CN112489419B

CN112489419B - Method and device for determining road capacity and storage medium

Info

Publication number: CN112489419B
Application number: CN202011174360.5A
Authority: CN
Inventors: 王萌; 唐欣; 熊福祥; 刘志远
Original assignee: Huawei Technologies Co Ltd
Current assignee: Huawei Technologies Co Ltd
Priority date: 2020-10-28
Filing date: 2020-10-28
Publication date: 2022-04-26
Anticipated expiration: 2040-10-28
Also published as: CN114882696A; CN114882696B; CN112489419A

Abstract

The embodiment of the application discloses a method and a device for determining road capacity and a storage medium, and belongs to the field of intelligent traffic. In the embodiment of the application, when the road capacity of the first time period is determined, the traffic supply attribute data and the traffic demand attribute data in the first time period are acquired, on one hand, if the acquired data is real-time data, the scheme can determine the road capacity in real time, namely, the real-time performance is higher, and on the other hand, if the acquired data is traffic data of a certain region and a certain time period, the scheme can determine the road capacity of the region in the time period, namely, the scheme can adapt to differences of different regions in different time periods. In addition, the data acquired by the scheme does not depend on manual measurement and historical experience, the accuracy of the determined road capacity is higher, and the labor cost is very low.

Description

Method and device for determining road capacity and storage medium

Technical Field

The embodiment of the application relates to the field of intelligent traffic, in particular to a method and a device for determining road capacity and a storage medium.

Background

The road capacity is also called road traffic capacity or road bearing capacity, and refers to the maximum number of vehicles passing through a certain point or a certain cross section on a road in unit time under the requirements of normal roads, traffic, control, running quality and the like. The road capacity reflects the maximum capacity of the road for bearing vehicle operation, and provides important basis for intelligent transportation, Intelligent Transportation System (ITS) and the like.

The related art refers to a calculation formula in a highway capacity manual (HMC) to calculate the road capacity, and this method is a commonly used method at present. In the method, firstly, the basic traffic capacity of a road is calculated through a calculation formula according to traffic signal timing and saturated headway on the road, and because the intersection in the road is considered to be the maximum bottleneck restricting the road capacity in the HMC, factors such as lane composition at the intersection, traffic flow proportion and the like restrict the road capacity, the reduction coefficient given in a manual is selected according to the factors, and the calculated basic traffic capacity is reduced through the reduction coefficient to obtain the road capacity. The traffic signal timing comprises a straight-line green signal ratio, a left-turning green signal ratio and a signal period, and the saturated head time distance can be obtained by dividing the shortest head distance of two front and back continuous vehicles on a road by the rear vehicle speed.

However, in the method for determining the road capacity according to the manual, the acquisition of the saturated headway generally depends on manual measurement or historical experience, but the manual measurement is high in cost and low in real-time performance, and the accuracy of the saturated headway determined according to the historical experience is low. In addition, the reduction coefficient given in the manual is difficult to adapt to the difference of different areas and different time periods. Therefore, a solution is needed to accurately acquire the road capacity in real time.

Disclosure of Invention

The embodiment of the application provides a method and a device for determining road capacity and a storage medium, the accuracy of the determined road capacity can be achieved, the method and the device can adapt to differences in different regions and different time periods, and the manual measurement cost can be reduced. The technical scheme is as follows:

in a first aspect, a method for determining road capacity is provided, the method comprising:

acquiring first traffic supply attribute data and first traffic demand attribute data, wherein the first traffic supply attribute data are used for representing road section attributes in a first time period, and the first traffic demand attribute data are used for representing traffic flow attributes in the first time period; and determining the road capacity in the first time period through a mapping relation between the road capacity and the traffic supply and demand attribute according to the first traffic supply attribute data and the first traffic demand attribute data, wherein the mapping relation is determined according to the traffic supply attribute data, the traffic demand attribute data and the road flow acquired by history, and the traffic supply and demand attribute comprises a traffic supply attribute and a traffic demand attribute.

In the embodiment of the application, when the road capacity of the first time period is determined, the traffic supply attribute data and the traffic demand attribute data in the first time period are acquired, on one hand, if the acquired data is real-time data, the scheme can determine the road capacity in real time, namely, the real-time performance is higher, and on the other hand, if the acquired data is traffic data of a certain region and a certain time period, the scheme can determine the road capacity of the region in the time period, namely, the scheme can adapt to differences of different regions in different time periods. In addition, the data acquired by the scheme does not depend on manual measurement and historical experience, the accuracy of the determined road capacity is higher, and the labor cost is very low.

Optionally, the embodiment of the present application is introduced by taking an example that the method is applied to a traffic center server included in a traffic center side.

Optionally, the obtaining, by the traffic center server, the first traffic demand attribute data includes: receiving traffic flow data sent by a data acquisition side and/or an edge side in a first time period, wherein the data acquisition side is used for acquiring traffic flow data in real time, and the edge side is used for collecting the traffic flow data acquired by the data acquisition side in real time and preprocessing the traffic flow data; and determining first traffic demand attribute data according to the received traffic flow data.

That is, the traffic center server can directly acquire traffic flow data from the data acquisition side, and can also acquire traffic flow data from the edge side. In other words, the data acquisition side can directly report the acquired traffic flow data to the traffic center server, and the data acquisition side can also firstly gather the acquired traffic flow data to the edge side, and report the data to the traffic center server after the data is preprocessed by the edge side.

Optionally, the data acquisition side comprises one or more traffic sensors, the edge side comprises one or more edge devices, and each of the one or more traffic sensors is any one of a sensor of an electric police station, a sensor of a geomagnetic coil/radar, a Global Navigation Satellite System (GNSS) sensor in the vehicle, and a positioning sensor of a mobile device in the vehicle.

Optionally, the traffic center server obtains first traffic supply attribute data, including: acquiring first traffic supply attribute data from the stored road attribute data; and/or acquiring the first traffic supply attribute data from the traffic management server.

For example, the traffic center server may obtain the road attribute data such as the number of lanes of a road section, the road width, the road speed limit, and the like from the stored data such as the road static topology, the road map, the road sign board, and the like, and the traffic center server may obtain the road attribute data such as the signal timing of an intersection, the traffic split ratio, and the like from the timing scheme stored in the traffic management server. Of course, the traffic center server can also obtain the road attribute data through other ways, such as obtaining the relevant information of the traffic signal lamp from the traffic signal machine.

Optionally, the first traffic provision attribute data comprises one or more of a following model parameter, a straight-ahead split, a left-turn split, a signal period and a number of lanes; the first traffic demand attribute data includes one or more of a vehicle type, a vehicle speed, a vehicle acceleration, a time occupancy, and a lane number.

It should be noted that, in the embodiment of the present application, the traffic center server receives a plurality of types of data included in the traffic flow data reported by the data acquisition side and/or the edge side, and the traffic center server determines the traffic supply and demand attribute data used for determining the road capacity in the embodiment of the present application by using methods such as data cleaning, feature extraction, feature screening, and the like. The traffic supply and demand attribute data determined in the embodiment of the present application include several data only as an example, and the embodiment of the present application does not limit this.

Two implementations for determining the road capacity are provided in the embodiments of the present application, and the two implementations are described in detail below.

The first implementation mode comprises the following steps:and determining the road capacity based on a Gaussian process and a Bayesian optimization method.

Optionally, determining the road capacity in the first time period according to the first traffic supply attribute data and the first traffic demand attribute data through a mapping relationship between the road capacity and the traffic supply and demand attribute, includes: inputting at least one of the first traffic supply attribute data and the first traffic demand attribute data into a Gaussian process model, and outputting a first supply-demand relation function, wherein the Gaussian process model is used for describing a mapping relation between road capacity and the traffic supply-demand attribute; a maximum value of the first supply-demand relationship function is determined as the road capacity in the first time period.

Optionally, determining a maximum value of the first supply-demand relationship function, and taking the maximum value as the road capacity in the first time period, includes: and determining the maximum value of the first supply-demand relation function through Bayesian optimization processing, and taking the maximum value as the road capacity in the first time period.

It should be noted that, the traffic center server takes which data of the first traffic supply attribute data and the first traffic demand attribute data is input as the gaussian process model, which data defines the constraint condition of the finally determined road capacity, and which data is specifically input as the gaussian process model may be selected according to the actual situation, which is not limited in the embodiment of the present application.

For example, the traffic center server inputs the first traffic provision attribute data into the gaussian process model, that is, the traffic center server only uses the first traffic provision attribute data as an input of the gaussian process model, so that the finally determined road capacity represents the maximum capability of the road to bear vehicle operation under the road condition defined by the first traffic provision attribute data, that is, the first traffic provision attribute data is used as a constraint condition for determining the road capacity.

Optionally, determining the road capacity in the first time period according to the first traffic supply attribute data and the first traffic demand attribute data through a mapping relationship between the road capacity and the traffic supply and demand attribute, includes: and inputting the first traffic supply attribute data and the first traffic demand attribute data into a Gaussian process model, and outputting the road capacity in the first time period, wherein the Gaussian process model is used for describing the mapping relation between the road capacity and the traffic supply and demand attribute. That is, the traffic center server directly uses both the first traffic supply attribute data and the first traffic demand attribute data as constraints for determining the road capacity.

It should be noted that, in the first implementation, a method of searching for the maximum value of the function by the bayesian optimization method is taken as an example to describe a method of determining the road capacity, but in the embodiment of the present application, the maximum value of the function can be searched not only by the bayesian optimization method, but also by other optimization methods, such as other gradient descent algorithms.

In addition, the gaussian process model is determined according to the traffic data acquired before the first time period, that is, before the gaussian process model is used, the gaussian process model needs to be obtained through simulation according to historical data.

Optionally, the method further comprises: acquiring multiple groups of first traffic data, wherein the multiple groups of first traffic data are determined before a first time period, and each group of first traffic data comprises traffic supply attribute data, traffic demand attribute data and road flow; and taking the traffic supply attribute data and the traffic demand attribute data in each group of first traffic data as input data, taking the road flow in each group of first traffic data as output data, and obtaining a Gaussian process model through Gaussian process fitting.

That is, the traffic center server takes the traffic supply and demand attribute data in the historical data as an input sample, takes the road flow in the historical data as a sample label, obtains a gaussian process model through gaussian process modeling according to the input sample and the sample label, and the obtained gaussian process model is used for describing the mapping relation between the road capacity and the traffic supply and demand attribute data.

In the embodiment of the application, because the traffic center server continuously collects the traffic flow data reported by the data acquisition side and/or the edge side, and the traffic center server can also continuously obtain the latest road attribute data from the traffic management center and the like, the traffic center server can also update the gaussian process model according to the continuously received traffic flow data and the continuously updated road attribute data.

Optionally, after the traffic center server obtains the gaussian process model through gaussian process fitting, the method further includes: acquiring road flow in a first time period; and updating the Gaussian process model according to the first traffic supply attribute data, the first traffic demand attribute data and the road flow in the first time period.

In the first implementation mode, dynamic calculation is carried out based on real-time traffic supply attribute data, real-time traffic states can be reflected, any prior assumption is not needed for traffic flow operation rules, and a data-driven mode is used for modeling traffic flow relation essence, so that modeling is more accurate. In addition, by using the characteristic engineering, the association between more traffic flow parameters and the flow is excavated, the relation between more characteristic parameters and the road flow is fitted through a Gaussian process, the dynamic characteristic of the traffic flow change of the road section can be better captured, the data-driven calculable characteristic attribute is more comprehensively used, and the calculation result is more accurate.

The second implementation mode comprises the following steps:a neural network based approach determines the road capacity.

In this implementation, the determining, by the traffic center server, the road capacity in the first time period according to the first traffic supply attribute data and the first traffic demand attribute data through the mapping relationship between the road capacity and the traffic supply and demand attribute includes: and inputting the first traffic supply attribute data and the first traffic demand attribute data into a trained neural network model, and outputting the road capacity in a first time period, wherein the trained neural network model is used for acquiring the mapping relation between the road capacity and the traffic supply and demand attribute.

In the embodiment of the application, before the traffic center server predicts the road capacity by using the neural network model, the neural network model needs to be obtained by training according to historical data.

Optionally, the method further comprises: acquiring multiple groups of second traffic data, wherein the multiple groups of second traffic data are determined before a first time period, and each group of second traffic data comprises traffic supply attribute data, traffic demand attribute data, road flow and actual passing time; taking each group of second traffic data and the stored initial parameters as the input of the neural network model to be trained, and obtaining a plurality of passing time errors corresponding to each group of second traffic data; and adjusting the network parameters of the neural network model to be trained through the back propagation of the plurality of transit time errors to obtain the trained neural network model.

Optionally, the neural network model to be trained includes an error correction module, a parameter calibration module and a traffic capacity estimation module, the initial parameters include an initial value of a road resistance parameter and a road section free driving time, and the road resistance parameter is a parameter of a highway Bureau (BPR) function; the traffic center server takes each group of second traffic data and the stored initial parameters as the input of the neural network model to be trained, obtains the passing time errors corresponding to each group of second traffic data, and obtains a plurality of passing time errors, wherein the passing time errors comprise: taking the traffic supply attribute data and the traffic demand attribute data included in each group of second traffic data as the input of a traffic capacity estimation module, and outputting the predicted road capacity corresponding to each group of second traffic data; taking the initial value of the road resistance parameter as the input of a parameter calibration module, and outputting the calibration value of the road resistance parameter; and taking the road flow and the actual passing time included by each group of second traffic data, and the predicted road capacity, the calibration value of the road resistance parameter and the free running time of the road section corresponding to each group of second traffic data as the input of the error correction module, and outputting the passing time error corresponding to each group of second traffic data.

Optionally, the traffic center server uses the road traffic and the actual transit time included in each group of second traffic data, and the predicted road capacity, the calibration value of the road resistance parameter, and the road segment free-running time corresponding to each group of second traffic data as inputs of the error correction module, and outputs the transit time error corresponding to each group of second traffic data, including: taking the road flow included by each group of second traffic data, and the predicted road capacity, the calibration value of the road resistance parameter and the free running time of the road section corresponding to each group of second traffic data as the input of a BPR function, and outputting the predicted passing time corresponding to each group of second traffic data through the BPR function; and taking the actual passing time included by each group of second traffic data and the predicted passing time corresponding to each group of second traffic data as the input of a loss function, and outputting the passing time error corresponding to each group of second traffic data through the loss function.

Optionally, the traffic center server adjusts network parameters of the neural network model to be trained through back propagation of the multiple transit time errors, and obtains the trained neural network model, including: and adjusting network parameters of the traffic capacity estimation module and the parameter calibration module through the back propagation of the plurality of traffic time errors to obtain a trained neural network model.

Because the road passing time is easy to collect, namely the road passing time is a traffic characteristic parameter which can be directly measured, the road passing time is easy to construct a loss function, and end-to-end training is easy to realize, in the second implementation mode, the road passing time is used as the sub-output of the BPR function, and the road capacity is used as the sub-output of the neural network model to be trained, so that the model training process is simplified. The required output road capacity is further converted through the BPR function, namely, the input, the output and the calculation process of the whole road resistance function are regarded as a large neural network, the traffic capacity estimation module is independently made into a small sub-network through the splitting of the road resistance function, and the road traffic time is regarded as the final output of the neural network model to be trained. Capturing a complex nonlinear mapping relation between various influencing factors (traffic supply and demand attributes) and the road section capacity through a neural network, combining the training process of the neural network with the parameter calibration of a road resistance function, and synchronously updating the parameters (such as weight and bias) of the neural network and the parameters of the road resistance function through a back propagation algorithm.

Optionally, the trained neural network model comprises a traffic capacity estimation module; the traffic center server inputs the first traffic supply attribute data and the first traffic demand attribute data into a trained neural network model and outputs the road capacity in a first time period, and the method comprises the following steps: and inputting the first traffic supply attribute data and the first traffic demand attribute data into a traffic capacity estimation module included in the trained neural network model, and outputting the road capacity in the first time period.

That is, the error correction module and the parameter calibration module included in the neural network model to be trained are both used for training the traffic capacity estimation module, and after the training is completed, the traffic capacity estimation module obtained only through the training can predict the road capacity.

Optionally, after obtaining the trained neural network model, the traffic center server further includes: acquiring road flow and actual passing time in a first time period; and updating the neural network model according to the first traffic supply attribute data, the first traffic demand attribute data, and the road flow and the actual passing time in the first time period.

That is, in the embodiment of the present application, the traffic center server can also update the neural network model according to the continuously received traffic flow data and the continuously updated road attribute data.

In the second implementation manner, the traffic center server can perform dynamic calculation based on real-time data, and can reflect a real-time traffic state. The complex nonlinear mapping relation between various influencing factors and the road capacity is captured through the neural network, and the determined road capacity is more accurate.

In a second aspect, there is provided a road capacity determination apparatus having a function of realizing the behavior of the road capacity determination method in the first aspect described above. The device for determining the road capacity comprises one or more modules, and the one or more modules are used for realizing the method for determining the road capacity provided by the first aspect.

That is, there is provided a road capacity determination apparatus including:

the system comprises a first acquisition module, a second acquisition module and a first display module, wherein the first acquisition module is used for acquiring first traffic supply attribute data and first traffic demand attribute data, the first traffic supply attribute data is used for representing road section attributes in a first time period, and the first traffic demand attribute data is used for representing traffic flow attributes in the first time period;

the determining module is used for determining the road capacity in the first time period through a mapping relation between the road capacity and the traffic supply and demand attribute according to the first traffic supply attribute data and the first traffic demand attribute data, the mapping relation is determined according to the traffic supply attribute data, the traffic demand attribute data and the road flow acquired historically, and the traffic supply and demand attribute comprises a traffic supply attribute and a traffic demand attribute.

Optionally, the determining module includes:

the processing submodule is used for inputting at least one of the first traffic supply attribute data and the first traffic demand attribute data into a Gaussian process model and outputting a first supply-demand relationship function, and the Gaussian process model is used for describing a mapping relationship between the road capacity and the traffic supply-demand attribute;

and the first determining submodule is used for determining the maximum value of the first supply-demand relation function, and the maximum value is used as the road capacity in the first time period.

Optionally, the first determining sub-module is configured to:

and determining the maximum value of the first supply-demand relation function through Bayesian optimization processing, and taking the maximum value as the road capacity in the first time period.

Optionally, the determining module includes:

and the second determining submodule is used for inputting the first traffic supply attribute data and the first traffic demand attribute data into a Gaussian process model and outputting the road capacity in the first time period, and the Gaussian process model is used for describing the mapping relation between the road capacity and the traffic supply and demand attribute.

Optionally, the apparatus further comprises:

a second obtaining module, configured to obtain multiple sets of first traffic data, where the multiple sets of first traffic data are determined before a first time period, and each set of first traffic data includes traffic supply attribute data, traffic demand attribute data, and road traffic;

the first learning module is used for taking the traffic supply attribute data and the traffic demand attribute data in each group of first traffic data as input data, taking the road flow in each group of first traffic data as output data, and obtaining a Gaussian process model through Gaussian process fitting.

Optionally, the apparatus further comprises:

the third acquisition module is used for acquiring the road flow in the first time period;

and the first updating module is used for updating the Gaussian process model according to the first traffic supply attribute data, the first traffic demand attribute data and the road flow in the first time period.

Optionally, the determining module includes:

and the third determining submodule is used for inputting the first traffic supply attribute data and the first traffic demand attribute data into the trained neural network model and outputting the road capacity in the first time period, and the trained neural network model is used for acquiring the mapping relation between the road capacity and the traffic supply and demand attribute.

Optionally, the apparatus further comprises:

a fourth obtaining module, configured to obtain multiple sets of second traffic data, where the multiple sets of second traffic data are determined before the first time period, and each set of second traffic data includes traffic supply attribute data, traffic demand attribute data, road traffic and actual transit time;

the second learning module is used for taking each group of second traffic data and the stored initial parameters as the input of the neural network model to be trained, obtaining the passing time error corresponding to each group of second traffic data, and obtaining a plurality of passing time errors;

and the third learning module is used for adjusting the network parameters of the neural network model to be trained through the back propagation of the plurality of passing time errors to obtain the trained neural network model.

Optionally, the neural network model to be trained includes an error correction module, a parameter calibration module and a traffic capacity estimation module, the initial parameters include an initial value of a road resistance parameter and a road section free driving time, and the road resistance parameter is a parameter of a BPR function;

the second learning module includes:

the first training submodule is used for taking the traffic supply attribute data and the traffic demand attribute data included in each group of second traffic data as the input of the traffic capacity estimation module and outputting the predicted road capacity corresponding to each group of second traffic data;

the second training submodule is used for taking the initial value of the road resistance parameter as the input of the parameter calibration module and outputting the calibration value of the road resistance parameter;

and the third training submodule is used for taking the road flow and the actual passing time included by each group of second traffic data, and the predicted road capacity, the calibration value of the road resistance parameter and the free running time of the road section corresponding to each group of second traffic data as the input of the error correction module, and outputting the passing time error corresponding to each group of second traffic data.

Optionally, the third training submodule is configured to:

taking the road flow included by each group of second traffic data, and the predicted road capacity, the calibration value of the road resistance parameter and the free running time of the road section corresponding to each group of second traffic data as the input of a BPR function, and outputting the predicted passing time corresponding to each group of second traffic data through the BPR function;

and taking the actual passing time included by each group of second traffic data and the predicted passing time corresponding to each group of second traffic data as the input of a loss function, and outputting the passing time error corresponding to each group of second traffic data through the loss function.

Optionally, the third learning module comprises:

and the fourth training submodule is used for adjusting the network parameters of the traffic capacity estimation module and the parameter calibration module through the back propagation of the plurality of traffic time errors to obtain a trained neural network model.

Optionally, the trained neural network model comprises a traffic capacity estimation module;

the third determination submodule is configured to:

and inputting the first traffic supply attribute data and the first traffic demand attribute data into a traffic capacity estimation module included in the trained neural network model, and outputting the road capacity in the first time period.

Optionally, the apparatus further comprises:

the fifth acquisition module is used for acquiring the road flow and the actual passing time in the first time period;

and the second updating module is used for updating the neural network model according to the first traffic supply attribute data, the first traffic demand attribute data, and the road flow and the actual passing time in the first time period.

Optionally, the first obtaining module includes:

the receiving submodule is used for receiving traffic flow data sent by a data acquisition side and/or an edge side in a first time period, the data acquisition side is used for acquiring traffic flow data in real time, and the edge side is used for collecting the traffic flow data acquired by the data acquisition side in real time and preprocessing the traffic flow data;

and the fourth determining submodule is used for determining the first traffic demand attribute data according to the received traffic flow data.

Optionally, the data acquisition side comprises one or more traffic sensors, the edge side comprises one or more edge devices, each of the one or more traffic sensors is any one of a sensor of an electric police gate, a sensor of a geomagnetic coil/radar, a GNSS sensor in a vehicle and a positioning sensor of a mobile device in a vehicle.

Optionally, the first obtaining module includes:

a first acquisition sub-module for acquiring first traffic supply attribute data from the stored road attribute data; and/or

And the second acquisition submodule is used for acquiring the first traffic supply attribute data from the traffic management server.

Optionally, the first traffic provision attribute data comprises one or more of a following model parameter, a straight-ahead split, a left-turn split, a signal period and a number of lanes;

the first traffic demand attribute data includes one or more of a vehicle type, a vehicle speed, a vehicle acceleration, a time occupancy, and a lane number.

In a third aspect, there is provided a traffic center side having a corresponding function of implementing the traffic center side in the method for determining road capacity in the first aspect described above. Optionally, the traffic center side comprises a traffic center server.

That is, the traffic center side is configured to obtain first traffic supply attribute data and first traffic demand attribute data, where the first traffic supply attribute data is used to represent link attributes in a first time period, and the first traffic demand attribute data is used to represent traffic flow attributes in the first time period; according to the first traffic supply attribute data and the first traffic demand attribute data, the road capacity in the first time period is determined through a mapping relation between the road capacity and the traffic supply and demand attribute, the mapping relation is determined according to the traffic supply attribute data, the traffic demand attribute data and the road flow acquired historically, and the traffic supply and demand attribute comprises a traffic supply attribute and a traffic demand attribute.

Optionally, the determining, by the traffic center side according to the first traffic supply attribute data and the first traffic demand attribute data, the road capacity in the first time period through a mapping relationship between the road capacity and the traffic supply and demand attribute includes: the first traffic supply attribute data and the first demand attribute data are processed through a Gaussian process model or a neural network model, and the road capacity in the first time period is determined. Wherein, the Gaussian process model or the neural network model is used for obtaining the mapping relation.

Optionally, the acquiring, by the traffic center side, the first traffic demand attribute data includes: receiving traffic flow data sent by a data acquisition side and/or an edge side in a first time period, wherein the data acquisition side is used for acquiring traffic flow data in real time, and the edge side is used for collecting the traffic flow data acquired by the data acquisition side in real time and preprocessing the traffic flow data; and determining first traffic demand attribute data according to the received traffic flow data.

Optionally, the traffic center side acquires first traffic supply attribute data, including: acquiring first traffic supply attribute data from the stored road attribute data; and/or acquiring the first traffic supply attribute data from the traffic management server.

In a fourth aspect, a data acquisition side is provided, where the data acquisition side has a corresponding function of implementing the data acquisition side in the method for determining road capacity in the first aspect.

That is, the data collection side is used to collect traffic flow data in real time. The data acquisition side is also used for sending the acquired traffic flow data to the edge side, and the edge side preprocesses the received traffic flow data and sends the preprocessed traffic flow data to the traffic center side. Or the data acquisition side sends the acquired traffic flow data to the traffic center side. After the traffic center side receives the traffic flow data, the road capacity is predicted according to the determination method of the road capacity provided in the above-described first aspect.

Optionally, the data collection side comprises one or more traffic sensors, each of which is any one of a sensor of an electric police checkpoint, a sensor of a geomagnetic coil/radar, a Global Navigation Satellite System (GNSS) sensor in a vehicle, and a positioning sensor of a mobile device in a vehicle.

In a fifth aspect, an edge side is provided, which has corresponding functions to implement the edge side in the method for determining road capacity in the first aspect.

That is, the edge side is used for collecting and preprocessing the traffic flow data sent by the data acquisition side, and the edge side is also used for sending the preprocessed traffic flow data to the traffic center side, and the traffic center side predicts the road capacity according to the method for determining the road capacity provided by the first aspect.

Optionally, the edge side comprises one or more edge devices and the data collection side comprises one or more traffic sensors, each edge side for collecting traffic flow data collected by at least one traffic sensor in the data collection side.

In a sixth aspect, a computer device is provided, which comprises a processor and a memory, wherein the memory is used for storing a program for executing the method for determining the road capacity provided by the first aspect, and storing data for realizing the method for determining the road capacity provided by the first aspect. The processor is configured to execute programs stored in the memory. The operating means of the memory device may further comprise a communication bus for establishing a connection between the processor and the memory.

In a seventh aspect, there is provided a computer-readable storage medium having stored therein instructions, which, when run on a computer, cause the computer to execute the method for determining a road capacity according to the first aspect.

In an eighth aspect, there is provided a computer program product containing instructions which, when run on a computer, cause the computer to perform the method of determining road capacity of the first aspect described above.

The technical effects obtained by the above second, third, fourth, fifth, sixth, seventh and eighth aspects are similar to the technical effects obtained by the corresponding technical means in the first aspect, and are not described herein again.

The technical scheme provided by the embodiment of the application can at least bring the following beneficial effects:

Drawings

Fig. 1 is a system architecture diagram according to a method for determining road capacity provided in an embodiment of the present application;

fig. 2 is a flowchart of a method for determining road capacity according to an embodiment of the present disclosure;

FIG. 3 is a schematic structural diagram of a computer device according to an embodiment of the present disclosure;

fig. 4 is a flowchart of a method for determining road capacity according to an embodiment of the present disclosure;

FIG. 5 is a schematic structural diagram of a neural network model to be trained according to an embodiment of the present disclosure;

fig. 6 is a flowchart of another road capacity determination method provided in the embodiment of the present application;

fig. 7 is a schematic structural diagram of a device for determining road capacity according to an embodiment of the present application.

Detailed Description

To make the objects, technical solutions and advantages of the embodiments of the present application more clear, the embodiments of the present application will be further described in detail with reference to the accompanying drawings.

In order to facilitate understanding of the embodiments of the present application, some terms referred to in the embodiments of the present application are first explained.

Road flow: the number of vehicles actually passing on the road.

Road capacity: the road traffic capacity or road bearing capacity refers to the maximum number of vehicles passing through a certain point or a certain cross section on a road in unit time under the requirements of normal roads, traffic, control, running quality and the like. I.e. the maximum value of the road flow under given conditions.

Traffic supply attribute data: the related data of the traffic supply attribute, namely the related data of the road attribute, is used for representing road infrastructure factors, such as dynamically-changed signal light information, dynamic speed limit of the road, road width, management and control strategies, lane number and the like. The signal light information includes a green signal ratio, a signal period, and the like.

Traffic demand attribute data: and the related data of the traffic demand attribute are used for representing the related data of the traffic flow, and comprise vehicle types, vehicle type proportions, vehicle speeds, vehicle acceleration, vehicle following model parameters and the like.

The green signal ratio: the time proportion of the traffic signal lamp available for the vehicle to pass in one signal period is indicated. I.e. the ratio of the active green time to the signal period within a signal period. For example, the direct green ratio is the ratio of the effective time of a direct green light to the signal period in one signal period, and the left-turning green ratio is the ratio of the effective time of a left-turning green light to the signal period in one signal period.

Following model parameters: i.e. the corresponding parameters of the following vehicle model, such as the safety distance. The following model is used for researching the corresponding behavior of the following vehicle caused by the change of the motion state of the front vehicle by using a dynamic method and mainly expresses the relation between the acceleration of the vehicle and the traffic parameters such as the distance between adjacent vehicles and the speed difference.

The road resistance function: the BPR function, also known as the federal highway administration function, also known as the road resistance function, is used for calculating the free-driving time of a road segment. The road resistance function is specifically shown in formula (1).

In the formula (1), t_iThe time required for actually passing a road, i.e. the actual transit time, t_i0The traffic capacity of the road is the free running time of the road section, Q is the traffic volume of the road at the time, namely the traffic flow, C is the actual traffic capacity of the road, namely the road capacity, and alpha and beta are road resistance parameters and are undetermined parameters.

Road section free travel time: also referred to as free flow time or link free transit time, etc., refers to the time it takes for a vehicle to travel freely through a road while the road is in a clear state.

Next, a system architecture according to an embodiment of the present application will be described. It should be noted that the system architecture and the service scenario described in the embodiment of the present application are for more clearly illustrating the technical solution of the embodiment of the present application, and do not constitute a limitation to the technical solution provided in the embodiment of the present application, and as a person having ordinary skill in the art knows that along with the evolution of the network architecture and the appearance of a new service scenario, the technical solution provided in the embodiment of the present application is also applicable to similar technical problems.

Fig. 1 is a system architecture diagram according to a method for determining road capacity provided in an embodiment of the present application. Referring to fig. 1, the system architecture includes a data acquisition side 101, an edge side 102, and a traffic center side 103, wherein the data acquisition side 101 includes one or more traffic sensors, the edge side 102 includes one or more edge devices, and the traffic center side 103 includes a traffic center server. Each edge device is connected with at least one traffic sensor in a wired or wireless mode for communication, and each edge device is connected with a traffic center server in a wired or wireless mode for communication.

In the embodiment of the present application, the data collection side 101 includes one or more traffic sensors for collecting traffic flow data in real time and transmitting the collected traffic flow data to the connected edge devices. Optionally, each of the one or more traffic sensors is any one of a sensor of an electric police station, a sensor of a geomagnetic coil/radar, a GNSS sensor in a vehicle, and a positioning sensor of a mobile device in a vehicle. The GNSS includes one or more of a Global Positioning System (GPS) in the united states, a beidou satellite positioning system in china, a glonass navigation system in russia, and a galileo navigation system in europe.

Optionally, the traffic sensors are mounted on the road and/or deployed on the vehicle. For example, sensors (detectors, cameras, etc.) of an electric police station are installed at both sides of a road, an intersection, etc., a sensor of a geomagnetic coil/radar is buried under a road surface of the road, and a GNSS sensor and a positioning sensor of a mobile device are deployed in a vehicle. Optionally, the vehicle includes a GNSS sensor in the vehicle, and the mobile device in the vehicle is a mobile phone, a tablet computer, or the like.

In the embodiment of the present application, the edge side 102 includes one or more edge devices for collecting and preprocessing the traffic flow data collected by the data collection side 101 in real time, and reporting the preprocessed traffic flow data to the traffic center server. That is, the edge device is used for aggregation, fusion and reporting of traffic flow data.

It should be noted that, because the traffic sensors included in the data acquisition side 101 are various, that is, the data acquisition side 101 includes a plurality of heterogeneous data generation sources, and the edge side 102 plays a role in data aggregation and data fusion, in this way, the data reported by the edge side 102 to the traffic center server is preprocessed data, which can reduce the calculation pressure of the traffic center server.

Optionally, each edge device is connected to some or all of the traffic sensors included in the data collection side 101 for collecting traffic flow data collected by the connected traffic sensors in real time.

Optionally, part or all of the traffic sensors included in the data acquisition side 101 are directly in communication connection with the traffic center server, and are configured to directly report the acquired traffic flow data to the traffic center server, and the traffic center server performs preprocessing on the received data. In the case that all traffic sensors included in the data acquisition side 101 are directly connected to the traffic center server in a communication manner, the system architecture may not include the edge side 102, and the traffic center side 103 performs preprocessing such as aggregation and fusion on the traffic flow data.

That is, for the data aggregation, the traffic sensor aggregates the acquired data to the edge side 102, and then the acquired data is aggregated to the traffic center side 103 by the edge side 102, or the traffic sensor may directly aggregate the acquired data to the traffic center side 103.

In the embodiment of the present application, the traffic center server included in the traffic center side 103 is configured to determine the first traffic demand attribute data according to the traffic flow data sent by the data collection side 101 and/or the edge device in the first time period. The traffic center side 103 is further configured to determine the road capacity in the first time period according to the first traffic supply attribute data and the first traffic demand attribute data by the road capacity determination method provided in the embodiment of the present application.

In an embodiment of the application, the traffic center server is configured to determine first traffic provision attribute data from road attribute data during a first time period. Optionally, the traffic center server stores therein all or part of the road attribute data for determining the first traffic provision attribute data, or the traffic center server does not store therein the road attribute data. In the case that the traffic center server does not store all data for determining the first traffic supply attribute data, the system architecture further includes a traffic management server, and the traffic center server is further configured to obtain some or all data for determining the first traffic supply attribute data from the traffic management server. Alternatively, the traffic center server may also be capable of acquiring a part of the road attribute data from the traffic signal. That is, the traffic center server obtains the road attribute data for determining the first traffic supply attribute data from various data sources, which is not limited in the embodiment of the present application.

For example, the traffic center server may obtain the road attribute data such as the number of lanes of a road section, the road width, the road speed limit, and the like from the stored data such as the road static topology, the road map, the road sign board, and the like, and the traffic center server may obtain the road attribute data such as the signal timing, the traffic split ratio, and the like of an intersection from the timing scheme stored in the traffic management server or from the traffic signal machine.

It should be noted that the traffic management server refers to a server of the traffic management center, and the server may be an entity physical machine or a virtual machine configured in the cloud.

Optionally, the traffic center server is configured to periodically and actively acquire and store the latest road attribute data from the traffic management server and/or the traffic signal. Optionally, the traffic management server and/or the traffic signal may also send the updated road attribute data to the traffic center server after updating the road attribute data. That is, the traffic center server may determine the real-time traffic supply attribute data through active query or passive reception, which is not limited in the embodiment of the present application.

It should be noted that, since the data collection side 101 collects traffic flow data in real time, the traffic center side 103 can determine real-time traffic demand attribute data, and the traffic center side 103 can also obtain real-time traffic supply attribute data, according to the present scheme, the traffic center server can determine the road capacity in real time.

In the embodiment of the present application, the data collection side 101 includes any one or more of the traffic sensors described above, and besides, any other sensor capable of collecting traffic flow data may take on the role of a traffic sensor in the embodiment of the present application, that is, the traffic sensor is not limited in the embodiment of the present application. Optionally, the edge device refers to an edge server, that is, the edge device is a server, and the edge device may also be another form of computing device. The traffic center server is a server, or a server cluster formed by a plurality of servers, or a cloud computing service center.

Next, taking the system architecture including a data acquisition side, an edge side, and a traffic center side as an example, the method for determining road capacity provided by the embodiment of the present application is introduced with reference to the system architecture.

Fig. 2 is a flowchart of a method for determining road capacity according to an embodiment of the present disclosure. Referring to fig. 2, the data acquisition side acquires traffic flow data through the traffic sensor and reports the data to the edge side, and the edge side aggregates the data acquired by the data acquisition side and reports the data to the traffic center side after preprocessing.

For the traffic center side, on the one hand, the traffic center server is able to predict the road capacity. Specifically, the traffic center server processes traffic flow data received in a first time period and stored static data through a stored bearing capacity calculation model to obtain a dynamic calculation result, that is, to obtain the road capacity in the first time period. The static data stored by the traffic center server comprises first traffic supply attribute data, the traffic center server determines first traffic demand attribute data according to traffic flow data received in a first time period, the first traffic demand attribute data and the stored static data are input into a bearing capacity calculation model, and a dynamic calculation result, namely the road capacity in the first time period, is output.

On the other hand, the traffic center server is able to train and update the load calculation model. Specifically, the traffic center server accumulates traffic flow data received within a period of time, performs data cleaning on the accumulated data, performs data storage on the traffic flow data after data cleaning, stores the traffic flow data in a historical database, also stores road attribute data acquired historically in a corresponding time period, performs feature extraction (feature screening) and the like on the traffic flow data and the road attribute data stored in the historical database by the traffic center server to obtain traffic demand attribute data and traffic supply attribute data for a training model, and trains to obtain a bearing capacity calculation model according to the traffic demand data and the traffic supply attribute data determined by the data accumulated historically. The traffic center server can also update the bearing capacity calculation model according to the traffic flow data acquired in real time and the latest traffic supply attribute data. For example, the load calculation model can be trained not only from data accumulated before the first time period, but also updated from data in the first time period.

It should be noted that, here, the mapping relationship between the road capacity and the traffic demand and supply attribute is described by the traffic center side through the stored bearing capacity calculation model. For the detailed description of the load bearing capability calculation model, reference is made to the detailed description of the embodiment of fig. 4 below, and the detailed description is omitted here.

Referring to fig. 3, fig. 3 is a schematic structural diagram of a computer device according to an embodiment of the present application. Optionally, the computer device is a traffic center server or edge device as shown in fig. 1 or fig. 2, and the computer device includes one or more processors 301, a communication bus 302, a memory 303, and one or more communication interfaces 304.

The processor 301 is a general-purpose Central Processing Unit (CPU), a Network Processor (NP), a microprocessor, or one or more integrated circuits such as an application-specific integrated circuit (ASIC), a Programmable Logic Device (PLD), or a combination thereof for implementing the present invention. Optionally, the PLD is a Complex Programmable Logic Device (CPLD), a field-programmable gate array (FPGA), a General Array Logic (GAL), or any combination thereof.

A communication bus 302 is used to transfer information between the above components. Optionally, the communication bus 302 is divided into an address bus, a data bus, a control bus, and the like. For ease of illustration, only one thick line is shown, but this does not mean that there is only one bus or one type of bus.

Optionally, the memory 303 is a read-only memory (ROM), a Random Access Memory (RAM), an electrically erasable programmable read-only memory (EEPROM), an optical disk (including a compact disc read-only memory (CD-ROM), a compact disc, a laser disk, a digital versatile disk, a blu-ray disk, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and that can be accessed by a computer, but is not limited to. The memory 303 is separate and connected to the processor 301 through the communication bus 302, or the memory 303 is integrated with the processor 301.

The communication interface 304 uses any transceiver or the like for communicating with other devices or communication networks. The communication interface 304 includes a wired communication interface and optionally a wireless communication interface. The wired communication interface is, for example, an ethernet interface. Optionally, the ethernet interface is an optical interface, an electrical interface, or a combination thereof. The wireless communication interface is a Wireless Local Area Network (WLAN) interface, a cellular network communication interface, or a combination thereof. When the computer device is a traffic center server, the communication interface 304 is used for communicating with the data acquisition side and/or the edge side, and when the computer device is an edge device, the communication interface 304 is used for communicating with the data acquisition side and the traffic center server.

Optionally, in some embodiments, the computer device comprises a plurality of processors, such as processor 301 and processor 305 shown in fig. 3. Each of these processors is a single core processor, or a multi-core processor. A processor herein optionally refers to one or more devices, circuits, and/or processing cores for processing data (e.g., computer program instructions).

Optionally, in some embodiments, the computer device further comprises an output device 306 and an input device 307. An output device 306 is in communication with the processor 301 and is capable of displaying information in a variety of ways. For example, the output device 306 is a Liquid Crystal Display (LCD), a Light Emitting Diode (LED) display device, a Cathode Ray Tube (CRT) display device, a projector (projector), or the like. The input device 307 is in communication with the processor 301 and is capable of receiving user input in a variety of ways. The input device 307 is, for example, a mouse, a keyboard, a touch screen device, a sensing device, or the like.

In some embodiments, the memory 303 is used to store program code 310 for performing aspects of the present application, and the processor 301 is capable of executing the program code 310 stored in the memory 303. The program code 310 includes one or more software modules, and the computer device can implement the method for determining road capacity provided in the embodiment of fig. 4 below by the processor 301 and the program code 310 in the memory 303.

For example, in the case that the first obtaining module 701 and the determining module 702 in the embodiment shown in fig. 7 are implemented by software, the program code 310 may include a first obtaining module for obtaining the first traffic supply attribute data and the first traffic demand attribute data, and a determining module for determining the road capacity in the first time period according to the first traffic supply attribute data and the first traffic demand attribute data.

Fig. 4 is a flowchart of a method for determining road capacity, which is applied to a traffic center server according to an embodiment of the present application. Referring to fig. 4, the method includes the following steps.

Step 401: the method comprises the steps of obtaining first traffic supply attribute data and first traffic demand attribute data, wherein the first traffic supply attribute data are used for representing road section attributes in a first time period, and the first traffic demand attribute data are used for representing traffic flow attributes in the first time period.

In the embodiment of the present application, as can be seen from the foregoing description of the system architecture related to the embodiment of the present application, the traffic center server receives traffic flow data sent by the data acquisition side and/or the edge side in the first time period, and determines the first traffic demand attribute data according to the received traffic flow data. The data acquisition side is used for acquiring traffic flow data in real time, and the edge side is used for collecting the traffic flow data acquired by the data acquisition side in real time and preprocessing the traffic flow data.

For data aggregation, the edge side performs data aggregation on the data acquisition side, for example, the traffic sensor aggregates the acquired data to the edge side, and then the edge side aggregates the acquired data to the traffic center side in a unified manner, or the traffic sensor directly aggregates the acquired data to the traffic center side.

Optionally, the data collection side comprises one or more traffic sensors, the edge side comprises one or more edge devices, each of the one or more traffic sensors is any one of a sensor of an electric police station, a sensor of a geomagnetic coil/radar, a Global Navigation Satellite System (GNSS) sensor in the vehicle, and a positioning sensor of a mobile device in the vehicle.

The traffic flow data collected by the sensors of the electric police checkpoint are called checkpoint data, and the core fields of each checkpoint data comprise checkpoint numbers, vehicle numbers, lane numbers, timestamps, instantaneous speeds and other fields. And the traffic center server or the edge device takes the total recorded data of the card port data of a road in a period of time as the road flow of the road in the period of time. Based on the instantaneous speed field included in the core field of the card port data of a road over a period of time, the traffic center server or the edge device can calculate relevant statistics of the vehicle speed, such as the arithmetic mean, harmonic mean, 85 quantile, median, non-0 proportional value, maximum, range, etc. of the vehicle speed. The range error is also called range error, and is used to represent the variation quantity (measures of variation) in the statistical data, which is the difference between the maximum value and the minimum value in the statistical data, i.e. the data obtained by subtracting the maximum value from the minimum value.

The traffic flow data collected by the sensor of the geomagnetic coil is called coil data, and the core field of each coil data comprises fields such as a coil number, a lane number, a timestamp, an instantaneous speed and a time occupancy rate. And the traffic center server or the edge device takes the total recorded number of the coil data of a road in a period of time as the road flow of the road in the period of time. Based on the instantaneous speed field and the time occupancy field included in the core field of the coil data of a road over a period of time, the traffic center server or the edge device can calculate the relevant statistics of the vehicle speed and the time occupancy, such as the arithmetic mean, harmonic mean, 85 quantiles, median, non-0 value proportion, maximum, range, etc. of the vehicle speed or the time occupancy, respectively.

Data collected by a GNSS sensor in the vehicle and a positioning sensor of a mobile device in the vehicle become floating vehicle data, and the core field of each floating vehicle data comprises fields such as a vehicle number, a timestamp, longitude, latitude, vehicle speed and the like. The traffic center server or the edge device may count the total number of vehicles in a given time window on a road, and obtain the road traffic of the road in the time window. Based on the vehicle speed included in the core field of the floating vehicle data, the traffic center server or edge device may calculate relevant statistics of vehicle speed, such as calculating an arithmetic mean, harmonic mean, 85 quantile, median, non-0 scale, maximum, range, etc. of vehicle speed.

As can be seen from the foregoing, traffic sensors are of various types, and acquired data are also diverse, and there are many redundant data, and even if there is edge device to perform data aggregation and data fusion, the traffic center side needs further data cleaning and feature extraction (feature screening). In the embodiment of the application, the method for the traffic center server to perform data cleansing may include one or more of abnormal data removal, data normalization, duplicate data de-redundancy, duplicate data averaging, and the like, and the method for feature extraction may include, but is not limited to, one or more of feature screening based on a decision tree, feature screening based on a random forest, feature screening based on a support vector machine, and the like.

Optionally, the first traffic demand attribute data comprises one or more of vehicle type, vehicle speed, vehicle acceleration, time occupancy and lane number. That is, in the embodiment of the present application, the traffic demand attribute data determined by the traffic center server after feature screening includes one or more of a vehicle type, a vehicle speed, a vehicle acceleration, a time occupancy rate, and a lane number. It should be noted that the method for feature screening and the time and environment of the collected raw data, which are adopted by the traffic center server, all affect what the finally determined traffic demand attribute data specifically includes, that is, in some other embodiments, the traffic demand attribute data may be different from what is described in the embodiments of the present application, and the embodiments of the present application do not limit what the traffic demand attribute data specifically includes.

And for the first traffic demand attribute data, the traffic center server determines the first traffic demand attribute data according to the traffic flow data received in the first time period. Illustratively, the traffic center server deletes redundant data in the received traffic flow data, retains several kinds of data remaining after feature screening, such as a bayonet number, a coil number, longitude and latitude, and the like, and performs preprocessing on the retained data, such as data normalization and the like.

Optionally, in this embodiment of the present application, the first traffic demand attribute data further includes data collection time. The traffic center server can intelligently analyze the data acquisition time and determine the traffic supply attribute data of the corresponding time period, namely determine the corresponding first traffic supply attribute data.

The foregoing describes an implementation manner in which the traffic center server obtains the first traffic demand attribute data, and then describes an implementation manner in which the traffic center server or the first traffic supply attribute data is obtained.

In the embodiment of the application, the traffic center server acquires the first traffic supply attribute data from the stored road attribute data and/or acquires the first traffic supply attribute data from the traffic management server.

For example, the traffic center server may obtain the road attribute data such as the number of lanes of a road section, the road width, the road speed limit, and the like from the stored data such as the road static topology, the road map, the road sign board, and the like, and the traffic center server may obtain the road attribute data such as the signal timing of an intersection, the traffic split ratio, and the like from the timing scheme stored in the traffic management server.

Alternatively, the traffic center server may also be capable of acquiring a part of the road attribute data from the traffic signal. For example, the traffic center server may acquire road attribute data such as signal timing at an intersection, a traffic split ratio, and the like from a traffic signal.

That is, the traffic center server obtains the road attribute data for determining the first traffic supply attribute data from various data sources, which is not limited in the embodiment of the present application.

Optionally, the first traffic provision attribute data comprises one or more of a following model parameter, a straight ahead split, a left turn split, a signal period and a number of lanes. That is, in the embodiment of the present application, the traffic demand attribute data determined by the traffic center server after feature screening includes one or more of a vehicle following model parameter, a straight-ahead split ratio, a left-turn split ratio, a signal period, and a number of lanes. It should be noted that the method for feature screening adopted by the traffic center server, the time and the environment of the collected raw data, and the like all have an influence on what the finally determined traffic provision attribute data specifically includes, that is, in some other embodiments, the traffic provision attribute data may be different from what is described in the embodiments of the present application, and the embodiments of the present application do not limit what the traffic provision attribute data specifically includes.

For the first traffic supply attribute data, the traffic center server acquires data in a first time period from the road attribute data stored in the traffic center server, and/or acquires the road attribute data in the first time period from the traffic management server, and determines the first traffic demand attribute data according to the acquired data. Illustratively, the traffic center server deletes redundant data in the acquired road attribute data, retains several kinds of data remaining after feature screening, and performs preprocessing, such as data normalization, on the retained data.

Optionally, in this embodiment of the present application, the first traffic supply attribute data further includes a corresponding effective date. And the traffic center server intelligently analyzes the corresponding effective date and determines the traffic demand attribute data corresponding to the data acquisition time, namely determines the corresponding first traffic demand attribute data.

Step 402: and determining the road capacity in the first time period through a mapping relation between the road capacity and the traffic supply and demand attribute according to the first traffic supply attribute data and the first traffic demand attribute data, wherein the mapping relation is determined according to the traffic supply attribute data, the traffic demand attribute data and the road flow acquired in history.

In the embodiment of the application, after the traffic center server acquires the first traffic supply attribute data and the first traffic demand attribute data, the traffic center server determines the road capacity in the first time period through the mapping relation between the road capacity and the traffic supply and demand attribute. The mapping relation is determined according to the traffic supply attribute data, the traffic demand attribute data and the road flow acquired by history, and the traffic supply and demand attribute comprises a traffic supply attribute and a traffic demand attribute.

It should be noted that the embodiments of the present application provide two implementations for determining the road capacity, and the two implementations are described in detail below.

That is, the traffic center server calculates the road capacity through a gaussian process model and bayesian optimization. The Gaussian process model is used for describing the mapping relation between the road capacity and the traffic supply and demand attributes.

In the embodiment of the application, the traffic center server inputs at least one of the first traffic supply attribute data and the first traffic demand attribute data into a gaussian process model, outputs a first supply-demand relation function, then determines a maximum value of the first supply-demand relation function through bayesian optimization, and takes the maximum value as the road capacity in a first time period.

Alternatively, the traffic center server not only uses the first traffic supply attribute data as an input of the gaussian process model, but also uses at least one of the first traffic demand attribute data as an input of the gaussian process model, so that the finally determined road capacity represents the maximum capacity of the road capable of carrying the vehicle operation under the road condition defined by the first traffic supply attribute data and under the condition defining the at least one traffic demand, that is, the first traffic supply attribute data and the at least one traffic demand are used as constraints for determining the road capacity.

For example, the traffic center server uses the first traffic supply attribute data and the vehicle speed included in the first traffic demand attribute data as input of a gaussian process model, that is, the first traffic supply attribute data is used to define the road condition, and the vehicle speed is used to define the traffic demand, so that the finally determined road capacity represents the maximum capacity of the road for the vehicle to run when the vehicle runs according to the vehicle speed under the defined road condition. For another example, the traffic center server uses the first traffic supply attribute data and the vehicle type as input of the gaussian process model, that is, the first traffic supply attribute data is used to define the road condition, and the vehicle type is used to define the traffic demand, so that the finally determined road capacity represents the maximum capacity of the road for the vehicle to run when all the vehicles running on the road under the defined road condition are the vehicle type.

Or the traffic center server inputs the first traffic supply attribute data and the first traffic demand attribute data into a Gaussian process model and outputs the road capacity in the first time period. That is, the traffic center server directly uses both the first traffic supply attribute data and the first traffic demand attribute data as constraints for determining the road capacity.

Optionally, while determining the maximum value of the first supply-demand relationship function, that is, while determining the road capacity, the traffic demand attribute data corresponding to the maximum value may also be determined, that is, the traffic demand data under a given constraint condition (for example, defining the first traffic supply attribute data as an input of a gaussian process model) is estimated, that is, the estimated traffic demand may also be output through a bayesian optimization process, and the estimated traffic demand data is used to further constrain the determined road capacity, that is, to describe what the traffic demand is under the road capacity, for example, what the average vehicle speed is approximately under the road capacity.

It should be noted that, in the embodiment of the present application, the traffic center server takes the first traffic demand attribute data as an input of the gaussian process model, and details of the first implementation manner are described above.

In the embodiment of the present application, the gaussian process model is determined according to the traffic data acquired before the first time period, that is, before the gaussian process model is used, the gaussian process model needs to be obtained through simulation according to historical data.

In the embodiment of the application, the traffic center server obtains multiple groups of first traffic data, the multiple groups of first traffic data are determined before a first time period, each group of first traffic data comprises traffic supply attribute data, traffic demand attribute data and road traffic, the traffic center server takes the traffic supply attribute data and the traffic demand attribute data in each group of first traffic data as input data, takes the road traffic in each group of first traffic data as output data, and obtains a gaussian process model through gaussian process fitting.

That is, the traffic supply and demand attribute data in the historical data is used as an input sample, the road flow in the historical data is used as a sample label, a gaussian process model is obtained through gaussian process modeling according to the input sample and the sample label, and the obtained gaussian process model is used for describing the mapping relation between the road capacity and the traffic supply and demand attribute data.

Optionally, in this embodiment of the application, the traffic center server may further obtain a road flow in the first time period, and update the gaussian process model according to the first traffic supply attribute data, the first traffic demand attribute data, and the road flow in the first time period.

Optionally, the traffic center server or the edge device determines the road traffic flow in the first time period according to the traffic flow data received in the first time period, and the edge device sends the determined road traffic flow in the first time period to the traffic center server when determining the road traffic flow in the first time period.

Illustratively, several implementations of determining road traffic are described herein including, but not limited to: counting the road flow in a first time period based on the number of records of the gate data of the electric police gate; or counting the road flow in the first time period based on the recording number of the sensors of the geomagnetic coil/radar; or calculating the time difference of two vehicles passing through a sensor of a geomagnetic coil/radar in the first time period, and taking the reciprocal of the time difference to obtain the road flow in the first time period; or calculating the number of vehicles matched with the road in the first time period based on GNSS positioning to obtain the road flow in the first time period.

It should be noted that, when the gaussian process model is updated by the first traffic supply attribute data and the first traffic demand attribute data, the first traffic supply attribute data and the first traffic demand attribute data are regarded as data included in the history data, and the first time period is regarded as a history time period, that is, the gaussian process model is updated by the history data.

Next, the gaussian process and the bayesian optimization described in the embodiments of the present application will be described first.

Gaussian Process (GP) is a random process in probability theory and mathematical statistics, and Gaussian process model is a statistical model in which observations occur in a continuous domain (e.g., time or space). In the gaussian process, each point in the continuous input space is associated with a normally distributed random variable. In addition, each finite set of these random variables has a multivariate normal distribution. The distribution of the gaussian process is a joint distribution of all those (an infinite number of) random variables.

Given N observations sampled from an unknown acquisition function H

A function f can be fitted using a gaussian process. For any input vector X (e.g., traffic supply attribute data) before training of the Gaussian process modelAnd traffic demand attribute data), the gaussian process can give a prior distribution of the output label Y (road traffic) at that time, as shown in equation (2).

In equation (2), m (x) is a prior mean, usually assumed to be 0, and K is a covariance matrix. The covariance matrix K of the above formula is of size N, where each element in K is generally determined by a kernel function K, having K_ij＝κ(x_i,x_j)，K_ijRepresenting a random variable x corresponding to the ith and j observations_iAnd x_jThe covariance between. Commonly used kernel functions include Radial Basis Function (RBF) kernel functions and Matern kernel functions, which are shown in equations (3) and (4), respectively.

The RBF kernel in equation (3) is a simplified form and contains a bandwidth parameter/. The Matern kernel in the formula (4) comprises a bandwidth parameter l and a smoothing coefficient v, H_vCorrecting Bessel function for the second kind, | · | | non-woven phosphor₂And the Euclidean distance is represented by the Poisson distribution of gamma (·). The embodiment of the present application is described by taking a kernel function as a Matern kernel function as an example.

The observed input vector and output vector are labeled as (x, y) (i.e., historical data), and the unobserved input vector and corresponding output vector are labeled as (x)_*,y_*) (functional relationship to be predicted) there is a joint Gaussian distribution, as shown in equation (5).

In formula (5), K ═ κ (x, x), K_*＝κ(x,x_*)，K_**＝κ(x_*,x_*)。

Obtaining posterior distribution of unobserved values according to equation (4):

wherein the content of the first and second substances,

after the posterior distribution of the unobserved values is obtained, the optimization of the kernel function parameters in the posterior distribution is realized by maximizing a Log Marginal Likelihood (LML) function, as shown in formula (6).

In the formula (6), θ is a kernel function parameter, for example, for an RBF kernel function, θ is a parameter l, and for a Matern kernel function, θ is a parameter l, v. Since the prior mean is generally assumed to be 0, equation (6) can be simplified to equation (7).

By calculating the partial derivative of the LML with respect to θ, various gradient-based optimization algorithms can be used to obtain the optimal kernel function parameters (i.e., parameters capable of maximizing the log-edge likelihood function), and in the embodiment of the present application, a multi-origin BFGS quasi-newton algorithm is used to solve the maximum value thereof, as shown in formula (8).

In equation (8), tr (·) represents a transpose operation of a matrix.

By the method, the functional relation Y ═ f (X) between the road flow Y and the traffic supply and demand attribute data X can be obtained, and a Gaussian process model is obtained.

After the gaussian process model is obtained, the maximum value of the road flow under the input, namely the road capacity, is obtained by inputting traffic data (such as inputting first traffic supply attribute data) and a bayesian optimization method.

Taking the example of inputting only the traffic supply attribute data, in this case, the road capacity is considered as the maximum flow rate at a certain traffic supply attribute, and the road capacity may be defined as equation (9).

In formula (9), C (x)_r) To be given at x_rRoad capacity of lower, x_rSupply attribute data for known traffic, x_tIs unknown traffic demand attribute data.

For the optimization problem, the highest point of the function in the traffic demand attribute change process is solved through Bayesian optimization, namely the maximum value of the supply-demand relation is solved.

It should be noted that bayesian optimization is a non-gradient optimization method, and for a function f (x) to be optimized_r,x_t) Based on the posterior probability distribution of the Gaussian process, a set of search rules is constructed, the input value of the function to be optimized can be continuously searched, and therefore the parameter combination which enables the function to be optimized to be maximized is quickly found.

An exemplary Bayesian optimization process for solving the maximum road flow is described as follows:

the input includes: function f (x) to be optimized_r,x_t) Knowing the parameter x_rNumber of initial sampling points n_starterMaximum number of optimization rounds n_explore，

Collection function H (x), parameter x to be optimized_tA search range bound, a probabilistic proxy model g.

The output includes: function f(x_r,x_t) Given a known parameter x_rMaximum value C and corresponding parameter x under the conditions_r。

And (3) solving: initializing an observed sample set X and a sample tag set Y, X comprising X_tAnd x_rAnd Y comprises Y;

X_starterrandom uniform sampling n within the search range bound_starterGroup x_tA value;

Y_starter＝f(x_r,X_starter)，X＝X∪X_starter，y＝y∪y_starter；

the optimal function value yopt is max (y), and the optimal parameter xopt is X [ arg max (y) ];

For i＝0:n_explore

training model g using data (X, y) to obtain posterior distribution

X＝X∪arg max H(x)，y＝y∪f(x_r,arg max H(x))；

yopt＝max(y)，xopt＝X[arg max(y)]；

End For

Return yopt，xopt

When the road capacity is estimated, the function to be optimized is a road flow estimation model, namely the introduced first supply-demand relation function, under the given traffic supply attribute condition, Bayesian optimization searches each traffic flow parameter, and determines a parameter combination capable of maximizing the road flow as the road capacity under the current supply condition.

An example provided by the embodiment of the present application is given below to illustrate the above optimization process.

1. Collecting traffic data, forming inputs x of an algorithmic model_tAnd x_r。

x_tAttribute data for traffic demand, including: vehicle number (vehno), vehicle type (vehtype), vehicle speed (speed), vehicle acceleration (acc), time occupancy (occ), lane number (nla), data acquisitionSet time (ptime). x is the number of_tAs shown in table 1.

TABLE 1

A	B	C	D	E	F	G
							vehno	vehtype	speed	acc	occ	nla	ptime
1	100	57.7	0	0.03	1	00:56.7
							2	100	52.9	0	0.1	1	02:03.6
3	100	50	0	0.01	1	02:19.7
							4	100	48.3	0	0.03	1	02:24.4
5	200	51.6	-0.28	0.02	2	02:37.6
							6	100	52.2	0.18	0.07	1	03:32.2
7	100	51.1	0	0.02	2	04:27.3
							8	100	52.1	0	0.06	1	04:40.5

x_rSupplying attribute data for traffic, comprising: following model parameters (a-C in table 2), straight going split (gr1), left turn split (gr2), signal period (cycle), number of lanes (nlanes), and corresponding effective date (supplied _ id). x is the number of_rAs shown in table 2.

TABLE 2

A	B	C	D	E	F	G	H
								w74ax	w74add	w74mult	gr1	gr2	cycletime	nlanes	supply_id
2.3	2.3	1.5	0.260274	0.157534	146	1	0
								3.5	3.3	2.9	0.21978	0.214286	182	1	1
2.2	2.3	2.9	0.220779	0.201299	154	1	2
								2.9	3	2.9	0.238806	0.171642	134	1	3
1.4	2.9	3.8	0.23913	0.173913	138	2	4
								3.1	2.4	1.5	0.278562	0.12069	116	1	5

2. X is to be_tAnd x_rThe input flow estimation algorithm is trained by using Gaussian process regression to obtain the functional relation g ═ f (x) between the road flow and the input_r,x_t) I.e. to obtain a gaussian process model. That is, the parameters of the kernel function are obtained through the training and learning of the Gaussian process, and the obtained parameters of the kernel function are used for determiningA gaussian process model is determined, for example, by obtaining parameters l-2.31627714 e +03 and v-1.34450588 e-01 of the Matern kernel in one experiment, which parameters determine the gaussian process model.

3. Inputting first traffic supply attribute data to g ═ f (x)_r,x_t) And obtaining the road capacity and the estimated traffic demand through a Bayesian optimization algorithm.

For example, first traffic supply attribute data including the number of lanes, the straight-ahead split, the left-turn split, and the following model parameters are input, as shown in table 3, and the road capacity and the predicted traffic demand are output.

TABLE 3

A	B	C	D	E	F
						nlanes	gr1	gr2	w74ax	w74add	w74mult
1	0.260274	0.157534	2.3	2.3	1.5
						1	0.21978	0.214286	3.5	3.3	2.9
1	0.220779	0.201299	2.2	2.3	2.9

As can be seen from the above, in the first implementation manner, firstly, traffic supply and demand attribute data with high correlation with road traffic, which is obtained based on feature screening (such as correlation analysis, decision tree model establishment, and the like), is used as historical data (training data), then, a correlation function between road traffic and the quantity of traffic supply and demand attributes is fitted according to the historical data, that is, a gaussian process model is obtained through gaussian process modeling, and finally, the real-time traffic supply attribute data is used as input of the gaussian process model, and a maximum value is solved through a bayesian optimization method, so that dynamic solution of road capacity is achieved.

In the embodiment of the application, the real-time traffic state can be reflected by performing dynamic calculation based on the real-time traffic supply attribute data, no prior assumption is needed for the traffic flow operation rule, and the modeling of the traffic flow relation essence is performed by using a data driving mode, so that the modeling is more accurate. In addition, by using the characteristic engineering, namely the traffic state calculated based on the original data, the association between more traffic flow parameters and the flow is mined, so that the calculation result is more accurate.

In the first implementation mode, the relation between more characteristic parameters and the road flow is fitted through a gaussian process, dynamic characteristics of road section traffic flow changes can be captured well, the limitation of three parameters in the traditional method (namely the road flow, the vehicle density and the vehicle speed in a traffic flow basic diagram (FD) fitting method) is broken through, and essential relation modeling of road section traffic flow can be performed by more comprehensively using data-driven calculable characteristic attributes, such as traffic states (green signal ratio, driving behavior and other characteristics) with higher correlation with the traffic flow, obtained through correlation analysis.

In the embodiment of the application, the traffic center server inputs the first traffic supply attribute data and the first traffic demand attribute data into a trained neural network model and outputs the road capacity in the first time period, wherein the trained neural network model is used for acquiring the mapping relation between the road capacity and the traffic supply and demand attribute.

In an implementation manner of determining the road capacity through the neural network model, the traffic supply and demand attribute data in the first time period are all used as input data, and the road capacity in the first time period is directly output through the neural network model. In addition, for the determination process of the first traffic supply attribute data and the first traffic demand attribute data, reference may be made to the related description in the foregoing first implementation manner, and details are not described here again.

In the embodiment of the application, the traffic center server obtains multiple groups of second traffic data, the multiple groups of second traffic data are determined before a first time period, each group of second traffic data comprises traffic supply attribute data, traffic demand attribute data, road traffic and actual traffic time, the traffic center server takes each group of second traffic data and stored initial parameters as input of a neural network model to be trained, traffic time errors corresponding to each group of second traffic data are obtained, multiple traffic time errors are obtained, network parameters of the neural network model to be trained are adjusted through back propagation of the multiple traffic time errors, and the trained neural network model is obtained.

That is, the traffic supply and demand attribute data and the road flow in the historical data are used as input samples, the actual passing time in the historical data is used as a sample label, the network parameters are adjusted through back propagation based on the passing time error, and finally the neural network model capable of predicting the road capacity is obtained through training.

It should be noted that there are many implementations of training the neural network model in the embodiment of the present application, and one of them is described next.

In the embodiment of the application, the neural network model to be trained comprises an error correction module, a parameter calibration module and a traffic capacity estimation module, wherein the initial parameters comprise an initial value of a road resistance parameter and road section free running time, and the road resistance parameter is a parameter of a BPR function.

The traffic center server takes the traffic supply attribute data and the traffic demand attribute data included in each group of second traffic data as the input of the traffic capacity estimation module, outputs the predicted road capacity corresponding to each group of second traffic data, takes the initial value of the road resistance parameter as the input of the parameter calibration module, outputs the calibration value of the road resistance parameter, takes the road flow and the actual traffic time included in each group of second traffic data, and takes the predicted road capacity, the calibration value of the road resistance parameter and the road section free running time corresponding to each group of second traffic data as the input of the error correction module, and outputs the traffic time error corresponding to each group of second traffic data.

In the embodiment of the application, the traffic center server takes the road traffic included by each group of second traffic data, and the predicted road capacity, the calibration value of the road resistance parameter and the road section free driving time corresponding to each group of second traffic data as the input of the BPR function, outputs the predicted passing time corresponding to each group of second traffic data through the BPR function, takes the actual passing time included by each group of second traffic data and the predicted passing time corresponding to each group of second traffic data as the input of the loss function, and outputs the passing time error corresponding to each group of second traffic data through the loss function. It will be appreciated that the error correction module is constructed from a BPR function and a loss function from which the traffic centre server determines the transit time error for back propagation.

In the embodiment of the application, the traffic center server adjusts the network parameters of the traffic capacity estimation module and the parameter calibration module through the back propagation of the multiple traffic time errors to obtain the neural network model. That is, the network parameters included in the traffic capacity estimation module and the parameter calibration module require error reverse iterative adjustment.

Optionally, the trafficability estimation module includes a plurality of hidden layers, network parameters (such as weight w and bias b) included in the plurality of hidden layers need to be adjusted by error backward iteration, and the parameter calibration module includes a parameter correction unit, and the network parameters included in the parameter correction unit also need to be adjusted by error backward iteration.

Exemplarily, the parameter modification module comprises a first parameter modification unit and a second parameter modification unit, and the first parameter modification unit comprises the network parameter w to be adjusted_αThe second parameter modification unit comprises a network parameter w to be adjusted_βWherein w is_αFor calculating the road resistance parameter alpha, w_βFor calculating the road resistance parameter beta.

In the embodiment of the application, after the training process, a trained neural network model is obtained, the neural network model can predict the road capacity, the neural network model comprises a traffic capacity estimation module, and the traffic center server inputs the first traffic supply attribute data and the first traffic demand attribute data into the traffic capacity estimation module included in the trained neural network model and outputs the road capacity in the first time period.

In the embodiment of the application, the traffic center server can also update the neural network model according to the continuously received traffic flow data and the continuously updated road attribute data.

Optionally, in this embodiment of the application, the traffic center server may further be configured to obtain a road traffic and an actual transit time in the first time period, and update the neural network model according to the first traffic supply attribute data, the first traffic demand attribute data, and the road traffic and the actual transit time in the first time period.

Optionally, there are many implementation manners for determining the road traffic in the first time period by the traffic center server, and reference is made to the related description in the foregoing first implementation manner, which is not described herein again, and in the second implementation manner, only the implementation manner for determining the actual transit time in the first time period by the traffic center server is described.

Illustratively, several implementations of determining the actual transit time include, but are not limited to: recording the passing time of two adjacent electric police checkpoints on the road, and calculating the time difference to obtain the actual passing time in the first time period; or recording the time of the vehicle at the starting positioning point and the ending positioning point of the road, and calculating the time difference to obtain the actual passing time in the first time period.

It should be noted that, when the neural network model is updated by the first traffic supply attribute data and the first traffic demand attribute data, the first traffic supply attribute data and the first traffic demand attribute data are regarded as data included in the history data, and the first time period is regarded as a history time period, that is, the neural network model is updated by the history data.

Fig. 5 is a schematic diagram of a neural network model to be trained according to an embodiment of the present application, and the process of training the neural network model described above will be described again with reference to fig. 5.

In fig. 5, the neural network model to be trained includes an error correction module, a parameter calibration module, and a traffic capacity estimation module.

Wherein the traffic capacity estimation module is a neural network with multiple hidden layers, and the input of the traffic capacity estimation module comprises traffic supply attribute data x_rAnd traffic demand attribute data x_tThe output includes the predicted link capacity C'.

Optionally, using the formula C ═ f_C(x_r,x_t) And explaining the complex nonlinear mapping relation characterized by the traffic capacity estimation module.

The parameter calibration module is used for correcting the parameters of the road resistance function. The parameter calibration module comprises a first parameter correction unit and a second parameter correction unit, wherein the input of the first parameter correction unit comprises a road resistance parameter alpha, the output of the first parameter correction unit comprises a calibration value alpha 'of the alpha, the input of the second parameter correction unit comprises a road resistance parameter beta, and the output of the second parameter correction unit comprises a calibration value beta' of the beta.

Optionally, the first parameter modification unit and the second parameter modification unit both use a weight to adjust the road resistance parameter (i.e. connected through the neural network with the offset b). The first parameter correcting unit and the second parameter correcting unit use f_α(α)＝w_αA and f_β(β)＝w_βBeta explains the respective characterized mapping.

And the error correction module is used for generating a transit time error in each iteration in the training process. The error correction module is constructed according to a BPR function and a loss function, wherein the input of the BPR function comprises free running time t of a road section₀The road flow v, the predicted road capacity C ', the calibration value α ' and the calibration value β ', the output comprising the predicted transit time t ', the input of the loss function comprising the actual transit time t and the predicted transit time t ', the output comprising the transit time error Δ t (not shown).

Optionally, the Loss function ═ Loss (t', t). In the embodiment of the present application, there are many methods for calculating the error, and the form of the loss function varies according to the method, for example, when the transit time error is calculated by using the mean square error method, the loss function is shown in formula (10).

From the above, the neural networks of the traffic capacity estimation module and the parameter calibration module have no training label, and the two modules are combined with the road resistance function included in the error correction module to obtain an observable label (road traffic time). The transit time error obtained by each iteration is used for back propagation so as to adjust the network parameters of the whole neural network model, including adjusting the network parameters of the transit capacity estimation module and the parameter calibration module.

By the method, the functional relation t (f) (x) between the BPR function and the traffic supply and demand attribute is obtained after the neural network model to be trained is trained, namely the trained neural network model can describe the mapping relation between the road traffic time and the traffic supply and demand attribute, and the trained traffic capacity estimation module can be used for obtaining the mapping relation between the road capacity and the traffic supply and demand attribute. Based on this, after the training, the neural network model actually used for predicting the road capacity may include only the trained traffic capacity estimation module, and exemplarily, the first traffic supply attribute data and the first traffic demand attribute data are input into the trained traffic capacity estimation module, and the road capacity is output.

In the second implementation manner, the traffic center server can perform dynamic calculation based on real-time data, and can reflect a real-time traffic state. It should be noted that, because the road traffic time is easy to collect, that is, the road traffic time is a traffic characteristic parameter that can be directly measured, the road traffic time is easy to construct a loss function, and end-to-end training is easy to implement. The road traffic time is used as the sub-output of the BPR function, and the road capacity is used as the sub-output of the neural network model to be trained, so that the model training process is simplified. The required output road capacity is further converted through the BPR function, namely, the input, the output and the calculation process of the whole road resistance function are regarded as a large neural network, the traffic capacity estimation module is independently made into a small sub-network through the splitting of the road resistance function, and the road traffic time is regarded as the final output of the neural network model to be trained. Capturing a complex nonlinear mapping relation between various influencing factors (traffic supply and demand attributes) and road capacity through a neural network, combining a training process of the neural network with parameter calibration of a road resistance function, and synchronously updating parameters (weight w and bias b) of the neural network and parameters (alpha and beta) of the road resistance function through a back propagation algorithm.

It should be noted that, in the embodiment of the present application, the neural network model for determining the road capacity may be regarded as a prediction regression model, and the neural network model may be constructed in various manners, for example, in the embodiment of the present application, the neural network model is constructed in a manner of multi-layer perception (MLP), and in other embodiments, the neural network model may also be constructed in a manner of logistic regression, random forest, ridge regression, and the like.

Fig. 6 is a flowchart of another road capacity determination method according to an embodiment of the present application. Referring to fig. 6, the process of determining the road capacity by the traffic center server includes data acquisition, training learning and dynamic calculation of the road capacity.

The data acquisition process comprises the steps of acquiring traffic flow data, obtaining traffic demand attribute data, obtaining road attribute data and obtaining traffic supply attribute data, the data acquisition module takes the traffic demand attribute data and the traffic supply attribute data before a determined first time period as historical data, and takes the traffic demand attribute data and the traffic supply attribute data in the determined first time period as real-time data. The process of training learning includes training a model according to the historical data by a model training algorithm to obtain a machine learning model, such as the gaussian process model or the neural network model described above. The road capacity dynamic calculation process comprises the steps of inputting real-time data into a machine learning model, and calculating the road capacity through the machine learning model.

It should be noted that, in the embodiment of the present application, the mapping relationship between the road capacity and the traffic supply and demand attribute is implicitly described through the machine learning model, and optionally, after the machine learning model capable of predicting the road capacity is obtained through the above method, the road capacity under any given constraint condition is predetermined and stored according to the machine learning model, that is, the mapping relationship between the road capacity and the traffic supply and demand attribute is explicitly stored, and the traffic center server predicts the road capacity according to the explicitly stored mapping relationship.

According to the embodiment of the application, the incidence relation between the road capacity and the traffic supply and demand attribute is mined from various traffic supply attributes and traffic demand attributes, and the calculation precision and accuracy of the road capacity are improved.

The method for determining the road capacity can predict the road capacity in real time, the predicted road capacity can be used for traffic real-time control and the like, and important bases are provided for intelligent traffic, an intelligent transportation system ITS and the like. Further, road saturation, intersection traffic capacity, congestion index, and the like can be further calculated from the predicted road capacity.

In summary, in the embodiment of the present application, when determining the road capacity in the first time period, the traffic supply attribute data and the traffic demand attribute data in the first time period are obtained, on one hand, if the obtained data is real-time data, the scheme may determine the road capacity in real time, that is, the real-time performance is higher, and on the other hand, if the obtained data is traffic data in a certain area and a certain time period, the scheme may determine the road capacity in the certain time period in the certain area, that is, the scheme may adapt to differences of different areas and different time periods. In addition, the data acquired by the scheme does not depend on manual measurement and historical experience, the accuracy of the determined road capacity is higher, and the labor cost is very low.

Fig. 7 is a schematic structural diagram of a road capacity determining apparatus 700 provided in an embodiment of the present application, where the road capacity determining apparatus 700 may be implemented by software, hardware, or a combination of the two as part or all of a computer device, and the computer device may be the traffic center server shown in fig. 1 or fig. 2. Referring to fig. 7, the apparatus 700 includes: a first obtaining module 701 and a determining module 702.

A first obtaining module 701, configured to obtain first traffic supply attribute data and first traffic demand attribute data, where the first traffic supply attribute data is used to represent a link attribute in a first time period, and the first traffic demand attribute data is used to represent a traffic flow attribute in the first time period; for a specific implementation, please refer to the detailed description of step 401 in the embodiment of fig. 4, which is not described herein again.

A determining module 702, configured to determine road capacity in a first time period according to a mapping relationship between road capacity and a traffic demand attribute according to the first traffic supply attribute data and the first traffic demand attribute data, where the mapping relationship is determined according to historically acquired traffic supply attribute data, traffic demand attribute data, and road traffic, and the traffic demand attribute includes a traffic supply attribute and a traffic demand attribute. For a specific implementation, please refer to the detailed description of step 402 in the embodiment of fig. 4, which is not described herein again.

Optionally, the determining module 702 includes:

and the first determining submodule is used for determining the maximum value of the first supply-demand relation function, and the maximum value is used as the road capacity in the first time period. For a detailed implementation, please refer to the detailed description of the first implementation of step 402 in the embodiment of fig. 4, which is not repeated here.

Optionally, the first determining sub-module is configured to:

and determining the maximum value of the first supply-demand relation function through Bayesian optimization processing, and taking the maximum value as the road capacity in the first time period. For a detailed implementation, please refer to the detailed description of the first implementation of step 402 in the embodiment of fig. 4, which is not repeated here.

Optionally, the determining module 702 includes:

and the second determining submodule is used for inputting the first traffic supply attribute data and the first traffic demand attribute data into a Gaussian process model and outputting the road capacity in the first time period, and the Gaussian process model is used for describing the mapping relation between the road capacity and the traffic supply and demand attribute. For a detailed implementation, please refer to the detailed description of the first implementation of step 402 in the embodiment of fig. 4, which is not repeated here.

Optionally, the apparatus 700 further comprises:

the first learning module is used for taking the traffic supply attribute data and the traffic demand attribute data in each group of first traffic data as input data, taking the road flow in each group of first traffic data as output data, and obtaining a Gaussian process model through Gaussian process fitting. For a detailed implementation, please refer to the detailed description of the first implementation of step 402 in the embodiment of fig. 4, which is not repeated here.

Optionally, the apparatus 700 further comprises:

and the first updating module is used for updating the Gaussian process model according to the first traffic supply attribute data, the first traffic demand attribute data and the road flow in the first time period. For a detailed implementation, please refer to the detailed description of the first implementation of step 402 in the embodiment of fig. 4, which is not repeated here.

Optionally, the determining module 702 includes:

and the third determining submodule is used for inputting the first traffic supply attribute data and the first traffic demand attribute data into the trained neural network model and outputting the road capacity in the first time period, and the trained neural network model is used for acquiring the mapping relation between the road capacity and the traffic supply and demand attribute. For a detailed implementation, please refer to the detailed description of the second implementation manner of step 402 in the embodiment of fig. 4, which is not repeated here.

Optionally, the apparatus 700 further comprises:

and the third learning module is used for adjusting the network parameters of the neural network model to be trained through the back propagation of the plurality of passing time errors to obtain the trained neural network model. For a detailed implementation, please refer to the detailed description of the second implementation manner of step 402 in the embodiment of fig. 4, which is not repeated here.

the second learning module includes:

and the third training submodule is used for taking the road flow and the actual passing time included by each group of second traffic data, and the predicted road capacity, the calibration value of the road resistance parameter and the free running time of the road section corresponding to each group of second traffic data as the input of the error correction module, and outputting the passing time error corresponding to each group of second traffic data. For a detailed implementation, please refer to the detailed description of the second implementation manner of step 402 in the embodiment of fig. 4, which is not repeated here.

Optionally, the third training submodule is configured to:

and taking the actual passing time included by each group of second traffic data and the predicted passing time corresponding to each group of second traffic data as the input of a loss function, and outputting the passing time error corresponding to each group of second traffic data through the loss function. For a detailed implementation, please refer to the detailed description of the second implementation manner of step 402 in the embodiment of fig. 4, which is not repeated here.

Optionally, the third learning module comprises:

the third determination submodule is configured to:

and inputting the first traffic supply attribute data and the first traffic demand attribute data into a traffic capacity estimation module included in the trained neural network model, and outputting the road capacity in the first time period. For a detailed implementation, please refer to the detailed description of the second implementation manner of step 402 in the embodiment of fig. 4, which is not repeated here.

Optionally, the apparatus 700 further comprises:

and the second updating module is used for updating the neural network model according to the first traffic supply attribute data, the first traffic demand attribute data, and the road flow and the actual passing time in the first time period. For a detailed implementation, please refer to the detailed description of the second implementation manner of step 402 in the embodiment of fig. 4, which is not repeated here.

Optionally, the first obtaining module 701 includes:

the receiving submodule is used for receiving traffic flow data sent by a data acquisition side and/or an edge side in a first time period, the data acquisition side is used for acquiring traffic flow data in real time, and the edge side is used for collecting the traffic flow data acquired by the data acquisition side in real time and preprocessing the traffic flow data; for a specific implementation, please refer to the detailed description of step 401 in the embodiment of fig. 4, which is not described herein again.

Optionally, the first obtaining module 701 includes:

And the second acquisition submodule is used for acquiring the first traffic supply attribute data from the traffic management server. For a specific implementation, please refer to the detailed description of step 401 in the embodiment of fig. 4, which is not described herein again.

the first traffic demand attribute data includes one or more of a vehicle type, a vehicle speed, a vehicle acceleration, a time occupancy, and a lane number. For a specific implementation, please refer to the detailed description of step 401 in the embodiment of fig. 4, which is not described herein again.

It should be noted that: the device for determining road capacity provided in the above embodiment is only illustrated by dividing the above functional modules when determining the road capacity, and in practical applications, the above function allocation may be completed by different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules to complete all or part of the above described functions. In addition, the determination apparatus for the road capacity and the determination method for the road capacity provided by the above embodiments belong to the same concept, and specific implementation processes thereof are detailed in the method embodiments and are not described herein again.

In the above embodiments, the implementation may be wholly or partly realized by software, hardware, firmware, or any combination thereof. When implemented in software, may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When loaded and executed on a computer, cause the processes or functions described in accordance with the embodiments of the application to occur, in whole or in part. The computer may be a general purpose computer, a special purpose computer, a network of computers, or other programmable device. The computer instructions may be stored on a computer readable storage medium or transmitted from one computer readable storage medium to another, for example, from one website, computer, server, or data center to another website, computer, server, or data center via wire (e.g., coaxial cable, fiber optic, Digital Subscriber Line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device, such as a server, a data center, etc., that includes one or more of the available media. The usable medium may be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., Digital Versatile Disk (DVD)), or a semiconductor medium (e.g., Solid State Disk (SSD)), among others. It is noted that the computer-readable storage medium referred to in the embodiments of the present application may be a non-volatile storage medium, in other words, a non-transitory storage medium.

It is to be understood that reference herein to "at least one" means one or more and "a plurality" means two or more. In the description of the embodiments of the present application, "/" means "or" unless otherwise specified, for example, a/B may mean a or B; "and/or" herein is merely an association describing an associated object, and means that there may be three relationships, e.g., a and/or B, which may mean: a exists alone, A and B exist simultaneously, and B exists alone. In addition, in order to facilitate clear description of technical solutions of the embodiments of the present application, in the embodiments of the present application, terms such as "first" and "second" are used to distinguish the same items or similar items having substantially the same functions and actions. Those skilled in the art will appreciate that the terms "first," "second," etc. do not denote any order or quantity, nor do the terms "first," "second," etc. denote any order or importance.

The above description is provided for illustrative embodiments of the present application and not for the purpose of limiting the present application, and any modifications, equivalents, improvements, etc. made within the spirit and principle of the present application should be included in the protection scope of the present application.

Claims

1. A method of determining road capacity, the method comprising:

acquiring first traffic supply attribute data and first traffic demand attribute data, wherein the first traffic supply attribute data are used for representing road section attributes in a first time period, and the first traffic demand attribute data are used for representing traffic flow attributes in the first time period;

determining the road capacity in the first time period by using a Gaussian process model or a trained neural network model according to the first traffic supply attribute data and the first traffic demand attribute data, wherein the Gaussian process model and the trained neural network model are both used for describing a mapping relation between the road capacity and the traffic supply and demand attribute, the mapping relation is determined according to the traffic supply attribute data, the traffic demand attribute data and the road flow acquired historically, and the traffic supply and demand attribute comprises the traffic supply attribute and the traffic demand attribute;

if the trained neural network model is used to determine the road capacity in the first time period, the method further comprises:

acquiring multiple groups of second traffic data, wherein the multiple groups of second traffic data are determined before the first time period, and each group of second traffic data comprises traffic supply attribute data, traffic demand attribute data, road flow and actual passing time;

taking each group of second traffic data and the stored initial parameters as the input of a neural network model to be trained, and obtaining a plurality of passing time errors corresponding to each group of second traffic data; and adjusting the network parameters of the neural network model to be trained through the back propagation of the plurality of transit time errors to obtain the trained neural network model.

2. The method of claim 1, wherein determining the road capacity within the first time period using a gaussian process model based on the first traffic supply attribute data and the first traffic demand attribute data comprises:

inputting at least one of the first traffic supply attribute data and the first traffic demand attribute data into the Gaussian process model, and outputting a first supply-demand relation function;

and determining the maximum value of the first supply-demand relation function, and taking the maximum value as the road capacity in the first time period.

3. The method of claim 2, wherein the determining a maximum value of the first supply-demand relationship function, the maximum value being taken as the road capacity in the first time period, comprises:

and determining the maximum value of the first supply-demand relation function through Bayesian optimization, and taking the maximum value as the road capacity in the first time period.

4. The method of claim 1, wherein determining the road capacity within the first time period using a gaussian process model based on the first traffic supply attribute data and the first traffic demand attribute data comprises:

inputting the first traffic supply attribute data and the first traffic demand attribute data into the Gaussian process model, and outputting the road capacity in the first time period.

5. The method of any of claims 2-4, wherein the method further comprises:

acquiring multiple groups of first traffic data, wherein the multiple groups of first traffic data are determined before the first time period, and each group of first traffic data comprises traffic supply attribute data, traffic demand attribute data and road flow;

and taking the traffic supply attribute data and the traffic demand attribute data in each group of first traffic data as input data, taking the road flow in each group of first traffic data as output data, and obtaining the Gaussian process model through Gaussian process fitting.

6. The method of claim 5, wherein after obtaining the Gaussian process model by Gaussian process fitting, further comprising:

acquiring the road flow in the first time period;

and updating the Gaussian process model according to the first traffic supply attribute data, the first traffic demand attribute data and the road flow in the first time period.

7. The method of claim 1, wherein determining the road capacity over the first time period using a trained neural network model from the first traffic supply attribute data and the first traffic demand attribute data comprises:

and inputting the first traffic supply attribute data and the first traffic demand attribute data into the trained neural network model, and outputting the road capacity in the first time period, wherein the trained neural network model is used for acquiring the mapping relation between the road capacity and the traffic supply and demand attribute.

8. The method of claim 1, wherein the neural network model to be trained comprises an error correction module, a parameter calibration module and a traffic capacity estimation module, the initial parameters comprise an initial value of a road resistance parameter and a road section free-running time, and the road resistance parameter is a parameter of a road bureau BPR function;

the step of obtaining the corresponding transit time error of each group of second traffic data by taking each group of second traffic data and the stored initial parameters as the input of the neural network model to be trained to obtain a plurality of transit time errors comprises the following steps:

taking the traffic supply attribute data and the traffic demand attribute data included in each group of second traffic data as the input of the traffic capacity estimation module, and outputting the predicted road capacity corresponding to each group of second traffic data;

taking the initial value of the road resistance parameter as the input of the parameter calibration module, and outputting the calibration value of the road resistance parameter;

and taking the road flow and the actual passing time included by each group of second traffic data, and the predicted road capacity, the calibration value of the road resistance parameter and the free running time of the road section corresponding to each group of second traffic data as the input of the error correction module, and outputting the passing time error corresponding to each group of second traffic data.

9. The method of claim 8, wherein the taking as input of the error correction module the road traffic and the actual transit time included in each set of second traffic data, and the predicted road capacity, the calibrated value of the road resistance parameter, and the road segment free-running time corresponding to each set of second traffic data, and outputting the transit time error corresponding to each set of second traffic data comprises:

taking the road flow included by each group of second traffic data, and the predicted road capacity, the calibration value of the road resistance parameter and the free running time of the road section corresponding to each group of second traffic data as the input of the BPR function, and outputting the predicted passing time corresponding to each group of second traffic data through the BPR function;

10. The method of claim 8, wherein adjusting network parameters of the neural network model to be trained by the plurality of back propagation of transit time errors to obtain the trained neural network model comprises:

and adjusting the network parameters of the traffic capacity estimation module and the parameter calibration module through the back propagation of the plurality of traffic time errors to obtain the trained neural network model.

11. The method of claim 9, wherein adjusting network parameters of the neural network model to be trained by the plurality of back propagation of transit time errors to obtain the trained neural network model comprises:

12. The method of claim 7, wherein the trained neural network model comprises the capacity estimation module;

the inputting the first traffic supply attribute data and the first traffic demand attribute data into a trained neural network model and outputting the road capacity in the first time period comprises:

inputting the first traffic supply attribute data and the first traffic demand attribute data into a traffic capacity estimation module included in the trained neural network model, and outputting the road capacity in the first time period.

13. The method of claim 1, wherein after obtaining the trained neural network model, further comprising:

acquiring road flow and actual passing time in a first time period;

and updating the neural network model according to the first traffic supply attribute data, the first traffic demand attribute data, and the road flow and the actual transit time in the first time period.

14. The method of any one of claims 8-11, wherein after obtaining the trained neural network model, further comprising:

acquiring road flow and actual passing time in a first time period;

15. The method of claim 12, wherein after obtaining the trained neural network model, further comprising:

acquiring road flow and actual passing time in a first time period;

16. The method of any of claims 1-4 and 8-11, wherein obtaining first traffic demand attribute data comprises:

receiving traffic flow data sent by a data acquisition side and/or an edge side in the first time period, wherein the data acquisition side is used for acquiring traffic flow data in real time, and the edge side is used for collecting and preprocessing the traffic flow data acquired by the data acquisition side in real time;

and determining the first traffic demand attribute data according to the received traffic flow data.

17. The method of claim 16, wherein the data collection side comprises one or more traffic sensors, the edge side comprises one or more edge devices, each of the one or more traffic sensors is any one of a sensor of an electric police station, a sensor of a geomagnetic coil/radar, a Global Navigation Satellite System (GNSS) sensor in a vehicle, and a positioning sensor of a mobile device in a vehicle.

18. The method of any of claims 1-4 and 8-11, wherein obtaining first traffic supply attribute data comprises:

acquiring the first traffic supply attribute data from the stored road attribute data; and/or

And acquiring the first traffic supply attribute data from a traffic management server.

19. The method of any of claims 1-4 and 8-11, wherein the first traffic supply attribute data includes one or more of a following model parameter, a straight-ahead split, a left-turn split, a signal period, and a number of lanes;

20. An apparatus for determining road capacity, the apparatus comprising:

a determining module, configured to determine road capacity in the first time period by using a gaussian process model or a trained neural network model according to the first traffic supply attribute data and the first traffic demand attribute data, where the gaussian process model and the trained neural network model are both used to describe a mapping relationship between road capacity and a traffic supply and demand attribute, the mapping relationship is determined according to traffic supply attribute data, traffic demand attribute data, and road traffic obtained historically, and the traffic supply and demand attribute includes the traffic supply attribute and the traffic demand attribute;

if the trained neural network model is used to determine the road capacity in the first time period, the apparatus further comprises:

the second learning module is used for taking each group of second traffic data and the stored initial parameters as the input of the neural network model to be trained, obtaining the passing time errors corresponding to each group of second traffic data, and obtaining a plurality of passing time errors;

21. The apparatus of claim 20, wherein the determining module comprises:

the processing submodule is used for inputting at least one of the first traffic supply attribute data and the first traffic demand attribute data into the Gaussian process model and outputting a first supply-demand relation function, and the Gaussian process model is used for describing a mapping relation between the road capacity and the traffic supply-demand attribute;

and the first determining submodule is used for determining the maximum value of the first supply-demand relation function, and taking the maximum value as the road capacity in the first time period.

22. The apparatus of claim 21, wherein the first determination submodule is to:

23. The apparatus of claim 20, wherein the determining module comprises:

and the second determining submodule is used for inputting the first traffic supply attribute data and the first traffic demand attribute data into the Gaussian process model and outputting the road capacity in the first time period, and the Gaussian process model is used for describing the mapping relation between the road capacity and the traffic supply and demand attribute.

24. The apparatus of any of claims 21-23, wherein the apparatus further comprises:

a second obtaining module, configured to obtain multiple sets of first traffic data, where the multiple sets of first traffic data are determined before the first time period, and each set of first traffic data includes traffic supply attribute data, traffic demand attribute data, and road traffic;

and the first learning module is used for taking the traffic supply attribute data and the traffic demand attribute data in each group of first traffic data as input data, taking the road flow in each group of first traffic data as output data, and obtaining the Gaussian process model through Gaussian process fitting.

25. The apparatus of claim 24, wherein the apparatus further comprises:

26. The apparatus of claim 20, wherein the determining module comprises:

27. The apparatus of claim 20, wherein the neural network model to be trained comprises an error correction module, a parameter calibration module and a traffic capacity estimation module, the initial parameters comprise an initial value of a road resistance parameter and a road segment free-running time, and the road resistance parameter is a parameter of a road bureau BPR function;

the second learning module includes:

and the third training submodule is used for taking the road flow and the actual passing time included by each group of second traffic data, the predicted road capacity corresponding to each group of second traffic data, the calibration value of the road resistance parameter and the free running time of the road section as the input of the error correction module, and outputting the passing time error corresponding to each group of second traffic data.

28. The apparatus of claim 27, wherein the third training submodule is to:

29. The apparatus of claim 27, wherein the third learning module comprises:

and the fourth training submodule is used for adjusting the network parameters of the traffic capacity estimation module and the parameter calibration module through the back propagation of the plurality of traffic time errors to obtain the trained neural network model.

30. The apparatus of claim 28, wherein the third learning module comprises:

31. The apparatus of claim 26, wherein the trained neural network model comprises the capacity estimation module;

the third determination submodule is configured to:

32. The apparatus of claim 20, wherein the apparatus further comprises:

33. The apparatus of any of claims 27-30, wherein the apparatus further comprises:

34. The apparatus of claim 31, wherein the apparatus further comprises:

35. The apparatus of any of claims 20-23, 27-30, wherein the first obtaining module comprises:

the receiving submodule is used for receiving traffic flow data sent by a data acquisition side and/or an edge side in the first time period, the data acquisition side is used for acquiring traffic flow data in real time, and the edge side is used for collecting the traffic flow data acquired by the data acquisition side in real time and preprocessing the traffic flow data;

36. The apparatus of claim 35, wherein the data collection side comprises one or more traffic sensors, the edge side comprises one or more edge devices, each of the one or more traffic sensors is any one of a sensor of an electric police gate, a sensor of a geomagnetic coil/radar, a Global Navigation Satellite System (GNSS) sensor in a vehicle, and a positioning sensor of a mobile device in a vehicle.

37. The apparatus of any of claims 20-23, 27-30, wherein the first obtaining module comprises:

a first acquisition sub-module configured to acquire the first traffic supply attribute data from the stored road attribute data; and/or

38. The apparatus of any of claims 20-23, 27-30, wherein the first traffic supply attribute data comprises one or more of a following model parameter, a straight-ahead split, a left-turn split, a signal period, and a number of lanes;

39. A computer-readable storage medium, in which a computer program is stored which, when being executed by a processor, carries out the method of any one of claims 1 to 19.