as low prediction accuracy and inability to obtain accurate and reliable traffic. Summary of the invention

[0004] The present application provides a flow rate prediction method, the method comprising:

[0005] Obtaining traffic data of multiple historical signal cycles before the current signal cycle at the target intersection; wherein the traffic data includes the traffic of each lane and the signal cycle duration of the historical signal cycle; wherein the signal cycle durations of different historical signal cycles are the same or different;

[0006] For each historical signal cycle, determine the historical flow rate of each lane corresponding to the historical signal cycle based on the traffic data of the historical signal cycle; wherein the historical flow rate of each lane corresponding to the historical signal cycle is determined based on the traffic flow of the lane and the length of the signal cycle;

[0007] A target flow rate corresponding to each lane in a target signal cycle is predicted based on the historical flow rates corresponding to each lane in the plurality of historical signal cycles, and the target signal cycle is located after the current signal cycle.

[0008] The present application provides a flow rate prediction device, the device comprising:

[0009] An acquisition module, used to acquire the flow data of a plurality of historical signal cycles before the current signal cycle of the target intersection; wherein the flow data includes the flow of each lane and the signal cycle duration of the historical signal cycle; wherein the signal cycle durations of different historical signal cycles are the same or different;

[0010] A processing module, for determining, for each historical signal cycle, a historical flow rate corresponding to each lane in the historical signal cycle based on the flow data of the historical signal cycle; wherein the historical flow rate corresponding to each lane in the historical signal cycle is determined based on the flow rate of the lane and the length of the signal cycle;

[0011] A target flow rate corresponding to each lane in a target signal cycle is predicted based on the historical flow rates corresponding to each lane in the plurality of historical signal cycles, and the target signal cycle is located after the current signal cycle.

[0012] The present application provides an electronic device, comprising: a processor and a machine-readable storage medium, wherein the machine-readable storage medium stores machine-executable instructions that can be executed by the processor; the processor is used to execute the machine-executable instructions to implement the flow rate prediction method of the above example of the present application.

[0013] It can be seen from the above technical solutions that in the embodiments of the present application, the target flow rate corresponding to the target signal cycle of each lane can be predicted based on the flow data of multiple historical signal cycles, and the flow rate in the future period (i.e., the target flow rate corresponding to the target signal cycle of each lane) can be predicted, thereby improving the prediction accuracy and the prediction precision, and obtaining an accurate and reliable flow rate. By introducing the flow rate of each lane at the signal cycle level and the signal cycle duration, the flow prediction is converted into a flow rate prediction, thereby realizing the flow rate prediction at the signal cycle level, and the flow prediction result can also be obtained based on the flow rate prediction result, thereby realizing the flow prediction at the signal cycle level, that is, a signal cycle level urban road network lane flow prediction method is proposed, which can obtain the prediction result at the signal cycle level and realize the flow rate prediction and flow prediction at the signal cycle level. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments of the present application or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this application. For ordinary technicians in this field, other drawings can also be obtained based on these drawings of the embodiments of the present application.

[0015] Figure 1 is a flow chart of a flow rate prediction method in one embodiment of the present application;

[0016] Figure 2 This is a schematic diagram of a process for obtaining flow data in one embodiment of the present application;

[0017] Figure 3 is a schematic diagram of a process for predicting a target flow rate in one embodiment of the present application;

[0018] Figure 4is a schematic diagram of a process for predicting a target flow rate and a target flow rate in one embodiment of the present application;

[0019] Figure 5A is a schematic diagram of the structure of a target prediction model in one embodiment of the present application;

[0020] Figure 5B is a schematic diagram of a training process of a target prediction model in one embodiment of the present application;

[0021] Figure 5C is a schematic diagram of a target flow rate prediction process in one embodiment of the present application;

[0022] Fig. 6A is a schematic diagram of the structure of a target prediction model in one embodiment of the present application;

[0023] Figure 6B is a schematic diagram of a training process of a target prediction model in one embodiment of the present application;

[0024] Figure 6C is a schematic diagram of a target flow rate and target flow prediction process in one embodiment of the present application;

[0025] Figure 7 is a schematic structural diagram of a flow rate prediction device in one embodiment of the present application;

[0026] Figure 8 It is a hardware structure diagram of an electronic device in one embodiment of the present application. DETAILED DESCRIPTION

[0027] The terms used in the embodiments of the present application are only for the purpose of describing specific embodiments, rather than limiting the present application. The singular forms of "a", "said" and "the" used in the present application and claims are also intended to include plural forms, unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used herein refers to any or all possible combinations of one or more associated listed items.

[0028] It should be understood that, although the terms first, second, third, etc. may be used to describe various information in the embodiments of the present application, these information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of the present application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, in addition, the word "if" used may be interpreted as "at..." or "when..." or "in response to determination".

[0029] A flow rate prediction method is proposed in the embodiment of the present application, see Figure 1 As shown, the method includes:

[0030] Step 101, obtaining traffic data of a target intersection for multiple historical signal cycles before the current signal cycle; wherein the traffic data may include the traffic of each lane and the signal cycle duration of the historical signal cycle; wherein the signal cycle durations of different historical signal cycles may be the same or different.

[0031] Step 102: for each historical signal cycle, determine the historical flow rate of each lane in the historical signal cycle based on the traffic data of the historical signal cycle; wherein the historical flow rate of each lane in the historical signal cycle can be determined based on the traffic flow of the lane and the length of the signal cycle.

[0032] Step 103 : predicting a target flow rate corresponding to a target signal cycle for each lane based on the historical flow rates corresponding to each lane in a plurality of historical signal cycles, where the target signal cycle is located after the current signal cycle.

[0033] Exemplarily, the signal cycle duration corresponding to the target signal cycle can also be obtained; for each lane, the target flow rate corresponding to the lane in the target signal cycle can be determined based on the target flow rate corresponding to the lane in the target signal cycle and the signal cycle duration of the target signal cycle.

[0034] Exemplarily, based on the target flow rate corresponding to the lane in the target signal cycle and the signal cycle duration of the target signal cycle, determining the target flow rate corresponding to the lane in the target signal cycle may include but is not limited to: i , the signal cycle duration C corresponding to the target signal cycle, and the obtained weight parameter matrix W i , determine the target flow rate Q corresponding to the lane in the target signal cycle i For example, the target flow rate Q corresponding to the target signal cycle of the lane is determined by the following formula: i :Q i =W i Cv i Of course, the above formula is only an example and is not intended to be limiting.

[0035] Exemplarily, determining the historical flow rate of each lane corresponding to the historical signal cycle based on the traffic data of the historical signal cycle may include, but is not limited to: using the following formula to determine the historical flow rate of the lane corresponding to the historical signal cycle: In the above formula, v i，j represents the historical flow rate of the i-th lane in the j-th historical signal cycle, Q i，j represents the traffic volume of the i-th lane in the j-th historical signal cycle, C jIndicates the signal cycle length of the j-th signal cycle.

[0036] In a possible implementation, the weight parameter matrix may be a parameter in a target prediction model, and the target prediction model may include but is not limited to a conversion layer and a prediction layer; based on this, the flow data of multiple historical signal cycles may be input to the conversion layer, and the historical flow rate corresponding to each lane in multiple historical signal cycles is determined by the conversion layer based on the flow data of multiple historical signal cycles. Then, the historical flow rate corresponding to each lane in multiple historical signal cycles is input to the prediction layer, and the prediction layer predicts the target flow rate corresponding to each lane in the target signal cycle based on the historical flow rate.

[0037] In a possible implementation, the target prediction model may further include an inverse conversion layer, based on which the target flow rate corresponding to each lane in the target signal cycle and the signal cycle duration of the target signal cycle may be input into the inverse conversion layer; the target flow rate corresponding to each lane in the target signal cycle is determined by the inverse conversion layer based on the signal cycle duration of the target signal cycle, the target flow rate and the weight parameters in the inverse conversion layer.

[0038] Exemplarily, the target prediction model can also be trained by the following steps: sample data and label data are obtained, the sample data can include first flow data of multiple sample signal cycles, the first flow data includes the flow and signal cycle duration of each lane; the label data can include second flow data of future signal cycles of multiple sample signal cycles, the second flow data includes the actual flow rate of each lane and the actual flow of each lane. The sample data is input to the initial prediction model, so that the initial prediction model determines the sample flow rate corresponding to each lane in the future signal cycle based on the sample data, and determines the sample flow corresponding to each lane in the future signal cycle based on the sample data; based on the difference between the sample flow rate and the actual flow rate, and the difference between the sample flow and the actual flow, the loss value is calculated, the network parameters of the initial prediction model are adjusted based on the loss value, and the target prediction model is determined based on the adjusted prediction model.

[0039] Exemplarily, the signal period is the total period required to release all phases of the target intersection once.

[0040] It can be seen from the above technical solutions that in the embodiments of the present application, the target flow rate corresponding to the target signal cycle of each lane can be predicted based on the flow data of multiple historical signal cycles, and the flow rate in the future period (i.e., the target flow rate corresponding to the target signal cycle of each lane) can be predicted, thereby improving the prediction accuracy and the prediction precision, and obtaining an accurate and reliable flow rate. By introducing the flow rate of each lane at the signal cycle level and the signal cycle duration, the flow prediction is converted into a flow rate prediction, thereby realizing the flow rate prediction at the signal cycle level, and the flow prediction result can also be obtained based on the flow rate prediction result, thereby realizing the flow prediction at the signal cycle level, that is, a signal cycle level urban road network lane flow prediction method is proposed, which can obtain the prediction result at the signal cycle level and realize the flow rate prediction and flow prediction at the signal cycle level.

[0041] The above technical solutions of the embodiments of the present application are described below in combination with specific application scenarios.

[0042] Before introducing the technical solutions of the embodiments of the present application, technical terms related to the present application are first introduced.

[0043] Equal interval prediction and non-equal interval prediction: Equal interval prediction refers to prediction based on time series data sampled at equal intervals, such as collecting multiple 5-minute time series data and making predictions based on multiple 5-minute time series data. Non-equal interval prediction refers to prediction based on time series data sampled at non-equal intervals (i.e., time series data with uneven sampling intervals), such as collecting 1-minute time series data, 2-minute time series data, 3-minute time series data, etc., and making predictions based on these time series data.

[0044] Signal control scheme: Based on the signal control scheme, traffic lights can be used to allocate the right of way to vehicles or pedestrians in certain directions in turn. That is, the signal control scheme is used to control the order of traffic lights in each direction.

[0045] Target intersection: The intersection to be predicted is called the target intersection, which can be any intersection. The target intersection can correspond to multiple phases. A phase is the signal state corresponding to one or more traffic flows that obtain the right of way at the same time. There can be multiple phases in a signal cycle. For example, the east-west direction is one phase and the north-south direction is another phase. The two constitute a signal cycle.

[0046] In addition, the target intersection may include multiple lanes. In this application, flow rate prediction and flow volume prediction are performed based on the lane granularity, such as 3 lanes in the north-south direction and 4 lanes in the east-west direction.

[0047] Signal cycle: the total cycle required to release all phases of the target intersection once, that is, the total cycle length required for the signal control scheme to release all phases once. For example, assuming that the target intersection corresponds to phase A and phase B, the time required to release phase A once is 50 seconds, and the time required to release phase B once is 100 seconds, then the signal cycle is 150 seconds. The total duration of different signal cycles can be the same or different. For example, in the time period from 8:00 to 9:00, the total duration of each signal cycle is 150 seconds, and in the time period from 9:00 to 10:00, the total duration of each signal cycle is 200 seconds.

[0048] Effective green light duration: the length of effective vehicle passage time remaining within a signal cycle or a single phase, after deducting the start-up loss time, driver reaction loss time, yellow light time and full red light time.

[0049] Traffic volume: the number of vehicles that pass through the target intersection via the lane within a certain statistical period, such as 5-minute traffic volume, 15-minute traffic volume, etc. For example, 5-minute traffic volume refers to the number of vehicles that pass through the target intersection via the lane within 5 minutes.

[0050] Flow rate: The number of vehicles passing through the target intersection through the lane within a unit time. If the statistical time is less than 1 hour, the number of vehicles passing per hour is calculated by dividing the flow rate by the length of the statistical period.

[0051] In intelligent transportation systems, it is often necessary to fine-tune the signal cycle and phase in combination with indicators such as traffic flow in the future to meet actual traffic needs. Therefore, it is necessary to predict traffic flow in the future. However, there is no reasonable prediction method in related technologies for predicting traffic flow in the future, and there are problems such as low prediction accuracy and inability to obtain accurate and reliable traffic flow.

[0052] For example, in related technologies, an equal-interval prediction scheme is usually adopted, that is, multiple time series data (i.e., traffic data) are sampled at equal intervals, such as collecting multiple 5-minute traffic data, and prediction is made based on multiple 5-minute traffic data. However, when the traffic data is collected at a 5-minute granularity, the 5-minute traffic data will include traffic data of multiple signal cycles, and the last signal cycle in the 5-minute time window may be cut off. When the traffic data at a 5-minute granularity is used to predict the traffic in the future, it is impossible to predict accurate and reliable traffic, which will result in a large loss of accuracy.

[0053] In response to the above findings, a flow rate prediction method is proposed in an embodiment of the present application. By introducing the flow rate of each lane at the signal cycle level and the signal cycle duration, the flow prediction is converted into a flow rate prediction, thereby realizing the flow rate prediction at the signal cycle level. The flow prediction result can also be obtained based on the flow rate prediction result, thereby realizing the flow prediction at the signal cycle level. That is, a method for predicting the lane flow of an urban road network at the signal cycle level is proposed, which can obtain the prediction result at the signal cycle level and realize the flow rate prediction and flow prediction at the signal cycle level.

[0054] By using the traffic data of the signal cycle for prediction, non-equal interval prediction can be achieved, that is, the traffic data of the signal cycle of 1 minute, the traffic data of the signal cycle of 2 minutes, the traffic data of the signal cycle of 3 minutes, etc. are collected, and non-equal interval prediction is achieved based on the traffic data of these signal cycles. The traffic in the future period is predicted with the traffic data at the signal cycle level, and accurate and reliable prediction results can be obtained. By proposing a signal cycle level road network lane traffic prediction scheme based on license plate recognition data, by introducing the traffic data of each lane at the signal cycle level and the signal timing scheme, the lane traffic prediction at the signal cycle level is achieved, and the problem of precision loss of the signal control algorithm caused by equal interval prediction is solved.

[0055] It should be noted that all data involved in this embodiment (such as traffic data, signal cycle duration, etc.) will only be obtained and used with the knowledge and authorization of the relevant users.

[0056] First, in order to achieve flow rate prediction at the signal cycle level, it is necessary to obtain traffic data at the signal cycle level. The traffic data can include the traffic flow corresponding to each lane in the signal cycle and the signal cycle duration of the signal cycle. Figure 2 As shown in FIG. 1 , a schematic diagram of a process for obtaining flow data is shown, and the process may include:

[0057] Step 201: Perform license plate recognition based on the image to obtain original vehicle passing data.

[0058] For example, by analyzing the image, the license plate identification can be obtained, and the time point when the license plate identification passes each intersection can be determined. The relationship between the license plate identification and the time point when the license plate identification passes each intersection constitutes the original vehicle passing data. That is, the original vehicle passing data may include multiple groups of vehicle data, each group of vehicle data includes the relationship between a license plate identification and the time point when the license plate identification passes each intersection.

[0059] Step 202: Count the number of vehicles passing by according to the signal period to obtain the flow rate corresponding to each lane.

[0060] For example, the target intersection may include multiple lanes, and all license plate signs passing the target intersection can be found from the original vehicle passing data. Based on the time point when the license plate sign passes each intersection, the driving direction of the license plate sign can be known, such as the driving direction can be east-west, west-east, north-south, north-south, etc., and then the lane corresponding to the license plate sign can be known.

[0061] For each lane of the target intersection, based on the lanes corresponding to the license plate logos of all vehicles passing through the target intersection, the number of vehicles passing each lane can be counted. In this way, for each signal cycle, the number of vehicles passing each lane in the signal cycle can be counted. For example, the number of vehicles passing lane A in signal cycle T1, the number of vehicles passing lane B in signal cycle T1, the number of vehicles passing lane A in signal cycle T2, the number of vehicles passing lane B in signal cycle T2, and so on can be counted.

[0062] After obtaining the number of vehicles passing the lane in the signal cycle, the flow rate of the lane in the signal cycle can be determined based on the number of vehicles passing, that is, the number of vehicles in the lane in the signal cycle. For example, the flow rate of lane A in signal cycle T1, the flow rate of lane B in signal cycle T1, the flow rate of lane A in signal cycle T2, the flow rate of lane B in signal cycle T2, and so on.

[0063] For example, after obtaining the flow rate corresponding to each lane in the signal cycle, preprocessing operations such as outlier filtering and missing value filling can be performed, and there is no restriction on this preprocessing process. Among them, outlier filtering refers to removing abnormal flow rate, and missing value filling refers to filling the missing flow rate.

[0064] Step 203: Acquire traffic data at the signal cycle level. The traffic data may include the traffic volume corresponding to each lane in the signal cycle and the signal cycle duration of the signal cycle.

[0065] For example, assuming that the target intersection corresponds to lane A and lane B, the traffic data includes the traffic corresponding to lane A in signal cycle T1, the traffic corresponding to lane B in signal cycle T1, the signal cycle duration of signal cycle T1 (the signal cycle duration can be a parameter of the traffic light at the target intersection, and there is no restriction on this acquisition method), the traffic corresponding to lane A in signal cycle T2, the traffic corresponding to lane B in signal cycle T2, the signal cycle duration of signal cycle T2, and so on, to obtain traffic data for multiple signal cycles.

[0066] It should be noted that the total durations of different signal cycles may be the same or different. For example, the total duration of the signal cycle T1 may be the same as or different from the total duration of the signal cycle T2.

[0067] Second, based on the traffic data of multiple signal cycles, the target flow rate corresponding to the target signal cycle of each lane can be predicted, see Figure 3 FIG. 1 is a flow chart of predicting the target flow rate, which includes:

[0068] Step 301: Obtain traffic data of multiple historical signal cycles before the current signal cycle. The traffic data may include the traffic of each lane and the signal cycle duration of the historical signal cycles.

[0069] The current signal cycle is the signal cycle currently being executed at the target intersection. The signal cycle before the current signal cycle is called the historical signal cycle. For example, the K signal cycles before the current signal cycle are taken as the historical signal cycles. The K signal cycles can be continuous signal cycles or discontinuous signal cycles.

[0070] For each historical signal cycle, you can obtain the traffic data of the historical signal cycle. For more information, see Figure 2 As shown, which will not be repeated here, the traffic data includes the traffic volume corresponding to each lane in the historical signal cycle and the signal cycle duration corresponding to the historical signal cycle.

[0071] Exemplarily, the total durations of different historical signal cycles may be the same, or the total durations of different historical signal cycles may be different, that is, traffic data of historical signal cycles of different durations may be acquired.

[0072] Step 302: For each historical signal cycle, determine the historical flow rate of each lane in the historical signal cycle based on the traffic data of the historical signal cycle, that is, determine the historical flow rate of the lane in the historical signal cycle based on the traffic flow and signal cycle length of the lane in the historical signal cycle.

[0073] For example, the historical flow rate of lane A in historical signal cycle T1 is determined based on the traffic volume corresponding to lane A in historical signal cycle T1 and the signal cycle duration of historical signal cycle T1; the historical flow rate of lane B in historical signal cycle T1 is determined based on the traffic volume corresponding to lane B in historical signal cycle T1 and the signal cycle duration of historical signal cycle T1; the historical flow rate of lane A in historical signal cycle T2 is determined based on the traffic volume corresponding to lane A in historical signal cycle T2 and the signal cycle duration of historical signal cycle T2; the historical flow rate of lane B in historical signal cycle T2 is determined based on the traffic volume corresponding to lane B in historical signal cycle T2 and the signal cycle duration of historical signal cycle T2, and so on.

[0074] In a possible implementation, the historical flow rate may be determined using the following formula: i =Q i / G i Among them, vi It is used to represent the historical flow rate of the i-th lane in the historical signal cycle, Q i It is used to represent the flow rate of the i-th lane in the historical signal cycle, G i The signal cycle length used to represent the historical signal cycle. Of course, the above formula is only an example, and there is no limitation on the method of determining the historical flow rate.

[0075] In another possible implementation, since the lane flow rate at the signal cycle level has a large volatility, the accuracy of direct prediction is relatively low. Therefore, the average flow rate of the m historical signal cycles before the jth historical signal cycle (including the jth historical signal cycle itself) can be taken to represent the historical flow rate of the jth historical signal cycle. The flow data of the n adjacent historical signal cycles are used as input to calculate the historical flow rate of each lane in the historical signal cycle. For example, the following formula (1) can be used to determine the historical flow rate of the lane in the historical signal cycle. Of course, formula (1) is only an example.

[0076]

[0077] In formula (1), v i，j represents the historical flow rate of the i-th lane (the i-th lane can be any lane) in the j-th historical signal cycle, Q i，j represents the traffic volume of the i-th lane in the j-th historical signal cycle, Q i，j-m+1 +Q i，j-m+2 +…+Q i，j It represents the sum of the traffic in the m historical signal cycles before the jth historical signal cycle. j represents the signal cycle length of the jth signal cycle, C j-m+1 +C j-m+2 +…+C j represents the sum of the signal cycle lengths of the m historical signal cycles before the jth historical signal cycle. In summary, it can be seen that the historical flow rate v i，j It represents the average flow rate of m historical signal cycles before the jth historical signal cycle.

[0078] Step 303: predict the target flow rate of each lane in the target signal period based on the historical flow rates of each lane in multiple historical signal periods. The target flow rate is the flow rate prediction result.

[0079] Exemplarily, the signal period following the current signal period may be referred to as the target signal period, i.e., the signal period within a period of time in the future. The number of target signal periods may be one or multiple. Taking one target signal period as an example, the target signal period may be the Mth signal period following the current signal period. M may be 1, 2, 3, etc., and there is no restriction on the value of M.

[0080] After obtaining the historical flow rates corresponding to each lane in multiple historical signal cycles, the target flow rate corresponding to each lane in the target signal cycle can be predicted based on these historical flow rates. There is no restriction on the prediction method of the target flow rate, as long as the target flow rate can be predicted. For example, the target flow rate corresponding to the target signal cycle of lane A and the target flow rate corresponding to the target signal cycle of lane B can be predicted.

[0081] For example, all historical flow rates can be input into a network model (such as a neural network model, a deep learning network model, etc.), and the network model outputs the target flow rate corresponding to each lane in the target signal period. Of course, other algorithms can also be used to obtain the target flow rate corresponding to each lane in the target signal period.

[0082] Third, based on the traffic data of multiple signal cycles, the target flow rate corresponding to each lane in the target signal cycle can be predicted, and the target flow corresponding to each lane in the target signal cycle can be predicted, see Figure 4 FIG. 1 is a flow chart of predicting target flow rate and target flow rate, and the process may include:

[0083] Step 401: Obtain traffic data of multiple historical signal cycles before the current signal cycle. The traffic data may include the traffic of each lane and the signal cycle duration of the historical signal cycles.

[0084] Step 402: for each historical signal cycle, determine the historical flow rate of each lane in the historical signal cycle based on the traffic data of the historical signal cycle, that is, determine the historical flow rate of the lane in the historical signal cycle based on the traffic flow and signal cycle length of the lane in the historical signal cycle.

[0085] Step 403: predict the target flow rate of each lane in the target signal period based on the historical flow rates of each lane in multiple historical signal periods. The target flow rate is the flow rate prediction result.

[0086] Step 404: Obtain the signal cycle duration corresponding to the target signal cycle.

[0087] Exemplarily, the user may configure the signal cycle duration corresponding to the target signal cycle, indicating that the target signal cycle will be controlled using the configured signal cycle duration. Alternatively, a certain algorithm may be used to obtain the signal cycle duration corresponding to the target signal cycle, without limitation.

[0088] Step 405: For each lane, based on the target flow rate corresponding to the lane in the target signal period and the signal period duration of the target signal period, determine the target flow rate corresponding to the lane in the target signal period. The target flow rate is the flow prediction result, that is, the flow rate within the target signal period.

[0089] For example, based on the target flow rate corresponding to the target signal cycle of lane A (i.e., the flow rate prediction result) and the signal cycle duration of the target signal cycle, the target flow rate corresponding to the target signal cycle of lane A is determined; based on the target flow rate corresponding to the target signal cycle of lane B and the signal cycle duration of the target signal cycle, the target flow rate corresponding to the target signal cycle of lane B is determined, and so on.

[0090] In a possible implementation, the target flow rate can be determined using the following formula: Q i =Cv i Among them, Q i represents the target flow rate of the i-th lane in the target signal cycle, C represents the signal cycle length of the target signal cycle, and v i It represents the target flow rate corresponding to the target signal period of the i-th lane. Of course, the above formula is only an example, and there is no limitation on the method of determining the target flow rate.

[0091] In a possible implementation, the target flow rate of the lane corresponding to the target signal period can be determined based on the target flow rate of the lane corresponding to the target signal period, the signal period duration corresponding to the target signal period, and the acquired weight parameter matrix. The weight parameter matrix can be a parameter in the target prediction model, a pre-configured parameter, or a parameter obtained by other means, without limitation. For example, the target flow rate can be determined by the following formula: Q i =W i Cv i .W i Represents the weight parameter matrix corresponding to the i-th lane, the weight parameter matrix W i The addition of can improve the adaptability of traffic flow prediction for different lanes and further improve the prediction accuracy. i The target prediction model can be trained by obtaining the target prediction model and the parameters in the target prediction model can be used as the weight parameter matrix without any restriction.

[0092] Fourth, based on the flow data of multiple signal cycles, the flow data of multiple signal cycles are input into the target prediction model to obtain the target flow rate corresponding to each lane in the target signal cycle, that is, the target prediction model predicts the target flow rate corresponding to each lane in the target signal cycle. The structure of the target prediction model, the training process of the target prediction model, and the target flow rate prediction process based on the target prediction model are described below.

[0093] See also Figure 5A As shown, it is a structural diagram of the target prediction model. The target prediction model may include a conversion layer and at least one prediction layer. Taking n prediction layers as an example, the target prediction model may be a prediction model based on a neural network, or a prediction model based on deep learning, such as a timing model or a timing model, etc. For example, it may be a prediction model based on LSTM (Long Short Term Memory), a prediction model based on GRU (Gate Recurrent Unit), a prediction model based on TCN, a prediction model based on STGCN, a prediction model based on ASTGCN, a prediction model based on STAWnet, etc. There is no restriction on the type of the target prediction model.

[0094] The input data of the target prediction model is the flow data of multiple historical signal cycles, which includes the flow of each lane and the signal cycle duration of the historical signal cycle. The total duration of different historical signal cycles can be the same or different, that is, the total duration of the historical signal cycle can be non-equal intervals. The output data of the target prediction model is the target flow rate corresponding to each lane in the target signal cycle, that is, the flow rate prediction result.

[0095] For the conversion layer, the conversion layer is used to convert the traffic data into the historical flow rate corresponding to each lane in the historical signal cycle. For example, the historical flow rate of each lane is obtained by the traffic volume of each lane in the historical signal cycle and the signal cycle duration corresponding to the historical signal cycle. Since the lane flow rate at the signal cycle level has a large volatility, the accuracy of directly using it for prediction is relatively low. Therefore, the average flow rate of the m historical signal cycles before the jth historical signal cycle (including the jth historical signal cycle itself) can be taken to represent the historical flow rate of the jth historical signal cycle. The traffic data of the adjacent n historical signal cycles are used as the input of the prediction layer, so that the prediction layer calculates the historical flow rate of each lane corresponding to the historical signal cycle. The calculation formula of the prediction layer can refer to the above formula (1), which will not be repeated here.

[0096] For n prediction layers, the n prediction layers can include multiple layers of neural network hidden layers, which are used to predict the target flow rate corresponding to the target signal period of each lane, that is, the flow rate prediction result. For example, the n prediction layers can perform convolution operations, activation operations, full connection operations, etc., and finally predict the target flow rate. For example, the prediction layer can be LSTM, GRU, TCN in the time series model, or STGCN, ASTGCN, STAWnet in the spatiotemporal series model, and there is no restriction on the structure of this prediction layer.

[0097] See also Figure 5B FIG. 1 is a schematic diagram of a training process of a target prediction model, which may include:

[0098] Step 511: Obtain a training set, which may include sample data and label data corresponding to the sample data, and the number of sample data may be multiple. For each sample data, the sample data includes first flow data of multiple sample signal cycles, the first flow data includes the flow and signal cycle duration of each lane, and the label data corresponding to the sample data includes second flow data of future signal cycles of multiple sample signal cycles, and the second flow data includes the actual flow rate of each lane.

[0099] For example, a certain signal cycle can be used as a reference signal cycle (i.e., the current signal cycle), the K signal cycles before the reference signal cycle can be used as K sample signal cycles (i.e., historical signal cycles), and the Mth signal cycle after the reference signal cycle can be used as a future signal cycle (i.e., the target signal cycle).

[0100] See also Figure 2 As shown, the flow data of each signal cycle can be obtained, that is, the first flow data of each sample signal cycle and the second flow data of the future signal cycle can be obtained. For each sample signal cycle, the first flow data of the sample signal cycle includes the flow of each lane in the sample signal cycle and the signal cycle length of the sample signal cycle. For the future signal cycle, the second flow data of the future signal cycle includes the real flow rate of each lane in the future signal cycle. Among them, since the flow of each lane in the future signal cycle can be known, the flow can be converted into the real flow rate.

[0101] Step 512: Obtain an initial prediction model. The initial prediction model may include a conversion layer and at least one prediction layer. For example, n prediction layers are used. Figure 5A As shown, the initial prediction model may be a pre-configured model or a model obtained by other methods, and there is no limitation on this.

[0102] Step 513: Input the sample data in the training set to the initial prediction model, so that the initial prediction model determines the sample flow rate corresponding to each lane in the future signal cycle based on the sample data.

[0103] Exemplarily, after the sample data (such as multiple sample data) in the training set is input into the initial prediction model, the conversion layer of the initial prediction model can convert the first flow data into the flow rate corresponding to each lane in the sample signal period, for example, the flow rate is converted by the above formula (1). Then, the flow rate corresponding to each lane in the sample signal period can be input into the prediction layer, and the sample flow rate corresponding to each lane in the future signal period is predicted by n prediction layers, and there is no restriction on this prediction process.

[0104] Step 514: Calculate the loss value based on the difference between the sample flow rate and the true flow rate, and adjust the network parameters of the initial prediction model based on the loss value to obtain an adjusted prediction model.

[0105] Assuming that the target intersection corresponds to lane A and lane B, the loss value can be calculated based on the difference between the sample flow rate of lane A in the future signal cycle and the actual flow rate of lane A in the future signal cycle, and the difference between the sample flow rate of lane B in the future signal cycle and the actual flow rate of lane B in the future signal cycle.

[0106] For example, a loss function can be designed in advance, the input of the loss function is the difference between the sample flow rate and the actual flow rate, and the output of the loss function is the loss value. Therefore, after substituting the difference between the sample flow rate and the actual flow rate (such as the difference between the sample flow rate and the actual flow rate of lane A, the difference between the sample flow rate and the actual flow rate of lane B) into the loss function, the loss value can be obtained, and there is no restriction on this loss function.

[0107] After obtaining the loss value, the network parameters of the initial prediction model (such as the network parameters of the conversion layer and the network parameters of each prediction layer) can be adjusted based on the loss value, such as by using the gradient descent method to adjust the network parameters. There is no restriction on this adjustment method, and the adjusted prediction model is obtained.

[0108] Step 515: Determine the target prediction model based on the adjusted prediction model. For example, if the adjusted prediction model has converged, the adjusted prediction model can be used as the target prediction model. If the adjusted prediction model has not converged, the adjusted prediction model can be used as the initial prediction model, and the process returns to step 513, and so on, until the adjusted prediction model has converged and the adjusted prediction model is used as the target prediction model.

[0109] Exemplarily, a test set may be obtained in the same manner as the training set, and the test set includes sample data and label data corresponding to the sample data. For each sample data, the sample data includes first flow data of multiple sample signal cycles, the first flow data includes the flow rate and signal cycle duration of each lane, and the label data corresponding to the sample data includes second flow data of future signal cycles of multiple sample signal cycles, and the second flow data includes the actual flow rate of each lane.

[0110] The prediction accuracy corresponding to the adjusted prediction model can be determined based on the test set. If the prediction accuracy corresponding to the adjusted prediction model is greater than the convergence threshold (the convergence threshold can be configured based on experience, such as 95%, 98%, etc.), it can be determined that the adjusted prediction model has converged. If the prediction accuracy corresponding to the adjusted prediction model is not greater than the convergence threshold, it can be determined that the adjusted prediction model has not converged. For example, after the sample data is input into the adjusted prediction model, if the prediction result is the same as the label data, the number of correct predictions is increased by 1, and if the prediction result is different from the label data, the number of wrong predictions is increased by 1. In this way, the number of correct predictions p1 and the number of wrong predictions p2 can be counted, and the prediction accuracy corresponding to the adjusted prediction model can be determined based on the number p1 and the number p2, such as the prediction accuracy can be p1 / (p1+p2).

[0111] In another possible implementation, if the number of iterations of the adjusted prediction model is greater than a preset number threshold (which can be configured based on experience), it is determined that the adjusted prediction model has converged. If the number of iterations of the adjusted prediction model is not greater than the preset number threshold, it is determined that the adjusted prediction model has not converged.

[0112] In another possible implementation, if the iteration duration of the adjusted prediction model is greater than a preset duration threshold (which can be configured based on experience), it is determined that the adjusted prediction model has converged. If the iteration duration of the adjusted prediction model is not greater than the preset duration threshold, it is determined that the adjusted prediction model has not converged.

[0113] Of course, the above are just a few examples, and this embodiment does not limit the convergence condition.

[0114] See also Figure 5C As shown, it is a schematic diagram of the target flow rate prediction process based on the target prediction model.

[0115] Step 521: Obtain traffic data of multiple historical signal cycles before the current signal cycle. The traffic data may include the traffic of each lane and the signal cycle duration corresponding to the historical signal cycle.

[0116] Exemplarily, the total durations of different historical signal cycles may be the same, or the total durations of different historical signal cycles may be different, that is, traffic data of historical signal cycles of different durations may be acquired.

[0117] Step 522: input the traffic data of multiple historical signal cycles to the conversion layer of the target prediction model. For each historical signal cycle, the historical flow rate of each lane in the historical signal cycle is determined based on the traffic data of the historical signal cycle through the conversion layer, that is, the historical flow rate of the lane in the historical signal cycle is determined based on the traffic flow of the lane in the historical signal cycle and the signal cycle length of the historical signal cycle.

[0118] For example, the conversion layer can use the above formula (1) to determine the historical flow rate, which will not be described in detail here.

[0119] Step 523: input the historical flow rates corresponding to each lane in multiple historical signal cycles to the prediction layer, and use the prediction layer to predict the target flow rate corresponding to each lane in the target signal cycle based on the historical flow rates.

[0120] For example, after obtaining the historical flow rates corresponding to each lane in multiple historical signal periods, the prediction layer can predict the target flow rate corresponding to each lane in the target signal period based on these historical flow rates. There is no restriction on the prediction method of this target flow rate, as long as the target flow rate can be predicted.

[0121] Fifth, the traffic data of multiple historical signal cycles and the signal cycle duration corresponding to the target signal cycle are input into the target prediction model. The target prediction model predicts the target flow rate and target flow rate corresponding to the target signal cycle for each lane. The structure of the target prediction model, the training process of the target prediction model, and the target flow rate and target flow rate prediction process based on the target prediction model are explained below.

[0122] See also Fig. 6A As shown, it is a schematic diagram of the structure of the target prediction model. The target prediction model may include a conversion layer, at least one prediction layer and an inverse conversion layer, taking n prediction layers as an example. Among them, the input data of the target prediction model are the flow data of multiple historical signal cycles and the signal cycle duration corresponding to the target signal cycle. The total duration of different historical signal cycles can be the same or different. The output data of the target prediction model can be the target flow rate corresponding to the target signal cycle for each lane (i.e., the flow rate prediction result), and the target flow corresponding to the target signal cycle for each lane (i.e., the flow prediction result).

[0123] For the conversion layer, the conversion layer is used to convert the traffic data into the historical flow rate corresponding to each lane in the historical signal cycle. For the n prediction layers, it is used to predict the target flow rate corresponding to each lane in the target signal cycle, that is, the flow rate prediction result. For the inverse conversion layer, the inverse conversion layer is used to calculate the target flow rate v corresponding to each lane in the target signal cycle. i (i.e., the flow rate prediction result of the prediction layer) and the signal cycle duration C (i.e., the input data) corresponding to the target signal cycle, combined with the weight parameter matrix W corresponding to each lane i , we can get the target flow Q corresponding to each lane in the target signal cycle i (i.e., the flow prediction result of the inverse conversion layer). The addition of the weight parameter matrix can improve the model's adaptability to the flow prediction of different lanes and further improve the prediction accuracy of the model. The specific formula can be shown as follows: i =W i Cv i .

[0124] See also Figure 6B FIG. 1 is a schematic diagram of a training process of a target prediction model, which may include:

[0125] Step 611: Acquire a training set, which may include sample data and label data corresponding to the sample data, and the number of sample data may be multiple. For each sample data, the sample data may include first flow data of multiple sample signal cycles and signal cycle duration of future signal cycles, and the first flow data may include the flow of each lane in the sample signal cycle and the signal cycle duration of the sample signal cycle. The label data corresponding to the sample data may include second flow data of the future signal cycle, and the second flow data may include the actual flow rate of each lane and the actual flow of each lane.

[0126] Step 612: Obtain an initial prediction model. The initial prediction model may include a conversion layer, at least one prediction layer, and an inverse conversion layer. Fig. 6A As shown, n prediction layers are taken as an example for explanation.

[0127] Step 613: Input the sample data in the training set to the initial prediction model, so that the initial prediction model determines the sample flow rate corresponding to each lane in the future signal cycle based on the sample data, and the initial prediction model determines the sample flow rate corresponding to each lane in the future signal cycle based on the sample data.

[0128] Exemplarily, after the sample data in the training set is input to the initial prediction model, the conversion layer of the initial prediction model can convert the first traffic data into the flow rate corresponding to each lane in the sample signal cycle. Then, the flow rate corresponding to each lane in the sample signal cycle can be input to the prediction layer, and the sample flow rate corresponding to each lane in the future signal cycle can be predicted through n prediction layers. Then, the sample flow rate corresponding to each lane in the future signal cycle and the signal cycle length of the future signal cycle can be input to the inverse conversion layer, so that the inverse conversion layer can obtain the target flow rate v corresponding to each lane in the future signal cycle. i and the signal cycle duration C of the future signal cycle, combined with the weight parameter matrix W corresponding to each lane i , get the sample flow Q corresponding to each lane in the future signal cycle i , such as Q i =W i Cv i .

[0129] Step 614, based on the difference between the sample flow rate and the actual flow rate, and the difference between the sample flow and the actual flow, the loss value is calculated, and the network parameters of the initial prediction model are adjusted based on the loss value to obtain the adjusted prediction model. For example, a loss function can be designed in advance. Assuming that the target intersection corresponds to lane A and lane B, the input of the loss function can be the difference between the sample flow rate of lane A in the future signal cycle and the actual flow rate of lane A in the future signal cycle, the difference between the sample flow of lane A in the future signal cycle and the actual flow of lane A in the future signal cycle, the difference between the sample flow of lane B in the future signal cycle and the actual flow of lane B in the future signal cycle, and the difference between the sample flow of lane B in the future signal cycle and the actual flow of lane B in the future signal cycle. The output of the loss function is the loss value. Obviously, after substituting the above differences into the loss function, the loss value can be obtained.

[0130] Step 615: Determine the target prediction model based on the adjusted prediction model. For example, if the adjusted prediction model has converged, the adjusted prediction model can be used as the target prediction model. If the adjusted prediction model has not converged, the adjusted prediction model can be used as the initial prediction model, and the process returns to step 613, and so on, until the adjusted prediction model has converged and the adjusted prediction model is used as the target prediction model.

[0131] See also Figure 6C FIG. 1 is a schematic diagram of a target flow rate prediction process and a target flow rate prediction process based on a target prediction model. The target flow rate prediction process and the target flow rate prediction process may include:

[0132] Step 621: Obtain traffic data of multiple historical signal cycles before the current signal cycle. The traffic data may include the traffic of each lane and the signal cycle duration of the historical signal cycles.

[0133] Exemplarily, the total durations of different historical signal cycles may be the same, or the total durations of different historical signal cycles may be different, that is, traffic data of historical signal cycles of different durations may be acquired.

[0134] Step 622: Obtain the signal cycle duration corresponding to the target signal cycle.

[0135] Step 623: Input the traffic data of multiple historical signal cycles to the conversion layer of the target prediction model. For each historical signal cycle, determine the historical flow rate of each lane in the historical signal cycle based on the traffic data of the historical signal cycle through the conversion layer, that is, determine the historical flow rate of the lane in the historical signal cycle based on the traffic flow of the lane in the historical signal cycle and the signal cycle length of the historical signal cycle.

[0136] Step 624: input the historical flow rates corresponding to each lane in multiple historical signal cycles to the prediction layer, and use the prediction layer to predict the target flow rate corresponding to each lane in the target signal cycle based on the historical flow rates.

[0137] Step 625: Input the target flow rate corresponding to the target signal period of each lane and the signal cycle duration corresponding to the target signal period to the inverse conversion layer. The inverse conversion layer determines the target flow rate corresponding to the target signal period of each lane based on the signal cycle duration, the target flow rate and the weight parameters in the inverse conversion layer.

[0138] For example, the inverse conversion layer can be configured by the target flow rate v corresponding to each lane in the target signal period. i The signal cycle duration C corresponding to the target signal cycle, combined with the weight parameter matrix W corresponding to each lane i , and obtain the target flow Q corresponding to each lane in the target signal cycle i For example, the inverse conversion layer uses the following formula to determine the target flow: Q i =W i Cv i , W i Represents the weight parameter matrix corresponding to the i-th lane, the weight parameter matrix W i The addition of can improve the adaptability of traffic flow prediction for different lanes and improve the prediction accuracy.

[0139] It can be seen from the above technical solutions that in the embodiment of the present application, the target flow rate corresponding to the target signal cycle of each lane can be predicted based on the flow data of multiple historical signal cycles, and the flow rate in the future period can be predicted, the prediction accuracy can be improved, the prediction precision can be improved, and accurate and reliable flow rate can be obtained. By introducing the traffic volume of each lane at the signal cycle level and the signal cycle duration, any equidistant flow prediction model can be converted into a non-equidistant prediction model, and the flow prediction can be converted into a flow rate prediction, thereby realizing the flow rate prediction at the signal cycle level. Based on the flow rate prediction result, the flow prediction result can also be obtained, thereby realizing the flow prediction at the signal cycle level, that is, a signal cycle level urban road network lane flow prediction method is proposed, which can obtain the prediction result at the signal cycle level and realize the flow rate prediction and flow prediction at the signal cycle level. By adding a conversion layer and an inverse conversion layer, any equidistant flow prediction model can be converted into a non-equidistant prediction model, and the prediction result of the signal cycle dimension can be directly output. By introducing a weight parameter matrix in the inverse conversion layer, the adaptability of the model to the flow prediction of different lanes can be improved, and the prediction precision is further improved.

[0140] By introducing the traffic volume and signal cycle duration of each lane at the signal cycle level, the original prediction of traffic volume is converted into a prediction of flow rate, thereby realizing the prediction of lane volume at the signal cycle level, solving the problem of precision loss of the signal control algorithm caused by equal interval prediction, and reducing the prediction lag problem when traffic volume changes suddenly during peak hours. By taking the average lane flow rate of the previous several signal cycles as the lane flow rate of the current signal cycle, the volatility of the data is reduced and the accuracy of the prediction algorithm is improved.

[0141] Based on the same application concept as the above method, a flow rate prediction device is proposed in the embodiment of the present application, see Figure 7 FIG. 1 is a schematic diagram of the structure of the flow rate prediction device, which may include:

[0142] An acquisition module 71 is used to acquire the flow data of a plurality of historical signal cycles of the target intersection before the current signal cycle; wherein the flow data includes the flow of each lane and the signal cycle duration of the historical signal cycle; wherein the signal cycle durations of different historical signal cycles are the same or different;

[0143] A processing module 72 is used to determine, for each historical signal cycle, a historical flow rate corresponding to each lane in the historical signal cycle based on the flow data of the historical signal cycle; wherein the historical flow rate corresponding to each lane in the historical signal cycle is determined based on the flow rate of the lane and the length of the signal cycle;

[0144] A target flow rate corresponding to each lane in a target signal cycle is predicted based on the historical flow rates corresponding to each lane in the plurality of historical signal cycles, and the target signal cycle is located after the current signal cycle.

[0145] Exemplarily, the acquisition module 71 is also used to obtain the signal cycle duration corresponding to the target signal cycle; the processing module 72 is also used to determine, for each lane, the target flow rate corresponding to the lane in the target signal cycle and the signal cycle duration of the target signal cycle.

[0146] Exemplarily, the processing module 72 determines the target flow rate corresponding to the target signal cycle of the lane based on the target flow rate corresponding to the target signal cycle and the signal cycle duration of the target signal cycle, and is specifically used to: i , the signal cycle duration C corresponding to the target signal cycle, and the acquired weight parameter matrix W i , the target flow rate Q corresponding to the lane in the target signal cycle is determined by the following formula i :Q i =W i Cv i .

[0147] Exemplarily, when the processing module 72 determines the historical flow rate corresponding to each lane in the historical signal cycle based on the flow data of the historical signal cycle, it is specifically used to: determine the historical flow rate corresponding to the lane in the historical signal cycle using the following formula: Among them, v i，j represents the historical flow rate of the i-th lane in the j-th historical signal cycle, Q i，j represents the traffic volume of the i-th lane in the j-th historical signal cycle, C j Indicates the signal cycle length of the j-th signal cycle.

[0148] Exemplarily, the weight parameter matrix is ​​a parameter in a target prediction model, and the target prediction model includes a conversion layer and a prediction layer; the processing module 72 is specifically used to: input the flow data of the multiple historical signal periods to the conversion layer, and determine the historical flow rate corresponding to each lane in the multiple historical signal periods based on the flow data through the conversion layer; input the historical flow rate corresponding to each lane in the multiple historical signal periods to the prediction layer, and predict the target flow rate corresponding to each lane in the target signal period based on the historical flow rate through the prediction layer.

[0149] Exemplarily, if the target prediction model also includes an inverse conversion layer, the processing module 72 is specifically used to: input the target flow rate corresponding to each lane in the target signal period and the signal cycle duration of the target signal period to the inverse conversion layer; determine the target flow rate corresponding to each lane in the target signal period through the inverse conversion layer based on the signal cycle duration, the target flow rate and the weight parameters in the inverse conversion layer.

[0150] Exemplarily, the processing module 72 is also used to train the target prediction model using the following steps: obtaining sample data and label data, the sample data including first flow data of multiple sample signal cycles, the first flow data including the flow and signal cycle duration of each lane; the label data including second flow data of future signal cycles of the multiple sample signal cycles, the second flow data including the actual flow rate of each lane and the actual flow of each lane; inputting the sample data into the initial prediction model, so that the initial prediction model determines the sample flow rate corresponding to each lane in the future signal cycle based on the sample data, and determines the sample flow corresponding to each lane in the future signal cycle based on the sample data; calculating the loss value based on the difference between the sample flow rate and the actual flow rate, and the difference between the sample flow and the actual flow, adjusting the network parameters of the initial prediction model based on the loss value, and determining the target prediction model based on the adjusted prediction model.

[0151] Based on the same application concept as the above method, an electronic device is proposed in the embodiment of the present application, see Figure 8 As shown, the electronic device includes: a processor 81 and a machine-readable storage medium 82, wherein the machine-readable storage medium 82 stores machine-executable instructions that can be executed by the processor 81; the processor 81 is used to execute the machine-executable instructions to implement the flow rate prediction method disclosed in the above example of this application.

[0152] Based on the same application concept as the above method, an embodiment of the present application also provides a machine-readable storage medium, on which a number of computer instructions are stored. When the computer instructions are executed by a processor, the flow rate prediction method disclosed in the above example of the present application can be implemented.

[0153] The above-mentioned machine-readable storage medium may be any electronic, magnetic, optical or other physical storage device, which may contain or store information, such as executable instructions, data, etc. For example, the machine-readable storage medium may be: RAM (Radom Access Memory), volatile memory, non-volatile memory, flash memory, storage drive (such as hard disk drive), solid state drive, any type of storage disk (such as CD, DVD, etc.), or similar storage medium, or a combination thereof.

[0154] The systems, devices, modules or units described in the above embodiments may be implemented by a computer entity or by a product having a certain function. A typical implementation device is a computer, and the specific form of the computer may be a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email transceiver device, a game console, a tablet computer, a wearable device or a combination of any of these devices.

[0155] For the convenience of description, the above device is described in various units according to their functions. Of course, when implementing the present application, the functions of each unit can be implemented in the same or multiple software and / or hardware.

[0156] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the embodiments of the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0157] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0158] Moreover, these computer program instructions can also be stored in a computer-readable memory that can guide a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory produce a product including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0159] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for executing on the computer or other programmable device to implement the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0160] The above is only an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. A flow rate prediction method, characterized in that: The method comprises: Obtaining traffic data of multiple historical signal cycles before the current signal cycle at the target intersection; wherein the traffic data includes the traffic of each lane and the signal cycle duration of the historical signal cycle; wherein the signal cycle durations of different historical signal cycles are the same or different; For each historical signal cycle, determine the historical flow rate of each lane corresponding to the historical signal cycle based on the traffic data of the historical signal cycle; wherein the historical flow rate of each lane corresponding to the historical signal cycle is determined based on the traffic flow of the lane and the length of the signal cycle; A target flow rate corresponding to each lane in a target signal cycle is predicted based on the historical flow rates corresponding to each lane in the plurality of historical signal cycles, and the target signal cycle is located after the current signal cycle.

2. The method according to claim 1, characterized in that The method further comprises: Obtaining a signal cycle duration corresponding to the target signal cycle; For each lane, a target flow rate corresponding to the lane in the target signal period is determined based on a target flow rate corresponding to the lane in the target signal period and a signal period duration of the target signal period.

3. The method according to claim 2, characterized in that Determining the target flow rate of the lane corresponding to the target signal cycle and the signal cycle duration of the target signal cycle comprises: Based on the target flow rate v corresponding to the lane in the target signal period i , the signal cycle duration C corresponding to the target signal cycle, and the acquired weight parameter matrix W i , the target flow rate Q corresponding to the lane in the target signal cycle is determined by the following formula i :Q i =W i Cv i .

4. The method according to claim 1, characterized in that: The determining of the historical flow rate of each lane corresponding to the historical signal period based on the flow data of the historical signal period includes: The following formula is used to determine the historical flow rate of the lane corresponding to the historical signal cycle: Among them, v i，j represents the historical flow rate of the i-th lane in the j-th historical signal cycle, Q i，j represents the traffic volume of the i-th lane in the j-th historical signal cycle, C j Indicates the signal cycle length of the j-th signal cycle.

5. The method according to any one of claims 1 to 4, characterized in that: The weight parameter matrix is ​​a parameter in a target prediction model, and the target prediction model includes a conversion layer and a prediction layer; Inputting the flow data of the multiple historical signal cycles into the conversion layer, and determining the historical flow rate of each lane corresponding to the multiple historical signal cycles based on the flow data through the conversion layer; The historical flow rates corresponding to each lane in the multiple historical signal cycles are input to the prediction layer, and the prediction layer predicts the target flow rate corresponding to each lane in the target signal cycle based on the historical flow rates.

6. The method according to claim 5, characterized in that If the target prediction model further includes an inverse conversion layer, the target flow rate corresponding to the target signal cycle of each lane and the signal cycle duration of the target signal cycle are input into the inverse conversion layer; The target flow rate of each lane corresponding to the target signal cycle is determined by the inverse conversion layer based on the signal cycle duration, the target flow rate and the weight parameter in the inverse conversion layer.

7. The method according to claim 6, characterized in that The method further comprises: training the target prediction model by the following steps: Acquire sample data and label data, wherein the sample data includes first flow data of a plurality of sample signal cycles, wherein the first flow data includes the flow rate and signal cycle duration of each lane; the label data includes second flow data of a future signal cycle of the plurality of sample signal cycles, wherein the second flow data includes a real flow rate of each lane and a real flow rate of each lane; Inputting the sample data into an initial prediction model, so that the initial prediction model determines the sample flow rate corresponding to each lane in the future signal cycle based on the sample data, and determines the sample flow rate corresponding to each lane in the future signal cycle based on the sample data; Based on the difference between the sample flow rate and the true flow rate, and the difference between the sample flow and the true flow, a loss value is calculated, the network parameters of the initial prediction model are adjusted based on the loss value, and the target prediction model is determined based on the adjusted prediction model.

8. A flow rate prediction device, characterized in that: The device comprises: An acquisition module, used to acquire the flow data of a plurality of historical signal cycles before the current signal cycle of the target intersection; wherein the flow data includes the flow of each lane and the signal cycle duration of the historical signal cycle; wherein the signal cycle durations of different historical signal cycles are the same or different; A processing module, for determining, for each historical signal cycle, a historical flow rate corresponding to each lane in the historical signal cycle based on the flow data of the historical signal cycle; wherein the historical flow rate corresponding to each lane in the historical signal cycle is determined based on the flow rate of the lane and the length of the signal cycle; A target flow rate corresponding to each lane in a target signal cycle is predicted based on the historical flow rates corresponding to each lane in the plurality of historical signal cycles, and the target signal cycle is located after the current signal cycle.

9. The device according to claim 8, characterized in that in, The acquisition module is further used to acquire the signal cycle duration corresponding to the target signal cycle; the processing module is further used to determine, for each lane, a target flow rate corresponding to the target signal cycle of the lane and the signal cycle duration of the target signal cycle; Wherein, the processing module determines the target flow rate corresponding to the target signal cycle of the lane based on the target flow rate corresponding to the target signal cycle and the signal cycle duration of the target signal cycle, and is specifically used to: i , the signal cycle duration C corresponding to the target signal cycle, and the acquired weight parameter matrix W i , the target flow rate Q corresponding to the lane in the target signal cycle is determined by the following formula i :Q i =W i Cv i ; Wherein, when the processing module determines the historical flow rate corresponding to each lane in the historical signal cycle based on the flow data of the historical signal cycle, it is specifically used to: determine the historical flow rate corresponding to the lane in the historical signal cycle using the following formula: Among them, v i，j represents the historical flow rate of the i-th lane in the j-th historical signal cycle, Q i，j represents the traffic volume of the i-th lane in the j-th historical signal cycle, C j represents the signal cycle length of the jth signal cycle; The weight parameter matrix is ​​a parameter in the target prediction model, and the target prediction model includes a conversion layer and a prediction layer; the processing module is specifically used to: input the flow data of the multiple historical signal cycles to the conversion layer, and determine the historical flow rate corresponding to each lane in the multiple historical signal cycles based on the flow data through the conversion layer; input the historical flow rate corresponding to each lane in the multiple historical signal cycles to the prediction layer, and predict the target flow rate corresponding to each lane in the target signal cycle based on the historical flow rate through the prediction layer; Wherein, if the target prediction model further includes an inverse conversion layer, the processing module is specifically used to: input the target flow rate corresponding to the target signal period of each lane and the signal cycle duration of the target signal period to the inverse conversion layer; determine the target flow rate corresponding to the target signal period of each lane through the inverse conversion layer based on the signal cycle duration, the target flow rate and the weight parameter in the inverse conversion layer; Wherein, the processing module is also used to train the target prediction model by adopting the following steps: obtaining sample data and label data, the sample data includes first flow data of multiple sample signal cycles, the first flow data includes the flow and signal cycle duration of each lane; the label data includes second flow data of future signal cycles of the multiple sample signal cycles, the second flow data includes the actual flow rate of each lane and the actual flow of each lane; the sample data is input into the initial prediction model, so that the initial prediction model determines the sample flow rate corresponding to each lane in the future signal cycle based on the sample data, and determines the sample flow corresponding to each lane in the future signal cycle based on the sample data; based on the difference between the sample flow rate and the actual flow rate, and the difference between the sample flow and the actual flow, calculate the loss value, adjust the network parameters of the initial prediction model based on the loss value, and determine the target prediction model based on the adjusted prediction model.

10. An electronic device, characterized in that: include: a processor and a machine-readable storage medium storing machine-executable instructions executable by the processor; The processor is used to execute machine executable instructions to implement the method described in any one of claims 1-7.

Images