Traffic flow estimation method and device, electronic equipment and storage medium

Abstract

This application discloses a traffic flow estimation method, apparatus, electronic device, and storage medium. The method includes: acquiring trajectory data of N connected vehicles at an intersection during the current traffic signal cycle; reconstructing the intersection shockwave based on the trajectory data of the N connected vehicles to obtain the maximum possible queue length of the intersection; calculating a probability compensation value based on the maximum possible queue length and the distance from the stopping point of the last connected vehicle among the N connected vehicles to the stop line of the intersection; estimating the queue length of the intersection based on the probability compensation value and the distance from the stopping point of the last connected vehicle among the N connected vehicles to the stop line of the intersection; and performing estimation processing based on the queue length of the intersection and the time it takes for the N connected vehicles to reach the stop line of the intersection to obtain the periodic traffic flow of the intersection during the current traffic signal cycle.

Description

Technical Field

[0001] This application relates to the field of traffic control technology, and in particular to traffic flow estimation methods, devices, electronic equipment, and storage media. Background Technology

[0002] In recent years, the development of Intelligent Transportation Systems (ITS) and Connected Vehicle (CV) technologies has provided tremendous opportunities for traffic volume estimation. In the past few years, researchers have utilized loop detector information and GPS data from CVs to estimate traffic volume. However, the installation and reliability of loop detectors limit the application and accuracy of this traffic volume estimation method. Therefore, how to accurately and efficiently estimate traffic flow using connected vehicle technology has become one of the current hot research topics. Summary of the Invention

[0003] This application provides a traffic flow estimation method, apparatus, electronic device, and storage medium, which can achieve accurate and efficient traffic flow estimation based on the trajectory data of networked vehicles.

[0004] In a first aspect, embodiments of this application provide a traffic flow estimation method, including:

[0005] The trajectory data of N connected vehicles at the intersection during the current traffic signal cycle are obtained, and the intersection shock wave is reconstructed based on the trajectory data of the N connected vehicles to obtain the maximum possible queue length of the intersection; N is an integer greater than or equal to 1.

[0006] The probability compensation value is calculated based on the maximum possible queue length and the distance from the parking point of the last connected vehicle among the N connected vehicles to the intersection parking line;

[0007] Based on the probability compensation value and the distance from the parking point of the last connected vehicle among the N connected vehicles to the parking line of the intersection, the queue length of the intersection is estimated.

[0008] The periodic traffic flow of the intersection during the current traffic signal cycle is obtained by estimating the queue length at the intersection and the arrival time of the N networked vehicles at the stop line at the intersection.

[0009] Secondly, embodiments of this application provide a traffic estimation method, including:

[0010] Obtain the traffic flow of P traffic signal cycles, wherein the traffic flow of each traffic signal cycle is obtained by the method in the first aspect.

[0011] Based on the P cycles of traffic flow, the periodic traffic arrival rate is estimated;

[0012] Based on the theory that periodic traffic flow follows a Poisson distribution, and using the traffic arrival rate, the traffic flow during the future target time period is estimated within the traffic signal period for which no connected vehicles were detected.

[0013] Thirdly, embodiments of this application provide a traffic flow estimation device, comprising:

[0014] The acquisition unit is used to acquire trajectory data of N connected vehicles at the intersection during the current traffic signal cycle, and reconstruct the intersection shock wave based on the trajectory data of the N connected vehicles to obtain the maximum possible queue length of the intersection; N is an integer greater than or equal to 1.

[0015] The calculation unit is used to calculate the probability compensation value based on the maximum possible queue length and the distance from the parking point of the last connected vehicle among the N connected vehicles to the intersection parking line;

[0016] An estimation unit is used to estimate the queue length at the intersection based on the probability compensation value and the distance from the parking point of the last connected vehicle among the N connected vehicles to the parking line at the intersection.

[0017] The processing unit is used to perform estimation processing based on the queue length of the intersection and the time when the N networked vehicles arrive at the stop line of the intersection to obtain the periodic traffic flow of the intersection in the current traffic signal cycle.

[0018] Fourthly, embodiments of this application provide a traffic flow estimation device, comprising:

[0019] The acquisition unit is used to acquire the traffic flow of P traffic signal cycles, wherein the traffic flow of each traffic signal cycle is determined by the method according to any one of claims 1-10; P is an integer greater than or equal to 1.

[0020] An estimation unit is used to estimate the periodic-level traffic arrival rate based on the P periods of traffic flow.

[0021] The estimation unit is also used to estimate the traffic flow in the future target time period within the traffic signal period for which no connected vehicles were detected, based on the traffic arrival rate, according to the theory that periodic traffic flow follows a Poisson distribution.

[0022] Fifthly, embodiments of this application provide an electronic device, including:

[0023] A processor suitable for implementing one or more computer programs; a computer storage medium storing one or more computer programs, the one or more computer programs being adapted to be loaded by the processor and executed by the aforementioned traffic flow estimation method of the first aspect and the traffic flow estimation method of the second aspect.

[0024] In a sixth aspect, embodiments of this application provide a computer storage medium storing a computer program, which, when executed by a processor of an electronic device, performs the traffic flow estimation method of the first aspect and the traffic flow estimation method of the second aspect.

[0025] In a seventh aspect, embodiments of this application provide a computer program product or a computer program, the computer program product including a computer program, which may refer to a computer program, the computer program being stored in a computer storage medium; a processor reading the computer program from the computer storage medium, the processor executing the computer program, causing an electronic device to execute the traffic flow estimation method of the first aspect and the traffic flow estimation method of the second aspect described above.

[0026] In this embodiment, for the case where connected vehicles are detected within the current traffic signal cycle, shockwave reconstruction is performed on the intersection based on the trajectory data of the detected N connected vehicles to obtain the maximum possible queue length. Then, a probability compensation value is calculated based on the maximum possible queue length and the distance from the stopping point of the last connected vehicle among the N connected vehicles to the intersection stop line. Furthermore, the queue length of the intersection is calculated based on the probability compensation value and the distance from the stopping point of the last connected vehicle to the intersection stop line. Finally, an estimation process is performed based on the queue length of the intersection and the arrival time of the N connected vehicles at the intersection stop line to obtain the periodic traffic flow of the intersection within the current traffic signal cycle. It can be seen that the above traffic flow estimation process utilizes the trajectory data of connected vehicles without using loop detectors, overcoming the limitation of existing technologies that rely on loop detectors, thus affecting estimation accuracy. This achieves traffic flow estimation using sparse connected vehicles while ensuring the accuracy of the estimated flow.

[0027] For traffic signal cycles in which no connected vehicles are detected, this application uses a Bayesian method to estimate traffic flow based on estimated periodic traffic flow from multiple historical traffic signal cycles. Existing technologies cannot estimate traffic flow when no connected vehicles are detected. Compared with existing technologies, this application overcomes the dependence of traffic flow estimation on whether connected vehicles are detected within a cycle, and can be applied to traffic flow estimation in more scenarios. Attached Figure Description

[0028] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0029] Figure 1 This is an overall framework diagram of traffic flow estimation provided in an embodiment of this application;

[0030] Figure 2 This is a schematic flowchart of a traffic flow estimation method provided in an embodiment of this application;

[0031] Figure 3a This application provides a spatiotemporal map of the trajectory of a networked vehicle when N=1.

[0032] Figure 3b This application provides a spatiotemporal map of the trajectory of a networked vehicle when N is greater than 1.

[0033] Figure 3c This application provides a spatiotemporal map of the trajectory of a networked vehicle when N equals 2 and M equals 1.

[0034] Figure 3d This application provides a spatiotemporal map of the trajectory of a networked vehicle when N is greater than or equal to 2 and M is greater than 1.

[0035] Figure 4 This is a flowchart of a shock wave correction process provided in an embodiment of this application;

[0036] Figure 5 This is a schematic diagram of another traffic flow estimation method provided in an embodiment of this application;

[0037] Figure 6a This application provides an embodiment of an intersection and a vehicle arrival model;

[0038] Figure 6b This is a schematic diagram illustrating the identification result of a time limit provided in an embodiment of this application;

[0039] Figure 6c This is the posterior distribution of traffic flow λ in the two stages provided in the embodiments of this application;

[0040] Figure 6d and Figure 6e The posterior distributions of southbound and northbound traffic arrival rates at intersections 2 and 3 are shown in the figure.

[0041] Figure 7 This is a schematic diagram of the structure of a traffic flow estimation device provided in an embodiment of this application;

[0042] Figure 8 This is a schematic diagram of another traffic flow estimation device provided in the embodiments of this application;

[0043] Figure 9 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0044] The technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. It should be noted that all information related to the object in social applications, such as avatars, virtual images, and names, in the following description of this application has been obtained with the permission of the respective object.

[0045] The emergence of connected vehicle (CV) technology has created new opportunities for traffic control. However, the amount of connected vehicle trajectory data is severely limited by the penetration rate of connected vehicles across different geographical locations and times. The goal of this study is to overcome these limitations and allow traffic flow estimation in any location or under any conditions, including low CV penetration.

[0046] To achieve the above objectives, this application proposes an improved queue length estimation method that uses sparse CV trajectory data to estimate the intersection traffic for each signal cycle; secondly, it uses a Bayesian inference method to estimate the traffic for cycles without CV trajectory data.

[0047] Specifically, for the case where connected vehicles are detected within the current traffic signal cycle, the intersection is reconstructed using shockwave data from the trajectory data of the N detected connected vehicles to obtain the maximum possible queue length. Then, a probability compensation value is calculated based on the maximum possible queue length and the distance from the stopping point of the last connected vehicle among the N connected vehicles to the intersection stop line. Furthermore, the queue length of the intersection is calculated based on the probability compensation value and the distance from the stopping point of the last connected vehicle to the intersection stop line. Finally, an estimation process is performed based on the queue length of the intersection and the arrival time of the N connected vehicles at the intersection stop line to obtain the periodic traffic flow of the intersection within the current traffic signal cycle. It is evident that the above traffic flow estimation process utilizes the trajectory data of connected vehicles without using loop detectors, overcoming the limitations of existing technologies that rely on loop detectors, thus achieving traffic flow estimation using sparse connected vehicles while ensuring the accuracy of the estimated flow.

[0048] For traffic signal cycles in which no connected vehicles are detected, this application uses a Bayesian method to estimate traffic flow based on estimated periodic traffic flow from multiple historical traffic signal cycles. Existing technologies cannot estimate traffic flow when no connected vehicles are detected. Compared with existing technologies, this application overcomes the dependence of traffic flow estimation on whether connected vehicles are detected within a cycle, and can be applied to traffic flow estimation in more scenarios.

[0049] The traffic flow estimation method provided in this application embodiment can be executed by an electronic device, which may include a terminal or a server. The terminal may include mobile phones, laptops, vehicle terminals, smart wearable devices, and other terminal devices. The server may refer to an independent physical server, a server cluster composed of multiple servers, or a cloud server capable of cloud computing.

[0050] See Figure 1 This document presents an overall framework diagram for traffic flow estimation provided in an embodiment of this application. First, trajectory data of multiple networked vehicles within a single traffic signal cycle is acquired. This trajectory data can be obtained via GPS. Based on the acquired trajectory data, the queue length and periodic traffic flow for a single traffic signal cycle are estimated. Then, using the queue length and periodic traffic flow obtained in the previous step as prior information, a Bayesian inference method is applied to estimate the traffic flow and time limits for different traffic arrival rates. Finally, traffic volume for different time periods, such as traffic volume after 10 minutes, half an hour, or one hour, can be derived from the traffic arrival rates.

[0051] See Figure 2 This is a flowchart illustrating a traffic flow estimation method provided in an embodiment of this application. Figure 2 The method shown can be executed by an electronic device, specifically by the processor of the electronic device. Figure 2 The method shown may include the following steps:

[0052] Step S201: Obtain the trajectory data of N connected vehicles at the intersection during the current traffic signal cycle, and reconstruct the intersection shock wave based on the trajectory data of the N connected vehicles to obtain the maximum possible queue length of the intersection.

[0053] The trajectory data of N connected vehicles can be obtained through GPS. The trajectory data of a single connected vehicle can include information such as the stopping point, stopping time, starting time, and arrival time at the intersection stop line during the current traffic signal cycle at the intersection. This application does not limit this information.

[0054] In one embodiment, reconstructing the intersection shockwave based on the trajectory data of the N connected vehicles to obtain the maximum possible queue length of the intersection may include: determining a first number of connected vehicles that stop at the intersection and a second number of connected vehicles that pass through the intersection based on the trajectory data of the N connected vehicles; and reconstructing the intersection shockwave based on the first and second numbers to obtain the maximum possible queue length of the intersection.

[0055] For example, if N is 3, and within the current traffic signal cycle, one of the three connected vehicles stops at the intersection, while two connected vehicles successfully pass through the intersection, then the first quantity is 1, and the second quantity is 2.

[0056] In the specific implementation, the intersection shock wave is reconstructed based on the first and second quantities to obtain the maximum possible queue length of the intersection, which can be achieved through the following steps S11 and S12:

[0057] Step S11: If N is greater than or equal to 1, and the first quantity is N and the second quantity is 0, then the assembly wave reconstruction and dissipation wave reconstruction are performed based on the trajectory data of the N networked vehicles. The first candidate queue length is calculated based on the reconstructed assembly wave and dissipation wave according to the shock wave theory, and the first candidate queue length is determined as the maximum possible queue length.

[0058] In other words, if N is greater than or equal to 1, the first quantity is N and the second quantity is 0. At this time, the detected connected vehicles have not passed through the intersection in the current traffic signal cycle and have stopped at the intersection. At this time, shock wave reconstruction and dissipation wave reconstruction are performed based on the trajectory data of these N connected vehicles. Then, according to the shock wave theory, the first candidate queue length is calculated based on the reconstructed condensation wave and dissipation wave. The obtained candidate queue length is determined as the maximum queue length.

[0059] As an optional implementation, when N equals 1, the assembly wave reconstruction and dissipation wave reconstruction are performed based on the trajectory data of the N connected vehicles. The first candidate queue length is calculated based on the reconstructed assembly wave and dissipation wave according to the shock wave theory, including: s11, performing shock wave reconstruction based on the distance from the position of the connected vehicle when it starts to join the queue to the intersection stop line and the time when the connected vehicle starts to join the queue, which are determined from the trajectory data; s12, performing dissipation wave reconstruction based on the distance from the position of the connected vehicle when it starts to the intersection stop line and the time when the connected vehicle starts, which are determined from the trajectory data.

[0060] See Figure 3aThis application provides a spatiotemporal map of the trajectory of a connected vehicle when N=1. In s11, the moment when a connected vehicle joins the queue refers to the moment when the speed of the connected vehicle is less than a speed threshold, for example, the speed threshold is set to 5. When the speed of a connected vehicle is detected to be less than 5, it is the moment when the connected vehicle joins the queue. The moment when a connected vehicle joins the queue can be denoted as... The distance from the position of a connected vehicle when it begins to join the queue to the stop line at the intersection can be denoted as... The reconstructed shock wave can be determined by the shock wave velocity W. g The constant term C of the reconstructed straight line from the shock wave g The linear equation of the shock wave after its formation, i.e., the shock wave after its re-emergence, can be expressed as: x = W g ·tC g .

[0061] Optionally, based on the distance from the location where the connected vehicle begins to join the queue to the stop line at the intersection, and the time when the connected vehicle begins to join the queue, determined from the trajectory data, shock wave reconstruction can be performed. Specifically, this can be based on the red light start time of the traffic signal cycle, determined by the time the connected vehicle begins to join the queue. The shock wave velocity is calculated by taking the distance from the starting position of the connected vehicle to the intersection stop line, and then using the difference between the starting time of queuing and the starting time of the red light. The ratio of this difference to the distance from the starting position of the connected vehicle to the intersection stop line is then used as the shock wave velocity. Based on the shock wave velocity, the distance from the starting position of the connected vehicle to the intersection stop line, and the starting time of queuing, a constant term for reconstructing the shock wave line is calculated. This can be done by multiplying the shock wave velocity and the starting time of queuing, and then subtracting the result of this multiplication. The result is used as the constant term for calculating the shock wave line reconstruction. The above steps can be expressed by the following formula:

[0062]

[0063] In s12, the reconstructed dissipated wave can be determined by the dissipated wave velocity W. d and the dissipation wave constant term C d The composition, i.e., the reconstructed equation of the dissipated wave line, can be expressed as: x = W d ·tC d Optionally, based on the distance from the location of the connected vehicle at startup to the stop line at the intersection, determined from the trajectory data. and the start time of the connected vehicle Reconstructing the dissipation wave can include: based on the start time of the connected vehicle, its position at the start time, and the distance from the intersection stop line. and the green light activation time within that traffic signal cycle. Calculate the dissipation wave velocity W d For example, the time when the green light turns on is subtracted from the time when the connected vehicle starts. Then, the distance from the starting position to the stop line at the intersection is compared with the result of the subtraction. The result is used as the dissipation wave velocity; based on the dissipation wave velocity W... d Connected vehicle startup time And the distance from the starting position to the intersection stop line. Calculate the dissipation wave constant term C d For example, the dissipation wave velocity is multiplied by the start time of the connected vehicle, and then the distance from the start position to the stop line at the intersection is subtracted from the multiplication result. The result is used as the dissipation wave constant. The above steps can be specifically represented by the following formula:

[0064]

[0065] Furthermore, after reconstructing the shock wave and dissipating wave in steps s11 and s12, the reconstructed shock wave and dissipating wave are determined when N=1. According to shock wave theory, the intersection point of the reconstructed shock wave and dissipating wave is calculated to obtain the first candidate queue length, which can be expressed as... (This can also be referred to as the maximum queue length) and the time t when the maximum queue length is reached. max In this case, the maximum possible queue length (which can be represented as L) max The length of the queue is equal to the maximum queue length, that is...

[0066] As another optional implementation, when N is greater than 1, that is, at least two connected vehicles are detected in the current traffic signal cycle and at least two connected vehicles stop at the intersection, in this scenario, the rally wave reconstruction and dissipation wave reconstruction are performed based on the trajectory data of the N connected vehicles, and the first candidate queue length is calculated based on the reconstructed rally wave and dissipation wave according to the shock wave theory. Alternatively, the shock wave reconstruction and dissipation wave reconstruction can be performed based on the trajectory data of the last two connected vehicles among the N connected vehicles, and then the first candidate queue length is calculated based on the reconstructed rally wave and dissipation wave according to the shock wave theory.

[0067] See Figure 3b Taking N=2 as an example, this illustrates the spatiotemporal trajectory diagram of a connected vehicle when N is greater than 1. Combined with... Figure 3bAssuming that the last two connected vehicles out of N connected vehicles can be represented as the nth connected vehicle and the (n-1th)th connected vehicle, then the shock wave reconstruction and dissipation wave reconstruction based on the trajectory data of the last two connected vehicles out of N connected vehicles can include:

[0068] s111. Based on the distance from the position of the nth connected vehicle when it starts to join the queue to the stop line of the intersection, the time when the nth connected vehicle starts to join the queue, and the distance from the position of the (n-1)th connected vehicle when it starts to join the queue to the stop line of the intersection, and the time when the (n-1)th connected vehicle starts to join the queue, shock wave reconstruction is performed; where n represents the last vehicle among N vehicles, and n-1 represents the vehicle before the last vehicle among N vehicles.

[0069] Optionally, suppose the distance from the position of the nth connected vehicle when it starts queuing to the stop line at the intersection can be represented as... The moment when the nth connected vehicle begins to join the queue can be represented as: The distance from the position of the (n-1)th connected vehicle when it starts joining the queue to the stop line at the intersection can be represented as: The moment when the (n-1)th connected vehicle begins to join the queue can be represented as:

[0070] The shockwave reconstruction in s11 can specifically include: determining the moment when the nth connected vehicle begins to join the queue. The moment when the (n-1)th connected vehicle begins to join the queue Perform the subtraction operation to obtain the difference at the first moment; calculate the distance from the position of the nth connected vehicle when it starts queuing to the stop line at the intersection. The distance from the position of the (n-1)th connected vehicle when it starts queuing to the stop line at the intersection. Perform a subtraction operation to obtain the first distance difference; divide the first distance difference by the first time difference to obtain the shock wave velocity (which can be expressed as W). g ); the shock wave velocity W g The moment when the nth connected vehicle begins to join the queue Perform a multiplication operation, using the distance from the position of the nth connected vehicle when it starts joining the queue to the stop line at the intersection. Subtracting the result of the multiplication operation, we obtain the shock wave constant term C. g The shock wave velocity W g and shock wave constant C g Substituting this into the equation of the straight line in the spacetime diagram, we can obtain the shock wave after the impact.

[0071] Assume the equation of the straight line in the spacetime graph can be expressed as x = W g·tC g The process of calculating the shock wave velocity and the shock wave constant can be expressed by the following formula:

[0072]

[0073] s112. Based on the distance from the starting position of the nth connected vehicle to the intersection stop line, the starting time of the nth connected vehicle, and the distance from the starting position of the (n-1)th connected vehicle to the intersection stop line and the starting time of the (n-1)th connected vehicle determined from the trajectory data, perform dissipation wave reconstruction.

[0074] Optionally, the distance from the starting position of the nth connected vehicle to the stop line at the intersection can be expressed as: The start-up time of the nth connected vehicle can be represented as The distance from the starting position of the (n-1)th connected vehicle to the stop line at the intersection can be expressed as: The start-up time of the (n-1)th connected vehicle can be represented as

[0075] In step S112, the dissipation wave reconstruction may specifically include: reconstructing the start time of the nth connected vehicle. Start-up time of the (n-1)th connected vehicle Perform a subtraction operation to obtain the second time difference value, which is the distance from the starting position of the nth connected vehicle to the stop line of the intersection. The distance from the starting position of the (n-1)th connected vehicle to the stop line at the intersection Perform a subtraction operation to obtain the second distance difference. Calculate the ratio of the second distance difference to the second time difference, and use the result as the dissipation wave velocity W. d The dissipation wave velocity W d Start-up time of the nth connected vehicle Perform a multiplication operation and use the distance from the starting position of the nth connected vehicle to the stop line at the intersection. Subtract the result of the multiplication operation from the result of the subtraction operation, and use the result as the dissipation constant term C. d Finally, the dissipated wave velocity and dissipated wave constant are substituted into the equation of the dissipated wave spacetime diagram to obtain the reconstructed dissipated wave.

[0076] Assuming the dissipation wave velocity is W d This indicates that the dissipation wave constant term uses C d This indicates that the equation of the dissipation wave spacetime diagram can be expressed as x = W. d ·tC d The above-mentioned dissipation wave velocity W d and the dissipation wave constant term Cd The method for determining this can be expressed by the following formula:

[0077]

[0078] Furthermore, after reconstructing the shock wave and dissipating wave through s111 and s112, the first candidate queue length can be obtained by calculating the intersection point of the reconstructed shock wave and dissipating wave according to the shock wave theory. Then, the first candidate queue length can be directly used as the maximum possible queue length.

[0079] Step S12: If N is greater than or equal to 2, and the first quantity is M, and the second quantity is NM, then shock wave reconstruction and dissipation wave reconstruction are performed based on the trajectory data of the M connected vehicles. The first candidate queue length is calculated based on the reconstructed condensation wave and dissipation wave according to the shock wave theory, and NM second candidate queue lengths are calculated based on the trajectory data of the NM connected vehicles. The minimum value among the first candidate queue length and the NM second candidate queue lengths is determined as the maximum possible queue length; M is an integer greater than 1, and M is less than N.

[0080] It should be understood that when N is greater than or equal to 2, it means that at least two connected vehicles were detected in the current traffic signal cycle. One of these connected vehicles was stopped at the intersection, and the other was passing through the intersection. We define the number of connected vehicles stopped at the intersection as M, and the number of connected vehicles passing through the intersection as NM, where M is an integer less than N. For example, if N = 2, M can be equal to 1; if N = 3, M can be equal to 1; or M can also be equal to 2.

[0081] In one embodiment, shock wave reconstruction and dissipation wave reconstruction are performed based on the trajectory data of the M connected vehicles. The first candidate queue length is calculated based on the reconstructed convective and dissipative waves according to shock wave theory. The specific implementation method is the same as steps s111 and s112 mentioned above, except that N is replaced with M in steps s111 and s112, which will not be elaborated further here. In short, based on the distance from the starting position of the last two connected vehicles among the M connected vehicles to the intersection stop line, their respective starting times, and the distance and starting time from their starting positions to the intersection stop line, the shock wave velocity, shock wave constant, dissipation wave velocity, and dissipation wave constant are calculated respectively. Then, shock wave reconstruction is performed based on the shock wave velocity and shock wave constant, and dissipation wave reconstruction is performed based on the dissipation wave velocity and dissipation wave constant. Finally, the first candidate queue length is determined based on the reconstructed shock wave and dissipation wave.

[0082] In one embodiment, the first candidate queue length can be expressed as: The queue length of NM second candidates can be expressed as... The value of i ranges from 1 to NM, and the maximum possible queue length can be represented as L. max After determining the first candidate queue length and multiple second candidate queue lengths, the minimum value among the first candidate queue length and the NM second candidate queue lengths can be determined as the maximum possible queue length, which can be expressed by the following formula:

[0083]

[0084] In one embodiment, in step S12, calculating NM second candidate queue lengths based on the trajectory data of the NM connected vehicles may include: for each of the NM connected vehicles, calculating the second candidate queue length of the connected vehicle through the following steps: (1) obtaining the time difference between the time when the connected vehicle passes the intersection stop line and the time when the traffic signal light allows passage within the current traffic signal cycle is turned on from the trajectory data; (2) obtaining the reconstructed dissipation wave velocity and the free flow velocity of the connected vehicle passing the intersection stop line, and adding the dissipation wave velocity and the free flow velocity after taking their reciprocals to obtain the addition result; (3) using the ratio of the time difference to the addition result as the second candidate queue length of the connected vehicle.

[0085] In (2) above, the free-flow velocity of each connected vehicle can be calculated as follows: subtract the time when the connected vehicle passes the stop line at the intersection from the time when it enters the intersection area to obtain a third time difference; then take the reciprocal of the third time difference to obtain the free-flow velocity of the connected vehicle. Assume the time when the connected vehicle enters the intersection area is expressed as... The time when a connected vehicle passes the stop line at an intersection is represented as follows: The free-flow velocity of the connected vehicle can be expressed as follows: The free-flow velocity of the connected vehicle (which can be expressed as Vb) can be calculated using the following formula:

[0086]

[0087] Assuming that in (1), the activation time of the traffic light (i.e., the green light) that allows passage within the current traffic signal cycle can be expressed as follows: In (3), the second candidate queue length of a connected vehicle can be expressed as The second candidate queue length of a connected vehicle, calculated using (1)-(3), can be expressed by the following formula:

[0088]

[0089] It should be understood that the above is only an example of calculating the second candidate queue length for a connected vehicle. For other connected vehicles, the same method can be used to calculate the corresponding second candidate queue length, which will not be elaborated here.

[0090] Based on the foregoing description, when N is greater than 2, it can be achieved through the following... Figure 3c and Figure 3d For example, let's introduce two possible scenarios when N is greater than 2. See [link / reference] Figure 3c Let N equal 2 and M equal 1, meaning that two connected vehicles are detected within the current traffic signal cycle: one connected vehicle is stopped at the intersection, and the other vehicle passes through the intersection. See also Figure 3d For example, N is greater than 2, and M is greater than 1, meaning that at least 4 connected vehicles are detected within the current traffic signal cycle, at least 2 are stopped at the intersection, and at least 2 are at the intersection. Figure 3c and Figure 3d The description can be found in the preceding text, and will not be repeated here.

[0091] In one embodiment, after determining the gathering wave velocity and dissipation wave velocity (both of which can be collectively referred to as shock wave velocity) through the preceding steps, this application can further perform correction processing on the reconstructed shock wave velocity based on Bayesian inference to ensure the accuracy of the shock wave velocity. See also Figure 4 This document provides a flowchart of a shock wave velocity correction process according to an embodiment of this application. Specifically, the correction process for the reconstructed shock wave velocity based on Bayesian inference may include: obtaining historical velocity data from the previous W traffic signal cycles, and calculating a log-normal distribution of the velocity data based on the W historical velocity data using Bayesian inference; determining whether the shock wave velocity requires correction processing based on the log-normal distribution of the velocity data; if the shock wave velocity requires correction processing, calculating a new log-normal distribution based on the W historical velocity data and the shock wave velocity using Bayesian inference; and performing correction processing on the shock wave velocity based on the new log-normal distribution.

[0092] The following combines the above steps and Figure 4 This paper specifically analyzes how to correct the shock wave velocity. It has been verified that the shock wave velocity follows a log-normal distribution, thus the distribution parameters can be estimated. Based on this, this application proposes a method for correcting the reconstructed shock wave velocity using Bayesian inference. In the specific implementation, it is assumed that the observation dataset used for Bayesian inference is W, W = (w1, ..., w...). n If W contains the historical wave velocities of W traffic signal cycles, then the prior set for Bayesian inference should be D = (x1, ..., x...). n Find the middle value of x. i =logw iTherefore, the normal distribution can be calculated using the Bayesian inference formula.

[0093] Furthermore, the need for correction processing of shock wave velocity can be determined based on the normal distribution of wave velocity. For shock wave velocities that need correction processing, then correction processing can be performed. This avoids the need for indiscriminate correction processing of all shock wave velocities, thus saving the resource consumption of electronic equipment.

[0094] In specific implementation, in order to determine the shock wave velocity (which can be expressed as w) i To determine whether corrective action is needed, a hypothesis testing method is proposed. The null hypothesis H0 is to test the shock wave velocity w. i It is an unbiased estimator, with the alternative hypothesis H. a To test the shock wave velocity w i This is a biased estimate, therefore correction is needed. Due to the continuity of traffic flow, the decision rule is as follows:

[0095]

[0096] in, and These are the upper and lower critical values, which are also normal distributions. up and down quantiles, of which, normal distribution It could be calculated using historical wave velocities from W traffic signal cycles as prior data.

[0097] Since the number of parked connected vehicles in a single traffic signal cycle will result in different confidence levels for the basic shock wave reconstruction results, this application defines the significance level α for the hypothesis test as follows:

[0098]

[0099] Where, N stopCV This indicates the number of connected parking vehicles detected within a traffic signal cycle.

[0100] Next, after determining through the above steps that the reconstructed shock wave velocity needs correction, the shock wave velocity requiring correction is added to the historical velocity data of W historical traffic signal cycles. A new normal distribution is then calculated again using the Bayesian inference formula. Assuming the currently reconstructed shock wave velocity is from the i-th traffic signal cycle, the shock wave velocity is represented as w. i Assuming W equals 5, adding the shock wave velocity to the historical wave velocity set yields a new set of wave velocities: W = (w i-5 , ..., w i The new normal distribution is calculated as follows: Correcting the shock wave velocity based on the new normal distribution can be...

[0101] It should be noted that after correcting the shock wave velocity, it is also necessary to determine the updated shock wave constant terms, including the gathering wave constant term or the dissipating wave constant term, based on the parking point or starting point corresponding to the last connected vehicle.

[0102] Step S202: Calculate the probability compensation value based on the maximum possible queue length and the distance from the parking point of the last connected vehicle among the N connected vehicles to the intersection parking line.

[0103] Optionally, if it is assumed that the vehicles arriving at the intersection during the current traffic signal cycle follow a Poisson distribution, step S202 may be implemented in the following ways: obtaining the average headway and the number of lanes in the direction of travel corresponding to the intersection; inputting the maximum possible queue length, the distance from the stopping point of the last connected vehicle to the stop line of the intersection, the average headway, and the number of lanes into the probability compensation value calculation rules for calculation to obtain the probability compensation value.

[0104] The average headway can be expressed as l v k is the number of lanes in the corresponding direction of travel. The probability compensation value is obtained by inputting the maximum possible queue length, the distance from the stopping point of the last connected vehicle to the stop line at the intersection, the average headway, and the number of lanes into the probability compensation value calculation rules. It can be expressed by the following formula:

[0105]

[0106] Where Error represents the probability compensation value, and P[(i-1)·k<X≤i·k] represents the error as i·l v The above formula essentially sums the probability of each error value and its probability of occurrence, using the result as the probability compensation value. Considering the different number of lanes, the probability of each error can be calculated using the following formula:

[0107]

[0108] Let t max The maximum possible queue length L max The time of arrival, t cv If the last connected vehicle stops, the time Δt of missing trajectory data can be calculated as follows:

[0109] Δt=t max -t cv

[0110] Therefore, P(X=x) is the probability of arriving at x connected vehicles during the time period when trajectory data is missing, and its calculation formula can be:

[0111]

[0112] Because Bayesian inference is used for correction in the shock wave reconstruction, the fit to the latter part of the shock wave is better, thus effectively reflecting the arrival rate during the red light period. Therefore, the Poisson distribution parameter λ in the above formula is determined by the maximum possible queue length L. max and the corresponding arrival time t max The calculation yields the following formula:

[0113]

[0114] Step S203: Estimate the queue length at the intersection based on the probability compensation value and the distance from the parking point of the last connected vehicle among the N connected vehicles to the intersection parking line.

[0115] In the specific implementation, the probability compensation value and the distance from the stopping point of the last connected vehicle to the intersection stop line are added together to obtain the queue length at the intersection. For example, suppose the probability compensation value is represented as Error, the distance from the stopping point of the last connected vehicle to the intersection stop line is represented as La, and the queue length at the intersection is represented as L. q Then L q =L a +Error.

[0116] Step S204: Based on the queue length at the intersection and the time it takes for N connected vehicles to arrive at the stop line at the intersection, estimate the periodic traffic flow at the intersection during the current traffic signal cycle.

[0117] Periodic traffic flow can be defined as the traffic volume at the intersection during the current traffic signal cycle.

[0118] In one embodiment, traffic flow processing is performed based on the queue length at the intersection and the arrival times of the N vehicles at the stop line at the intersection to obtain the traffic flow at the intersection during the current traffic signal cycle. This may include:

[0119] First, using the kernel density estimation method, based on the arrival times of the N connected vehicles at the intersection stop line, the probability density function of the total arrival time of the connected vehicles and the traffic arrival volume during the red light time are estimated. The probability density function can be expressed as:

[0120]

[0121] Where φ(x) is the Gaussian kernel, h is the smoothing bandwidth, and T r，iThe relative arrival time of connected vehicles is defined with the start time of the red light in each traffic signal cycle as zero.

[0122] Secondly, the rate of change is calculated based on the probability density function according to the rules for calculating the rate of change. Specifically, the rate of change can be the integral of the total number of arrivals during the entire traffic signal cycle divided by the total number of arrivals during the red light period. The formula for calculating the rate of change can be expressed as:

[0123]

[0124] Finally, the ratio of the queue length to the average vehicle spacing is calculated, and this ratio is multiplied by the rate of change to obtain the periodic traffic flow within the current traffic signal cycle. Let the periodic traffic flow be represented by Q, the average vehicle spacing by lv, and the queue length by L. real Traffic cycle flow can be calculated using the following formula:

[0125]

[0126] In this embodiment, for the case where connected vehicles are detected within the current traffic signal cycle, shockwave reconstruction is performed on the intersection based on the trajectory data of the detected N connected vehicles to obtain the maximum possible queue length. Then, a probability compensation value is calculated based on the maximum possible queue length and the distance from the stopping point of the last connected vehicle among the N connected vehicles to the intersection stop line. Furthermore, the queue length of the intersection is calculated based on the probability compensation value and the distance from the stopping point of the last connected vehicle to the intersection stop line. Finally, an estimation process is performed based on the queue length of the intersection and the arrival time of the N connected vehicles at the intersection stop line to obtain the periodic traffic flow of the intersection within the current traffic signal cycle. It can be seen that the above traffic flow estimation process utilizes the trajectory data of connected vehicles without using loop detectors, overcoming the limitation of existing technologies that rely on loop detectors, thus affecting estimation accuracy. This achieves traffic flow estimation using sparse connected vehicles while ensuring the accuracy of the estimated flow.

[0127] Based on the embodiments of the traffic flow estimation method described above, this application also provides another traffic flow estimation method, see [link to relevant documentation]. Figure 5 This is a flowchart illustrating another traffic flow estimation method provided in an embodiment of this application. Figure 5 The main challenge is how to estimate periodic traffic flow for traffic signal cycles when no connected vehicles arrive. Figure 5 The traffic flow estimation method may specifically include the following steps:

[0128] Step S501: Obtain the traffic flow of P traffic signal cycles.

[0129] It should be noted that the periodic traffic flow for each traffic signal cycle can be calculated using... Figure 2 The traffic flow estimation method described above is used for determination. For detailed implementation methods, please refer to... Figure 2 The description will not be repeated here.

[0130] Step S502: Estimate the periodic traffic arrival rate based on P periods of traffic flow.

[0131] Assume that the traffic volume in one cycle follows a Poisson distribution Q ~ Poisson(λ), where λ is the Poisson distribution parameter, also known as the traffic arrival rate. In this embodiment, P cycles of traffic flow are used as priors, and the distribution followed by λ is estimated using a Bayesian inference method based on the Markov Chain Monte Carlo (MCMC) method.

[0132] Assume that the periodic traffic flow Q follows a pattern of Q ~ Poisson(λ), P(λ) is the prior distribution of λ, and D is the observed data (i.e., the periodic traffic flow during at least one traffic signal cycle with connected vehicles stopped). The distribution of the Poisson distribution parameter λ can be calculated using the following Bass inference formula:

[0133] P(λ|D)∝P(D|λ)·P(λ)

[0134] Here, the prior distribution of λ is assumed to be an exponential distribution λ~Exp(β), where β is the mean of the observed data.

[0135] In one embodiment, step S502, which estimates the periodic traffic arrival rate based on P periods of traffic flow, may include: calculating a prior distribution of the traffic arrival rate based on the P periods of traffic flow using a Bayesian inference formula; the prior distribution of the traffic arrival rate is an exponential distribution such as λ ~ Exp(β), where the parameter β in the exponential distribution is the mean of the P periods of traffic flow; obtaining an initial traffic arrival rate, and iterating over a target number of times based on the prior distribution and the initial traffic arrival rate to obtain the traffic arrival rate;

[0136] In the j-th iteration, a random traffic arrival rate j is selected; j is an integer greater than or equal to 1 and less than the target number; the probability of the traffic arrival rate j being used as the traffic arrival rate is calculated according to the Monte Carlo rule; if the probability is greater than the probability of the traffic arrival rate j-1 determined in the (j-1)-th iteration, then the traffic arrival rate j is used as the traffic arrival rate; where, when j equals 1, the traffic arrival rate j-1 refers to the initial traffic arrival rate; if the probability is less than or equal to the probability of the traffic arrival rate j-1 determined in the (j-1)-th iteration, then the probability of the traffic arrival rate j-1 is used as the traffic arrival rate.

[0137] Here, assuming the initial traffic arrival rate is denoted as λ(0), the following uses the j-th iteration as an example to introduce how to iterate the target number based on the prior distribution and the initial traffic arrival rate to obtain the traffic arrival rate.

[0138] 1. In the j-th iteration, a random traffic arrival rate j is selected from the prior distribution λ~Exp(β), which can also be expressed as λ(j);

[0139] 2. Calculate the probability that the traffic arrival rate j is used as the traffic arrival rate according to the Monte Carlo rule. The specific calculation can be expressed by the following formula:

[0140]

[0141] Where P(j) represents the probability that traffic arrival rate j is used as the traffic arrival rate.

[0142] 3. If the probability determined in step 2 is greater than the probability determined in the (j-1)th iteration, then the traffic arrival rate j is taken as the traffic arrival rate; otherwise, the traffic arrival rate j-1 determined in the (j-1)th iteration is retained as the traffic arrival rate. Wherein, when j equals 1, the traffic arrival rate j-1 refers to the initial traffic arrival rate.

[0143] The above 1-3 are just examples of one iteration. After each iteration, the traffic arrival rate will be updated and will maintain the result of the previous iteration. After completing the target number of iterations, the determined traffic arrival rate will be used as the final determined traffic arrival rate.

[0144] Step S503: Based on the theory that periodic traffic flow follows a Poisson distribution, and based on the traffic arrival rate, estimate the periodic traffic flow within the future target time period of the traffic signal period in which no connected vehicles were detected.

[0145] After determining the traffic arrival rate λ in step S502, based on the theory that the periodic traffic arrival rate follows a Poisson distribution, i.e., Q ~ Poisson(λ), the periodic traffic flow in the future target time period can be estimated within the traffic signal period in which no connected vehicles are detected. For example, the periodic traffic flow in the next 10 minutes, the next half hour, or the next hour.

[0146] In this embodiment, for traffic signal cycles in which no connected vehicles are detected, traffic flow is estimated using a Bayesian method based on the estimated periodic traffic flow from multiple historical traffic signal cycles. Existing technologies cannot estimate traffic flow when no connected vehicles are detected. Compared with existing technologies, this application overcomes the dependence of traffic flow estimation on whether connected vehicles are detected within a cycle, and can be applied to traffic flow estimation in more scenarios.

[0147] Steps S501-S503 can estimate the traffic arrival rate λ for any predefined time range. However, the traffic flow throughout the day and the traffic flow of connected vehicles follow similar time-varying Poisson distributions. Therefore, assuming there are M different traffic flow phases for P periods, each with a different arrival rate, this can be expressed as follows:

[0148]

[0149] Furthermore, the time-varying Poisson distribution parameters also follow the same exponential distribution, as shown below:

[0150] λ1~Exp(β CV )

[0151] λ2~Exp(β CV )

[0152] …

[0153] λ M-1 ~Exp(β) CV )

[0154] Where, β CV It is the reciprocal of the arithmetic mean of the periodic traffic flow, i.e.

[0155]

[0156] The specific value of M can be determined by Bayesian inference and time boundary identification using the P periods of traffic flow as prior information. Further, for each time boundary, based on the traffic flow within that time boundary, the periodic traffic arrival rate is estimated, and according to the theory that the traffic flow for each time boundary follows a Poisson distribution, the traffic flow within each time boundary is estimated based on the traffic arrival rate. The correspondence between time boundaries, traffic arrival rate, and traffic flow is established and stored. The calculation method for the traffic arrival rate for each time boundary can be the same as in steps S501-S503, and will not be repeated here.

[0157] In this embodiment of the application, in order to simulate the changes in traffic volume in actual traffic flow, a model is established as follows: Figure 6a The intersection and vehicle arrival model described above, and the intersection signal timing scheme are as follows: the traffic signal cycle is 150 seconds, and the green light time for left turns and straight-ahead is 50 seconds each.

[0158] First, the time limit is identified based on the periodic traffic flow of P traffic signal cycles. If the identification result is 100, then the actual time limit is the 96th cycle. Figure 6b As shown, in Figure 6bThe bottom part represents the posterior time limit, the middle part represents the posterior traffic arrival rate, and the top part represents the time limit and the corresponding periodic traffic flow.

[0159] In this embodiment, based on multiple periodic traffic flows, the posterior distribution of traffic flow λ corresponding to two stages is obtained through 10,000 MCMC processes combined with Bayesian inference, as shown below. Figure 6c As shown in Table 1 below, the mean absolute percentage error of traffic flow estimation using this method is as low as 5.5%.

[0160] Table 1

[0161]

[0162] Meanwhile, the embodiments of this application also measured the corresponding queue length estimation results, as shown in Table 2 below. Table 2 shows the error of the cycle traffic flow estimated by the queue length when the penetration rate is 5%. As can be seen from Table 2, the average absolute error of this method is 15.93%, while the reason why the former error is relatively lower may be because Bayesian inference eliminates the influence of some extreme estimates.

[0163] Table 2

[0164]

[0165] Finally, based on the estimation results of periodic traffic flow and time limits, the periodic traffic flow of vehicles without network connectivity can be estimated, and further, the traffic flow for 10 minutes, half an hour, and hours can be obtained.

[0166] The above verification shows that the traffic flow estimation method of this application has strong robustness to the estimation of Poisson arrival rate, time limit, and time period traffic flow; the traffic flow estimation method of this application solves the influence of sparse trajectory on flow estimation and can be applied to low-permeability road network environment.

[0167] To test the performance of the traffic flow estimation method on a real-world dataset, this application applies the method to the NGSIM Peachtree dataset, which contains 5 intersections and 6 arterial road sections. The NGSIM Peachtree dataset contains vehicle trajectory data for all vehicles on the street, including Vehicle_ID, Global_Time, Local_X, Local_Y, v_Vel, and Direction. Since only intersections 2 and 3 contain all southbound (SB) and northbound (NB) entrance vehicle trajectories, these two intersections were chosen as the case study targets. Furthermore, this dataset only contains 15 minutes of vehicle trajectories, approximately 10 cycles.

[0168] To simulate GPS trajectories for CV under low penetration conditions, we applied a traffic flow estimation method to estimate traffic arrival rates by randomly sampling vehicle trajectories from the NGSIM Peachtree dataset with a probability of 5%. Since the data only contains 15 minutes of vehicle trajectories, only one traffic arrival rate exists. The posterior distributions of southbound and northbound traffic arrival rates at intersections 2 and 3 are as follows: Figure 6d and Figure 6e As shown in Table 3, the estimates of λ and absolute error using the law of large numbers are given, with the mean absolute error of all estimates being 5.2%.

[0169] Table 3

[0170]

[0171] The results above demonstrate that this method can also be applied to estimate actual traffic volume using data at extremely low penetration rates (5%). Even though some estimates have an absolute error of 9.8% (southbound approach of intersection 3), the estimated traffic volume becomes more accurate with more CV trajectories recorded. Therefore, this method can be applied even with low CV data penetration.

[0172] Based on the above embodiments of the traffic flow estimation method, this application provides a traffic flow estimation device, see [link to relevant documentation]. Figure 7 This is a schematic diagram of a traffic flow estimation device provided in an embodiment of this application. Figure 7 The control device shown can operate the following units:

[0173] The acquisition unit 701 is used to acquire trajectory data of N connected vehicles at the intersection during the current traffic signal cycle, and reconstruct the intersection shock wave based on the trajectory data of the N connected vehicles to obtain the maximum possible queue length of the intersection; N is an integer greater than or equal to 1.

[0174] The calculation unit 702 is used to calculate a probability compensation value based on the maximum possible queue length and the distance from the parking point of the last connected vehicle among the N connected vehicles to the intersection parking line.

[0175] The estimation unit 703 is used to estimate the queue length of the intersection based on the probability compensation value and the distance from the parking point of the last connected vehicle among the N connected vehicles to the parking line of the intersection.

[0176] The processing unit 704 is used to perform estimation processing based on the queue length of the intersection and the time when the N networked vehicles arrive at the stop line of the intersection to obtain the periodic traffic flow of the intersection in the current traffic signal cycle.

[0177] In one embodiment, when the acquisition unit 701 reconstructs the intersection shock wave based on the trajectory data of the N networked vehicles to obtain the maximum possible queue length of the intersection, it performs the following steps:

[0178] Based on the trajectory data of the N connected vehicles, determine a first number of connected vehicles that stopped at the intersection and a second number of connected vehicles that passed through the intersection.

[0179] The shock wave at the intersection is reconstructed based on the first quantity and the second quantity to obtain the maximum possible queue length at the intersection.

[0180] In one embodiment, when the acquisition unit 701 reconstructs the intersection shock wave based on the first quantity and the second quantity to obtain the maximum possible queue length of the intersection, it performs the following steps:

[0181] If N is greater than or equal to 1, and the first quantity is N and the second quantity is 0, then the assembly wave reconstruction and dissipation wave reconstruction are performed based on the trajectory data of the N connected vehicles. The first candidate queue length is calculated based on the reconstructed assembly wave and dissipation wave according to the shock wave theory, and the first candidate queue length is determined as the maximum possible queue length.

[0182] If N is greater than or equal to 2, and the first quantity is M, and the second quantity is NM, then based on the trajectory data of the M connected vehicles, the assembly wave reconstruction and dissipation wave reconstruction are performed. According to the shock wave theory, the first candidate queue length is calculated based on the reconstructed assembly wave and dissipation wave, and NM second candidate queue lengths are calculated based on the trajectory data of the NM connected vehicles. The minimum value among the first candidate queue length and the NM second candidate queue lengths is determined as the maximum possible queue length; M is an integer greater than 1, and M is less than N.

[0183] In one embodiment, when the acquisition unit 701 performs assembly wave reconstruction and dissipation wave reconstruction based on the trajectory data of the L networked vehicles, and calculates the first candidate queue length based on the reconstructed assembly wave and dissipation wave according to the shock wave theory, L is equal to N, and N is greater than 1; or L is equal to M; the following steps are performed:

[0184] Based on the distance from the position of the nth connected vehicle when it starts queuing to the intersection stop line, the time when the nth connected vehicle starts queuing, and the distance from the position of the (n-1)th connected vehicle when it starts queuing to the intersection stop line, the rally wave is reconstructed; where n represents the last vehicle in the N vehicles, and n-1 represents the vehicle before the last vehicle in the N vehicles.

[0185] Based on the distance from the starting position of the nth connected vehicle to the stop line of the intersection, the starting time of the nth connected vehicle, and the distance from the starting position of the (n-1)th connected vehicle to the stop line of the intersection and the starting time of the (n-1)th connected vehicle determined from the trajectory data, dissipation wave reconstruction is performed.

[0186] In one embodiment, if L equals N and N equals 1, then when the acquisition unit 701 performs aggregate wave reconstruction and dissipation wave reconstruction based on the trajectory data of the L networked vehicles, it performs the following steps:

[0187] Based on the distance from the location where the connected vehicle begins to join the queue to the intersection stop line and the time when the connected vehicle begins to join the queue, determined from the trajectory data, a rallying wave reconstruction is performed. Based on the distance from the location where the connected vehicle starts to the intersection stop line and the time when the connected vehicle starts, determined from the trajectory data, a dissipation wave reconstruction is performed.

[0188] In one embodiment, if L equals M, then when the acquisition unit 701 calculates the NM second candidate queue lengths based on the trajectory data of the NM networked vehicles, it performs the following steps:

[0189] For each of the NM connected vehicles, the second candidate queue length of the connected vehicles is calculated through the following steps:

[0190] From the trajectory data, obtain the time difference between the time when the connected vehicle passes the stop line at the intersection and the time when the traffic light that allows passage is turned on in the current traffic signal cycle;

[0191] Based on the reconstruction of the dissipated wave, the dissipated wave velocity and the free flow velocity are determined, and the reciprocals of the dissipated wave velocity and the free flow velocity are added together to obtain the addition result.

[0192] The ratio of the time difference to the summation result is used as the second candidate queue length for the connected vehicles.

[0193] In one embodiment, the processing unit 704 is further configured to:

[0194] The reconstructed shock wave velocity is corrected based on Bayesian inference. The reconstructed shock wave velocity includes the reconstructed condensation wave velocity and the reconstructed dissipation wave velocity.

[0195] In one embodiment, when the processing unit 704 performs correction processing on the reconstructed shock wave velocity based on Bayesian inference, it executes the following steps:

[0196] Obtain the historical wave velocity of the previous W traffic signal cycles, and calculate the log-normal distribution of the wave velocity based on the W historical wave velocities using Bayesian inference;

[0197] If the shock wave velocity needs to be corrected based on the log-normal distribution of wave velocity, then a new log-normal distribution is calculated based on W historical wave velocities and the shock wave velocity using Bayesian inference.

[0198] The shock wave velocity is corrected based on the new log-normal distribution.

[0199] In one embodiment, if the vehicles arriving at the intersection within the current traffic signal cycle follow a Poisson distribution, the calculation unit 702 performs the following steps when calculating the probability compensation value based on the maximum possible queue length and the distance from the stopping point of the last connected vehicle among the N connected vehicles to the stop line at the intersection:

[0200] Obtain the average headway and the number of lanes in the direction of travel corresponding to the intersection;

[0201] The probability compensation value is obtained by inputting the maximum possible queue length, the distance from the parking point of the last connected vehicle to the intersection parking line, the average headway, and the number of lanes into the probability compensation value calculation rules.

[0202] In one embodiment, when the processing unit 704 performs traffic flow estimation based on the queue length of the intersection and the arrival times of the N vehicles at the intersection stop line to obtain the traffic flow of the intersection during the current traffic signal cycle, it performs the following steps:

[0203] The probability density function of arrival time is estimated based on the arrival time of the N networked vehicles at the intersection stop line using the kernel density estimation method.

[0204] The rate of change is obtained by performing calculations based on the probability density function according to the rate of change calculation rules.

[0205] Calculate the ratio of the queue length to the average vehicle distance, and multiply the ratio by the rate of change to obtain the periodic traffic flow within the current traffic signal cycle.

[0206] In this embodiment, for the case where connected vehicles are detected within the current traffic signal cycle, shockwave reconstruction is performed on the intersection based on the trajectory data of the detected N connected vehicles to obtain the maximum possible queue length. Then, a probability compensation value is calculated based on the maximum possible queue length and the distance from the stopping point of the last connected vehicle among the N connected vehicles to the intersection stop line. Furthermore, the queue length of the intersection is calculated based on the probability compensation value and the distance from the stopping point of the last connected vehicle to the intersection stop line. Finally, an estimation process is performed based on the queue length of the intersection and the arrival time of the N connected vehicles at the intersection stop line to obtain the periodic traffic flow of the intersection within the current traffic signal cycle. It can be seen that the above traffic flow estimation process utilizes the trajectory data of connected vehicles without using loop detectors, overcoming the limitation of existing technologies that rely on loop detectors, thus affecting estimation accuracy. This achieves traffic flow estimation using sparse connected vehicles while ensuring the accuracy of the estimated flow.

[0207] Based on the foregoing embodiments of the traffic flow estimation method and the traffic flow device, this application also provides another traffic flow estimation device, see [link to relevant documentation]. Figure 8 This is a schematic diagram of another traffic flow estimation device provided in an embodiment of this application. See also... Figure 8 The traffic flow estimation device can operate the following units:

[0208] The acquisition unit 801 is used to acquire the traffic flow of P traffic signal cycles, wherein the traffic flow of each traffic signal cycle is determined by the method described in any one of claims 1-10; P is an integer greater than or equal to 1.

[0209] Estimation unit 802 is used to estimate the periodic traffic arrival rate based on the P periodic traffic flows;

[0210] The estimation unit 802 is further configured to estimate the traffic flow in the future target time period within the traffic signal period for which no connected vehicles were detected, based on the traffic arrival rate, according to the theory that periodic traffic flow follows a Poisson distribution.

[0211] In one embodiment, when estimating the periodic traffic arrival rate based on the P periods of traffic flow, the estimation unit 802 performs the following steps:

[0212] The prior distribution of traffic arrival rate is calculated based on the traffic flow over the P periods using the Bayesian inference formula; the prior distribution of traffic arrival rate is an exponential distribution, where the parameter of the exponential distribution is the mean of the traffic flow over the P periods.

[0213] Obtain the initial traffic arrival rate, and iterate the target number based on the prior distribution and the initial traffic arrival rate to obtain the traffic arrival rate;

[0214] For the j-th iteration, a random traffic arrival rate j is selected based on the prior distribution; j is an integer greater than or equal to 1 and less than the target number.

[0215] The probability of the traffic arrival rate j being used as the traffic arrival rate is calculated according to the Monte Carlo rule.

[0216] If the probability is greater than the probability of the traffic arrival rate j-1 determined in the (j-1)th time, then the traffic arrival rate j is taken as the traffic arrival rate; where, when j equals 1, the traffic arrival rate j-1 refers to the initial traffic arrival rate.

[0217] If the probability is less than or equal to the probability of the traffic arrival rate j-1 determined in the (j-1)th time, then the traffic arrival rate j-1 is taken as the traffic arrival rate.

[0218] In one embodiment, the traffic flow estimation device further includes a processing unit, the processing unit being used for:

[0219] The P-cycle traffic flow is used as prior information for Bayesian inference and time boundary identification to obtain multiple time boundaries.

[0220] For each time limit, based on the traffic flow within the time limit, the periodic traffic arrival rate is estimated, and based on the theory that the traffic flow at each time limit follows a Poisson distribution, the traffic flow within each time limit is estimated based on the traffic arrival rate.

[0221] Establish and store the correspondence between time limits, traffic arrival rates, and traffic flow.

[0222] In this embodiment, for traffic signal cycles in which no connected vehicles are detected, traffic flow is estimated using a Bayesian method based on the estimated periodic traffic flow from multiple historical traffic signal cycles. Existing technologies cannot estimate traffic flow when no connected vehicles are detected. Compared with existing technologies, this application overcomes the dependence of traffic flow estimation on whether connected vehicles are detected within a cycle, and can be applied to traffic flow estimation in more scenarios.

[0223] Based on the above-described traffic flow estimation method and control device embodiments, this application also provides an electronic device, see [link to relevant documentation]. Figure 9 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Figure 7The control device shown may include a processor 701, an input interface 702, an output interface 703, and a computer storage medium 703. The processor 701, input interface 702, output interface 703, and computer storage medium 703 may be connected via a bus or other means.

[0224] The computer storage medium 704 can be stored in the memory of the control device. The computer storage medium 704 is used to store computer programs, and the processor 701 is used to execute the computer programs stored in the computer storage medium 704. The processor 701 (or CPU (Central Processing Unit)) is the computing and control core of the control device, and it is suitable for implementing one or more computer programs, specifically suitable for loading and executing the aforementioned traffic flow estimation method.

[0225] This application embodiment also provides a computer storage medium (memory), which is a memory device of a control device used to store programs and data. It is understood that the computer storage medium here may include the built-in storage medium of the control device, or it may include an extended storage medium supported by a data processing device. The computer storage medium provides storage space that stores the operating system of the control device. Furthermore, one or more computer programs suitable for loading and execution by the processor 701 are also stored in this storage space. It should be noted that the computer storage medium here may be a high-speed RAM memory, or a non-volatile memory, such as at least one disk storage device; optionally, it may also be at least one computer storage medium located remotely from the aforementioned processor.

[0226] In one embodiment, one or more computer programs stored in the computer storage medium may be loaded by processor 901 and executed as described above for traffic flow estimation.

[0227] In this embodiment, for the case where connected vehicles are detected within the current traffic signal cycle, shockwave reconstruction is performed on the intersection based on the trajectory data of the detected N connected vehicles to obtain the maximum possible queue length. Then, a probability compensation value is calculated based on the maximum possible queue length and the distance from the stopping point of the last connected vehicle among the N connected vehicles to the intersection stop line. Furthermore, the queue length of the intersection is calculated based on the probability compensation value and the distance from the stopping point of the last connected vehicle to the intersection stop line. Finally, an estimation process is performed based on the queue length of the intersection and the arrival time of the N connected vehicles at the intersection stop line to obtain the periodic traffic flow of the intersection within the current traffic signal cycle. It can be seen that the above traffic flow estimation process utilizes the trajectory data of connected vehicles without using loop detectors, overcoming the limitation of existing technologies that rely on loop detectors, thus affecting estimation accuracy. This achieves traffic flow estimation using sparse connected vehicles while ensuring the accuracy of the estimated flow.

[0228] For traffic signal cycles in which no connected vehicles are detected, this application uses a Bayesian method to estimate traffic flow based on estimated periodic traffic flow from multiple historical traffic signal cycles. Existing technologies cannot estimate traffic flow when no connected vehicles are detected. Compared with existing technologies, this application overcomes the dependence of traffic flow estimation on whether connected vehicles are detected within a cycle, and can be applied to traffic flow estimation in more scenarios.

Claims

1. A traffic flow estimation method, characterized in that, include: The trajectory data of N connected vehicles at the intersection during the current traffic signal cycle is obtained, and the intersection shock wave is reconstructed based on the trajectory data of the N connected vehicles to obtain the maximum possible queue length of the intersection; N is an integer greater than or equal to 1. The probability compensation value is calculated based on the maximum possible queue length and the distance from the parking point of the last connected vehicle among the N connected vehicles to the intersection parking line; Based on the probability compensation value and the distance from the parking point of the last connected vehicle among the N connected vehicles to the parking line of the intersection, the queue length of the intersection is estimated. Based on the queue length at the intersection and the time it takes for the N connected vehicles to arrive at the stop line at the intersection, the periodic traffic flow at the intersection during the current traffic signal cycle is estimated. The step of reconstructing the intersection shock wave based on the trajectory data of the N connected vehicles to obtain the maximum possible queue length at the intersection includes: Based on the trajectory data of the N connected vehicles, determine a first number of connected vehicles that stopped at the intersection and a second number of connected vehicles that passed through the intersection. The shock wave at the intersection is reconstructed based on the first quantity and the second quantity to obtain the maximum possible queue length at the intersection. The step of reconstructing the intersection shock wave based on the first quantity and the second quantity to obtain the maximum possible queue length of the intersection includes: If N is greater than or equal to 1, and the first quantity is N and the second quantity is 0, then the assembly wave reconstruction and dissipation wave reconstruction are performed based on the trajectory data of the N connected vehicles. The first candidate queue length is calculated based on the reconstructed assembly wave and dissipation wave according to the shock wave theory, and the first candidate queue length is determined as the maximum possible queue length. If N is greater than or equal to 2, and the first quantity is M, and the second quantity is NM, then based on the trajectory data of M connected vehicles, the assembly wave reconstruction and dissipation wave reconstruction are performed. According to the shock wave theory, the first candidate queue length is calculated based on the reconstructed assembly wave and dissipation wave, and NM second candidate queue lengths are calculated based on the trajectory data of NM connected vehicles. The minimum value among the first candidate queue length and the NM second candidate queue lengths is determined as the maximum possible queue length; M is an integer greater than 1, and M is less than N. Specifically, based on the trajectory data of L connected vehicles, the assembly wave and dissipation wave are reconstructed. The first candidate queue length is calculated based on the reconstructed assembly wave and dissipation wave according to shock wave theory. L is equal to N, and N is greater than 1; or L is equal to M; including: Based on the distance from the position of the nth connected vehicle when it starts queuing to the intersection stop line, the time when the nth connected vehicle starts queuing, and the distance from the position of the (n-1)th connected vehicle when it starts queuing to the intersection stop line, the rally wave is reconstructed; where n represents the last vehicle in the N vehicles, and n-1 represents the vehicle before the last vehicle in the N vehicles. Based on the distance from the starting position of the nth connected vehicle to the stop line of the intersection, the starting time of the nth connected vehicle, and the distance from the starting position of the (n-1)th connected vehicle to the stop line of the intersection and the starting time of the (n-1)th connected vehicle determined from the trajectory data, dissipation wave reconstruction is performed.

2. The method according to claim 1, characterized in that, If L equals N, and N equals 1, then based on the trajectory data of the L networked vehicles, reconstructing the aggregate wave and reconstructing the dissipation wave, including: Based on the distance from the location where the connected vehicle begins to join the queue to the intersection stop line and the time when the connected vehicle begins to join the queue, determined from the trajectory data, a rallying wave reconstruction is performed. Based on the distance from the location where the connected vehicle starts to the intersection stop line and the time when the connected vehicle starts, determined from the trajectory data, a dissipation wave reconstruction is performed.

3. The method according to claim 1, wherein if L equals M, the step of calculating NM second candidate queue lengths based on the trajectory data of NM networked vehicles includes: For each of the NM connected vehicles, the second candidate queue length of the connected vehicles is calculated through the following steps: From the trajectory data, obtain the time difference between the time when the connected vehicle passes the stop line at the intersection and the time when the traffic light that allows passage is turned on in the current traffic signal cycle; Based on the reconstruction of the dissipated wave, the dissipated wave velocity and the free flow velocity are determined, and the reciprocals of the dissipated wave velocity and the free flow velocity are added together to obtain the addition result. The ratio of the time difference to the summation result is used as the second candidate queue length for the connected vehicles.

4. The method according to claim 1, characterized in that, The method further includes: The reconstructed shock wave velocity is corrected based on Bayesian inference. The reconstructed shock wave velocity includes the reconstructed condensation wave velocity and the reconstructed dissipation wave velocity.

5. The method according to claim 4, characterized in that, The Bayesian inference-based correction of the reconstructed shock wave velocity includes: Obtain the historical wave velocity of the previous W traffic signal cycles, and calculate the log-normal distribution of the wave velocity based on the W historical wave velocities using Bayesian inference; If the shock wave velocity needs to be corrected based on the log-normal distribution of wave velocity, then a new log-normal distribution is calculated based on W historical wave velocities and the shock wave velocity using Bayesian inference. The shock wave velocity is corrected based on the new log-normal distribution.

6. The method according to claim 1, characterized in that, If the vehicles arriving at the intersection within the current traffic signal cycle follow a Poisson distribution, the calculation of the probability compensation value based on the maximum possible queue length and the distance from the stopping point of the last connected vehicle among the N connected vehicles to the stop line at the intersection includes: Obtain the average headway and the number of lanes in the direction of travel corresponding to the intersection; The probability compensation value is obtained by inputting the maximum possible queue length, the distance from the parking point of the last connected vehicle to the intersection parking line, the average headway, and the number of lanes into the probability compensation value calculation rules.

7. The method according to claim 1, characterized in that, The estimation process, based on the queue length at the intersection and the arrival times of the N networked vehicles at the intersection stop line, yields the periodic traffic flow at the intersection within the current traffic signal cycle, including: The probability density function of arrival time is estimated based on the arrival time of the N networked vehicles at the intersection stop line using the kernel density estimation method. The rate of change is obtained by performing calculations based on the probability density function according to the rate of change calculation rules. Calculate the ratio of the queue length to the average vehicle distance, and multiply the ratio by the rate of change to obtain the periodic traffic flow within the current traffic signal cycle.

8. A traffic flow estimation method, characterized in that, include: Obtain P cycle traffic flows for P traffic signal cycles, wherein the cycle traffic flow for each traffic signal cycle is determined by the method described in any one of claims 1-7; P is an integer greater than or equal to 1. Based on the P cycles of traffic flow, the periodic traffic arrival rate is estimated; Based on the theory that periodic traffic flow follows a Poisson distribution, and using the traffic arrival rate, the periodic traffic flow within the future target time period is estimated within the traffic signal period for which no connected vehicles were detected.

9. The method according to claim 8, characterized in that, The method of estimating the periodic traffic arrival rate based on the P periods of traffic flow includes: Based on the P cycles of traffic flow, the prior distribution of traffic arrival rate is calculated using the Bayesian inference formula. The prior distribution of the traffic arrival rate is an exponential distribution, where the parameter of the exponential distribution is the mean of traffic flow over P periods. Obtain the initial traffic arrival rate, and iterate the target number based on the prior distribution and the initial traffic arrival rate to obtain the traffic arrival rate; For the j-th iteration, a random traffic arrival rate j is selected based on the prior distribution; j is an integer greater than or equal to 1 and less than the target number. The probability of the traffic arrival rate j being used as the traffic arrival rate is calculated according to the Monte Carlo rule. If the probability is greater than the probability of the traffic arrival rate j-1 determined in the (j-1)th time, then the traffic arrival rate j is taken as the traffic arrival rate; where, when j equals 1, the traffic arrival rate j-1 refers to the initial traffic arrival rate. If the probability is less than or equal to the probability of the traffic arrival rate j-1 determined in the (j-1)th time, then the traffic arrival rate j-1 is taken as the traffic arrival rate.

10. The method according to claim 9, characterized in that, The method further includes: The P-cycle traffic flow is used as prior information for Bayesian inference and time boundary identification to obtain multiple time boundaries. For each time limit, based on the traffic flow within the time limit, the periodic traffic arrival rate is estimated, and based on the theory that the traffic flow at each time limit follows a Poisson distribution, the traffic flow within each time limit is estimated based on the traffic arrival rate. Establish and store the correspondence between time limits, traffic arrival rates, and traffic flow.

11. A traffic flow estimation device for implementing the traffic flow estimation method as described in any one of claims 1-7, characterized in that, include: The acquisition unit is used to acquire trajectory data of N connected vehicles at the intersection during the current traffic signal cycle, and reconstruct the intersection shock wave based on the trajectory data of the N connected vehicles to obtain the maximum possible queue length of the intersection; N is an integer greater than or equal to 1. The calculation unit is used to calculate the probability compensation value based on the maximum possible queue length and the distance from the parking point of the last connected vehicle among the N connected vehicles to the intersection parking line; An estimation unit is used to estimate the queue length at the intersection based on the probability compensation value and the distance from the parking point of the last connected vehicle among the N connected vehicles to the parking line at the intersection. The processing unit is used to perform estimation processing based on the queue length of the intersection and the time when the N networked vehicles arrive at the stop line of the intersection to obtain the periodic traffic flow of the intersection in the current traffic signal cycle.

12. A traffic flow estimation device, characterized in that, include: The acquisition unit is used to acquire the traffic flow of P traffic signal cycles, wherein the traffic flow of each traffic signal cycle is determined by the method described in any one of claims 1-7; P is an integer greater than or equal to 1. An estimation unit is used to estimate the periodic-level traffic arrival rate based on the P periods of traffic flow. The estimation unit is also used to estimate the traffic flow in the future target time period within the traffic signal period for which no connected vehicles were detected, based on the traffic arrival rate, according to the theory that periodic traffic flow follows a Poisson distribution.

13. An electronic device, characterized in that, include: A processor is used to implement one or more computer programs; A computer storage medium storing one or more computer programs, said one or more computer programs being adapted to be loaded by a processor and executed by the traffic flow estimation method as described in any one of claims 1-7, or the traffic flow estimation method as described in any one of claims 8-10.

14. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the traffic flow estimation method as described in any one of claims 1-7, or the traffic flow estimation method as described in any one of claims 8-10.

Images