Traffic signal light control method based on multi-source heterogeneous data

Abstract

This invention provides a traffic signal control method based on multi-source heterogeneous data, which can be applied to the field of traffic signal control technology. The method includes: acquiring historical traffic information of multiple intersections within a historical period; inputting the historical traffic information into a prediction model to obtain predicted traffic parameters for multiple traffic flow phases at the current intersection in future periods and signal phase information of various traffic lights; updating the green light duration in the traffic control model based on the predicted traffic parameters and signal phase information until the queue length in the future period reaches a target value, thus obtaining the target green light duration and the target cycle duration for the future period. The traffic control model is constructed based on the queue length and congestion density of various traffic lights for multiple traffic flow phases in future periods; and controlling various traffic lights according to the target green light duration and target cycle duration.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of traffic signal control, and more particularly to a traffic signal lamp control method based on multi-source heterogeneous data. BACKGROUND

[0002] Multi-source heterogeneous traffic data of intersections collected through various sensors has been widely applied to traffic signal control. Multi-source heterogeneous traffic data increases the diversity of traffic information of intersections, which can analyze the traffic conditions of the intersections from multiple aspects, and is beneficial to the control of traffic signal lamps.

[0003] However, the traffic operation of the intersection is also affected by the external environment (such as the conflict of vehicle flow of other intersections), so the control accuracy of the traffic signal lamp is low, which easily causes traffic congestion and safety accidents, and reduces the traffic efficiency of the intersection. SUMMARY

[0004] In view of the above problems, the present application provides a traffic signal lamp control method based on multi-source heterogeneous data.

[0005] According to a first aspect of the present application, a traffic signal lamp control method based on multi-source heterogeneous data is provided, comprising: obtaining historical traffic information of a plurality of intersections in a historical period, wherein the plurality of intersections include a current intersection and an upstream intersection of the current intersection; inputting the historical traffic information into a prediction model to obtain predicted traffic parameters of a plurality of vehicle flow phases of the current intersection in a future period and signal phase information of a plurality of traffic signal lamps, wherein the plurality of vehicle flow phases represent directions in which vehicles can travel in the current intersection; updating a length of a green signal in a traffic control model based on the predicted traffic parameters and the signal phase information until a queue length in the future period reaches a target value, to obtain a target green signal length of the green signal and a target cycle length of the future period, wherein the plurality of traffic signal lamps include the green signal, the queue length represents a number of vehicles waiting to pass through the current intersection, and the traffic control model is constructed based on queue lengths and congestion densities of the plurality of traffic signal lamps in the future period for the plurality of vehicle flow phases; and controlling the plurality of traffic signal lamps according to the target green signal length and the target cycle length.

[0006] According to an embodiment of the present application, the predicted traffic parameters include a predicted arrival rate, a predicted saturation release rate, and a predicted congestion density, the predicted arrival rate represents a number of vehicles traveling to the current intersection within a preset time, and the signal phase information includes a signal phase sequence and a phase loss length.

[0007] According to an embodiment of the present invention, the historical cycle includes the current cycle, and the future cycle includes the next cycle of the current cycle. The traffic control model is constructed based on the following: determining the shock wave velocity based on the predicted arrival rate, predicted saturation release rate, and predicted congestion density of the next cycle; constructing sub-models of the queue lengths of various traffic lights based on the shock wave velocity, the queue length at the end of the current cycle, and the duration of the green light signal, so as to determine the queue lengths of various traffic lights at the current intersection in the next cycle; determining the first target duration and the second target duration in the next cycle based on the signal phase sequence, phase loss duration, and the duration of the green light signal; constructing a sub-model of the congestion density of the current intersection based on the first target duration, the second target duration, and the queue lengths of various traffic lights in the next cycle, so as to determine the congestion density of the current intersection in the next cycle; and constructing a traffic control model for the current intersection based on the sub-models of the queue lengths of various traffic lights and the congestion density sub-model.

[0008] According to an embodiment of the present invention, the signal phase sequence is a first red light phase, a green light phase, a yellow light phase, and a second red light phase. The signal phase information also includes the duration of the first red light phase and the duration of the second red light phase. Based on the shock wave velocity, the queue length at the end of the current cycle, and the duration of the green light signal, a sub-model of the queue length for each of the various traffic lights is constructed, including: determining the queue length of the first red light phase based on the shock wave velocity and the duration of the first red light phase, and constructing a sub-model of the cumulative queue length of the first red light phase based on the sum of the queue length of the first red light phase and the queue length at the end of the current cycle, so as to determine the cumulative queue length of the first red light phase at the current intersection in the next cycle. The system calculates the queue length based on the predicted saturation release rate, predicted congestion density, and green light signal duration. It then constructs a sub-model of the remaining queue length for the green light phase based on the cumulative queue length of the first red light phase and the remaining queue length of the green light phase, to determine the remaining queue length of the green light phase at the current intersection in the next cycle. Finally, it calculates the queue length for the second red light phase based on the shock wave velocity and the duration of the second red light phase, and constructs a sub-model of the cumulative queue length for the second red light phase based on the sum of the queue length of the second red light phase and the remaining queue length of the green light phase, to determine the cumulative queue length of the second red light phase at the current intersection in the next cycle.

[0009] According to an embodiment of the present invention, determining the first target duration and the second target duration in the next cycle based on the signal phase sequence, the phase loss duration, and the duration of the green light signal includes: determining the first target duration before the start of the green light phase and the second target duration after the end of the green light phase in the next cycle according to the signal phase sequence, the duration of the first red light phase, the duration of the green light signal, the duration of the yellow light phase, and the duration of the second red light phase, wherein the duration of the yellow light phase is the phase loss duration.

[0010] According to an embodiment of the present invention, a sub-model of the congestion density of the current intersection is constructed based on a first target duration, a second target duration, and the queue lengths of various traffic lights in the next cycle. This includes: determining a first congestion parameter for the first red light phase based on the first target duration, the queue length at the end of the current cycle, and the queue length of the first red light phase; determining a second congestion parameter for the second red light phase based on the second target duration, the remaining queue length of the green light phase, and the queue length of the second red light phase; and constructing a sub-model of congestion density based on the first congestion parameter, the second congestion parameter, and the predicted congestion density.

[0011] According to an embodiment of the present invention, updating the duration of the green light signal in the traffic control model based on predicted traffic parameters and signal phase information until the queue length in a future cycle reaches a target value, thereby obtaining the target green light duration and the target cycle duration for the future cycle, includes: inputting predicted traffic parameters and signal phase information into the traffic control model; updating the duration of the green light signal in the traffic control model based on the constraint that the congestion density in the traffic control model is less than or equal to the critical density, until the cumulative queue length of the second red light phase in multiple consecutive preset cycles in the future cycle reaches a target value, thereby obtaining the target green light duration; and determining the target cycle duration based on the target green light duration, the phase loss duration, the duration of the first red light phase, and the duration of the second red light phase.

[0012] According to an embodiment of the present invention, the prediction model includes a forget gate, an input gate, an output gate, and a memory unit. The prediction model is trained in the following manner: traffic training information of multiple training intersections during the training period is input into the initial prediction model to obtain the initial prediction value of the current training intersection. The initial prediction value includes initial traffic parameters and initial signal phase information. The multiple training intersections include the current training intersection and its upstream training intersection. The model parameters of the initial prediction model are updated based on the error between the initial prediction value and the true value until the error meets a preset condition to obtain the prediction model. The true value includes the true traffic parameters and true signal phase information of the current training intersection.

[0013] According to an embodiment of the present invention, the model parameters include the weights and biases of the forget gate, input gate, output gate, and memory unit, respectively; the model parameters of the initial prediction model are updated based on the error between the initial predicted value and the true value until the error meets a preset condition to obtain the prediction model, including: determining the gradient of the t-th model parameter according to the t-th updated model parameter, the step size coefficient in the gradient descent algorithm, and the t-th error, where t is an integer greater than or equal to 1; updating the t-th model parameter according to the t-th model parameter, the gradient, and the step size coefficient to obtain the t+1-th updated model parameter.

[0014] According to an embodiment of the present invention, historical traffic information includes: historical driving speeds of vehicles passing through multiple intersections collected via inductive loop detectors, historical traffic flow at multiple intersections collected via vehicle-mounted locators, and historical queue lengths at multiple intersections collected via roadside cameras.

[0015] A second aspect of the present invention provides a traffic signal control device based on multi-source heterogeneous data, comprising: an acquisition module for acquiring historical traffic information of multiple intersections within a historical period, wherein the multiple intersections include the current intersection and its upstream intersection; an input module for inputting the historical traffic information into a prediction model to obtain predicted traffic parameters for multiple traffic flow phases of the current intersection in future periods and signal phase information of multiple traffic lights, wherein the multiple traffic flow phases represent the directions in which vehicles can travel in the current intersection; an update module for updating the duration of the green light signal in the traffic control model based on the predicted traffic parameters and signal phase information until the queue length in the future period reaches a target value, thereby obtaining the target green light duration and the target period duration of the future period, wherein the multiple traffic lights include the green light signal, the queue length represents the number of vehicles waiting to pass through the current intersection, and the traffic control model is constructed based on the queue length and congestion density of the multiple traffic lights for each of the multiple traffic flow phases in the future period; and a control module for controlling the multiple traffic lights according to the target green light duration and the target period duration.

[0016] A third aspect of the present invention provides an electronic device comprising: one or more processors; and a memory for storing one or more computer programs, wherein the one or more processors execute the one or more computer programs to implement the steps of the method described above.

[0017] A fourth aspect of the present invention also provides a computer-readable storage medium having a computer program or instructions stored thereon, wherein the computer program or instructions, when executed by a processor, implement the steps of the above-described method.

[0018] A fifth aspect of the present invention also provides a computer program product, including a computer program or instructions that, when executed by a processor, implement the steps of the above-described method.

[0019] According to embodiments of the present invention, by inputting historical traffic information from the current intersection and upstream intersections into a prediction model, and considering the impact of upstream intersections on the current intersection's traffic flow, the predicted traffic parameters and signal phase information for each of the multiple traffic flow phases at the current intersection in future cycles can be accurately obtained. Since the traffic control model is constructed based on the queue lengths and congestion densities of various traffic signals for each of the multiple traffic flow phases in future cycles, the queue lengths and congestion densities in the traffic control model accurately reflect the dynamic traffic conditions of the intersection. Therefore, the duration of the green light signal in the traffic control model is updated based on the predicted traffic parameters and signal phase information until the queue length in the future cycle reaches the target value, obtaining the target green light duration and the target cycle duration for the future cycle. This achieves high-precision traffic flow prediction and signal timing optimization based on the dynamic traffic conditions of the intersection. Controlling multiple traffic lights according to the target green light duration and target cycle duration ensures that vehicles can pass efficiently at the current intersection, improving the current intersection's traffic efficiency. Attached Figure Description

[0020] The above-described features, other objects, and advantages of the present invention will become clearer from the following description of embodiments of the invention with reference to the accompanying drawings, in which:

[0021] Figure 1 The diagram illustrates an application scenario of a traffic signal control method based on multi-source heterogeneous data according to an embodiment of the present invention.

[0022] Figure 2 A flowchart of a traffic signal control method based on multi-source heterogeneous data according to an embodiment of the present invention is shown;

[0023] Figure 3 Traffic flow heatmaps at different times at the current intersection are shown according to embodiments of the present invention;

[0024] Figure 4 The prediction results at the current intersection are shown according to an embodiment of the present invention;

[0025] Figure 5A Simulation results for signal timing according to embodiments of the present invention are shown;

[0026] Figure 5B A structural block diagram of a traffic signal control system based on multi-source heterogeneous data according to an embodiment of the present invention is shown;

[0027] Figure 6 A structural block diagram of a traffic signal control device based on multi-source heterogeneous data according to an embodiment of the present invention is shown;

[0028] Figure 7A block diagram of an electronic device suitable for implementing a traffic signal control method based on multi-source heterogeneous data according to an embodiment of the present invention is shown. Detailed Implementation

[0029] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the invention. In the following detailed description, numerous specific details are set forth to provide a thorough understanding of the embodiments of the invention for ease of explanation. However, it will be apparent that one or more embodiments may be practiced without these specific details. Furthermore, descriptions of well-known structures and techniques are omitted in the following description to avoid unnecessarily obscuring the concept of the invention.

[0030] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the invention. The terms “comprising,” “including,” etc., as used herein indicate the presence of the stated features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0031] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein are to be interpreted in a manner consistent with the context of this specification, and not in an idealized or overly rigid way.

[0032] When using expressions such as "at least one of A, B and C", they should generally be interpreted in accordance with the meaning that is commonly understood by those skilled in the art (e.g., "a system having at least one of A, B and C" should include, but is not limited to, a system having A alone, a system having B alone, a system having C alone, a system having A and B, a system having A and C, a system having B and C, and / or a system having A, B and C, etc.).

[0033] Traffic signal control using multi-source heterogeneous data collected from roadside cameras, vehicle-mounted sensors, ground loops, and other environmental monitoring equipment has become widely used. However, the duration of traffic signals is often affected by unforeseen events, which may cause traffic congestion and safety accidents.

[0034] One implementation method is a centralized signal timing control method, which requires massive amounts of real-time traffic information and high-performance computing resources. It has high requirements for the types and quantities of data, as well as computing performance. Therefore, it is not suitable for real-time control of large-scale traffic networks, and it is even more difficult to accurately control the duration of traffic light signals in order to achieve efficient vehicle passage at intersections.

[0035] Another implementation method is the distributed signal timing control method, which uses roadside controllers and vehicle-to-everything (V2X) technology. It can share data from multiple intersections in the area in real time and is more suitable for traffic coordination optimization of multiple intersections. However, it is difficult to guarantee local optimization of traffic at a single intersection, and the actual control accuracy of traffic signals is low, which reduces the traffic efficiency of intersections.

[0036] In view of this, embodiments of the present invention provide a traffic signal control method based on multi-source heterogeneous data, comprising: acquiring historical traffic information of multiple intersections in a historical period, wherein the multiple intersections include the current intersection and its upstream intersection; inputting the historical traffic information into a prediction model to obtain predicted traffic parameters for multiple traffic flow phases of the current intersection in future periods and signal phase information of multiple traffic lights, wherein the multiple traffic flow phases represent the directions in which vehicles can travel in the current intersection; updating the duration of the green light signal in the traffic control model based on the predicted traffic parameters and signal phase information until the queue length in the future period reaches a target value, thereby obtaining the target green light duration and the target period duration of the future period, wherein the multiple traffic lights include the green light signal, the queue length represents the number of vehicles waiting to pass through the current intersection, and the traffic control model is constructed based on the queue length and congestion density of multiple traffic lights for each of the multiple traffic flow phases in the future period; and controlling the multiple traffic lights according to the target green light duration and the target period duration.

[0037] Figure 1 The diagram illustrates an application scenario of a traffic signal control method based on multi-source heterogeneous data according to an embodiment of the present invention.

[0038] like Figure 1 As shown, application scenario 100 according to this embodiment may include a vehicle 110, a traffic light 120 at an upstream intersection, and a traffic light 130 at the current intersection. The driver will drive the vehicle through the upstream intersection according to the instructions of the traffic light 120. After passing the upstream intersection, the current intersection has multiple traffic flow phases, such as a left-turn phase, a straight-ahead phase, and a right-turn phase. The driver will then pass through the current intersection according to the instructions of the traffic light 130. The traffic light 130 at the current intersection may include red, green, and yellow lights. The traffic light 130 at the current intersection may have multiple signal phases; for example, it may be in a green light phase when the green light is on, a yellow light phase when the yellow light is on, and a red light phase when the red light is on.

[0039] For example, when traffic light 130 at the current intersection shows a green light, vehicle 110 can proceed straight through along its lane. If the green light phase allows a right turn, the vehicle can turn into the right lane. It should be noted that traffic light 130 at the current intersection can be set to a left-turn green light; when the left-turn green light is on, vehicle 110 can turn into the left lane.

[0040] For example, when traffic light 130 at the current intersection is red, vehicles going straight or turning left must wait behind the stop line, while vehicles turning right may turn slowly. When traffic light 130 at the current intersection is yellow, vehicles that have already crossed the stop line into the current intersection (such as vehicles going straight or turning left) must continue to proceed, while vehicles that have not crossed the stop line must slow down and stop behind the stop line.

[0041] A roadside camera can be installed on traffic light 130 at the current intersection to collect the queue length of vehicles waiting to cross. Inductive loop detectors can be installed on the road between the upstream intersection and the current intersection to collect the speed of vehicles passing through the current intersection. Vehicle locators can be installed on vehicle 110 and other vehicles to collect traffic flow data at the current intersection. It should be noted that traffic flow data can also be obtained using inductive loop detectors.

[0042] It should be noted that, Figure 1 The number of vehicles 110 shown is merely illustrative. In reality, there could be multiple vehicles, and different vehicles might be traveling in different directions. The traffic light 130 at the current intersection can be configured with a green light for left-turning traffic as needed, and there are no restrictions here.

[0043] Figure 2 A flowchart of a traffic signal control method based on multi-source heterogeneous data according to an embodiment of the present invention is shown.

[0044] like Figure 2 As shown, the traffic signal control method based on multi-source heterogeneous data in this embodiment includes operations S210 to S240.

[0045] In operation S210, historical traffic information of multiple intersections within a historical period is obtained, including the current intersection and its upstream intersections.

[0046] According to embodiments of the present invention, the plurality of intersections may include the current intersection and a plurality of upstream intersections of the current intersection. For example, such as Figure 1 The upstream intersection of the current intersection shown is a T-junction. The upstream intersection of a T-junction can include two crossroads (…). Figure 1(Not shown). Therefore, the upstream intersection of the current intersection may include one T-junction and two cross-junctions.

[0047] According to an embodiment of the present invention, the historical period can be an adjacent period to the future period. For example, the current period is the 5th period, the historical period can be the 1st to 5th periods, and the future period can be the 6th period.

[0048] For example, a historical cycle can include multiple historical sub-cycles, and a future cycle can also include multiple future sub-cycles. The current cycle is the 5th sub-cycle, the historical cycles can be sub-cycles 1 to 5, and the future cycles can be sub-cycles 6 to 7.

[0049] It should be noted that the cycle is determined according to the type of traffic light signal. For example, the duration of the cycle can be the duration after the traffic light flashes in sequence according to the indication of each signal.

[0050] According to embodiments of the present invention, historical traffic information may include historical traffic information of the current intersection and historical traffic information of multiple upstream intersections. For example, historical traffic communication information may include the duration of each traffic signal, vehicle speed, number of vehicles after each signal ends, and congestion duration at the intersection.

[0051] In operation S220, historical traffic information is input into the prediction model to obtain the predicted traffic parameters of multiple traffic flow phases at the current intersection in the future cycle and the signal phase information of multiple traffic lights. Among them, multiple traffic flow phases represent the directions in which vehicles can travel at the current intersection.

[0052] According to embodiments of the present invention, multiple traffic flow phases may include straight-ahead phases, left-turn phases, right-turn phases, etc. Furthermore, multiple traffic flow phases are not limited to straight-ahead phases, left-turn phases, and right-turn phases; various left-turn phases and various right-turn phases can be set according to the actual intersection conditions.

[0053] According to embodiments of the present invention, the prediction model can be a machine learning algorithm, such as a support vector machine algorithm, a random forest algorithm, or a neural network algorithm. For example, historical traffic information of multiple traffic flow phases at the current intersection and historical traffic information of multiple traffic flow phases at the upstream intersection are processed according to time information to obtain traffic state features. The traffic state features are then input into the prediction model to obtain the predicted traffic parameters for each of the multiple traffic flow phases at the current intersection in future periods and the signal phase information of various traffic lights.

[0054] According to an embodiment of the present invention, predicted traffic parameters can characterize the traffic status of the current intersection in a future period.

[0055] According to embodiments of the present invention, the signal phase is determined based on multiple signals of a traffic light. The signal phase information can be information from multiple traffic lights. For example, the multiple signals of a traffic light may include a green light signal. For example, the signal phase information of multiple traffic lights may include the flashing sequence and flashing duration of the multiple traffic lights, etc.

[0056] In operation S230, the duration of the green light signal in the traffic control model is updated based on predicted traffic parameters and signal phase information until the queue length in the future cycle reaches the target value, thus obtaining the target green light duration and the target cycle duration for the future cycle. Among them, there are multiple traffic lights including green lights, and the queue length represents the number of vehicles waiting to pass through the current intersection. The traffic control model is constructed based on the queue length and congestion density of multiple traffic lights in the future cycle for each of the multiple traffic flow phases.

[0057] According to embodiments of the present invention, congestion density is used to describe the critical index at which traffic flow shifts from "smooth" to "congested". The numerical range of congestion density is greater than the critical density and less than the blockage density. The critical density can be the critical point at which the intersection has the highest traffic efficiency. The blockage density can be the density when vehicles at the intersection are completely stationary.

[0058] According to an embodiment of the present invention, since the queue length and congestion density of various traffic signals in the future cycle of various traffic flow phases can accurately reflect the dynamic traffic situation of the intersection, a traffic control model is constructed based on the queue length and congestion density of various traffic signals in the future cycle of various traffic flow phases.

[0059] According to embodiments of the present invention, the target value can minimize the cumulative delay over future cycles. For example, after repeatedly updating the duration of the green light signal in the traffic control model, multiple queue lengths are obtained. The target value can be the minimum of these multiple queue lengths.

[0060] When operating S240, various traffic lights are controlled according to the target green light duration and target cycle duration.

[0061] According to an embodiment of the present invention, the green light in a variety of traffic lights flashes according to a target green light duration. The duration of the sequential flashing of the various traffic lights is the target cycle duration.

[0062] For example, an n-order phase-triggered discrete optimization algorithm can be used to dynamically adjust the duration of the green light signal for each traffic flow phase in the traffic control model to achieve the global optimum of intersection traffic efficiency (such as minimum delay and maximum throughput).

[0063] According to embodiments of the present invention, by inputting historical traffic information from the current intersection and upstream intersections into a prediction model, and considering the impact of upstream intersections on the current intersection's traffic flow, the predicted traffic parameters and signal phase information for each of the multiple traffic flow phases at the current intersection in future cycles can be accurately obtained. Since the traffic control model is constructed based on the queue lengths and congestion densities of various traffic signals for each of the multiple traffic flow phases in future cycles, the queue lengths and congestion densities in the traffic control model accurately reflect the dynamic traffic conditions of the intersection. Therefore, the duration of the green light signal in the traffic control model is updated based on the predicted traffic parameters and signal phase information until the queue length in the future cycle reaches the target value, obtaining the target green light duration and the target cycle duration for the future cycle. This achieves high-precision traffic flow prediction and signal timing optimization based on the dynamic traffic conditions of the intersection. Controlling multiple traffic lights according to the target green light duration and target cycle duration ensures that vehicles can pass efficiently at the current intersection, improving the current intersection's traffic efficiency.

[0064] According to an embodiment of the present invention, historical traffic information includes: historical driving speeds of vehicles passing through multiple intersections collected via inductive loop detectors, historical traffic flow at multiple intersections collected via vehicle-mounted locators, and historical queue lengths at multiple intersections collected via roadside cameras.

[0065] According to embodiments of the present invention, inductive loop detectors can be installed at the current intersection and the upstream intersection to obtain the historical driving speed of vehicles at the intersection. Roadside cameras are installed along the roadside at the current intersection and the upstream intersection to collect the historical queue lengths at the current intersection and the upstream intersection.

[0066] It should be noted that the traffic information involved in the technical solution of the present invention is all information and data authorized by the driver or fully authorized by all parties. The collection, storage, use, processing, transmission, provision, disclosure and application of the relevant data all comply with relevant laws, regulations and standards, necessary confidentiality measures have been taken, and they do not violate public order and good morals. Corresponding operation interfaces are provided for drivers to choose to authorize or refuse.

[0067] According to an embodiment of the present invention, traffic state characteristics under multi-source heterogeneous data are determined based on the historical driving speed, historical traffic flow and historical queue length of vehicles at multiple intersections.

[0068] Based on vehicle-mounted locators, historical traffic flow data for the current intersection at the current time and previous times are obtained. Using time-series analysis algorithms, historical traffic flow data for the upstream intersection at the current time and previous times are obtained. The historical traffic flow data, historical traffic flow, and historical queue length are fused to form traffic state features under multi-source monitoring, possessing real-time performance and robustness. These traffic state features are then used as the input sequence to the prediction model.

[0069] In one embodiment, the prediction model may include an intersection control agent, multiple sub-region control agents, and a network control agent, with each of the multiple sub-regions (such as multiple sub-regions including the current intersection and multiple upstream intersections) having its own corresponding sub-region control agent.

[0070] Intersection control agents are used to collect multi-source heterogeneous information. At the intersection, various sensors are deployed to collect data such as traffic flow, speed, and queue length in real time. This heterogeneous data is then cleaned, transformed, and noise and redundancy are removed, and initially fused into usable traffic state features.

[0071] The sub-region control agent is responsible for aggregation and coordination within its sub-region. It collects traffic state characteristics from each intersection within the sub-region and compares and analyzes the traffic conditions at different intersections. Based on traffic demand and the overall situation, it coordinates and optimizes the control strategies for each intersection to ensure orderly traffic within the sub-region.

[0072] The network control agent integrates multi-source heterogeneous data fusion control models at the network-wide level. It integrates information from all sub-regions to assess traffic conditions from a global perspective. Based on the assessment results, it uniformly adjusts and optimizes the control strategies for each sub-region.

[0073] The intelligent agent possesses autonomous decision-making capabilities and can work collaboratively. The intersection control agent is responsible for locally collecting multi-source heterogeneous information such as traffic flow, speed, and queue length, which is then cleaned, transformed, and initially fused. Based on regional traffic characteristics, sub-region control agents are established to collect information from each intersection, compare different states, and then further summarize and coordinate. Based on the information from all sub-regions of the network, a network control agent is constructed, and global adjustments and optimizations are performed to form a multi-source heterogeneous data fusion control model.

[0074] In another embodiment, the historical traffic flow of the upstream intersection is obtained based on the vehicle locator as shown in the following formula (1):

[0075] (1);

[0076] yes Intersection (upstream intersection) No. Historical traffic flow at any given time yes Intersection No. Historical traffic flow at any given time yes Intersection No. Historical traffic flow at any given time It is an integer greater than 1. For integers greater than 2, from 1 to A moment is a period of time within a historical cycle.

[0077] The historical traffic flow at the current intersection is obtained based on the vehicle locator, as shown in the following formula (2):

[0078] (2);

[0079] yes Intersection (current intersection) No. Historical traffic flow at any given time yes Intersection No. Historical traffic flow at any given time yes Intersection No. Historical traffic flow at any given moment.

[0080] Historical traffic flow data, historical traffic volume, and historical queue length are fused together to form traffic state characteristics under multi-source monitoring, possessing real-time performance and robustness. These traffic state characteristics are then used as input sequences into the prediction model.

[0081] The prediction model can be a long short-term memory network, while the traffic control model can be constructed from the queue lengths and congestion densities of various traffic lights in future cycles for different traffic flow phases. The queue lengths and congestion densities are derived from traffic flow wave theory. Therefore, combining long short-term memory networks and traffic flow wave models provides clear numerical definitions and enables end-to-end joint optimization of traffic flow prediction and signal timing.

[0082] According to an embodiment of the present invention, the predicted traffic parameters include the predicted arrival rate, the predicted saturation release rate, and the predicted congestion density. The predicted arrival rate represents the number of vehicles approaching the current intersection within a preset time. The signal phase information includes the signal phase sequence and the phase loss duration.

[0083] According to an embodiment of the present invention, the predicted saturation release rate characterizes the first target (maximum) number of vehicles under the continuous traffic flow condition during the duration of the green light signal.

[0084] According to an embodiment of the present invention, congestion density characterizes the second target (maximum) number of vehicles that the current intersection can accommodate when vehicles are stopped at the current intersection.

[0085] According to an embodiment of the present invention, the signal phase sequence can be the order in which multiple traffic lights flash sequentially. For example, if multiple traffic lights include red, green, and yellow lights, and they flash sequentially in the order of green, yellow, and red, then the signal phase sequence can be the red phase, green phase, and yellow phase.

[0086] According to an embodiment of the present invention, the phase loss duration can be the yellow light phase duration.

[0087] According to an embodiment of the present invention, the historical cycle includes the current cycle, and the future cycle includes the next cycle of the current cycle. The traffic control model is constructed based on the following: determining the shock wave velocity based on the predicted arrival rate, predicted saturation release rate, and predicted congestion density of the next cycle; constructing sub-models of the queue lengths of various traffic lights based on the shock wave velocity, the queue length at the end of the current cycle, and the duration of the green light signal, so as to determine the queue lengths of various traffic lights at the current intersection in the next cycle; determining the first target duration and the second target duration in the next cycle based on the signal phase sequence, phase loss duration, and the duration of the green light signal; constructing a sub-model of the congestion density of the current intersection based on the first target duration, the second target duration, and the queue lengths of various traffic lights in the next cycle, so as to determine the congestion density of the current intersection in the next cycle; and constructing a traffic control model for the current intersection based on the sub-models of the queue lengths of various traffic lights and the congestion density sub-model.

[0088] According to an embodiment of the present invention, the shock wave velocity The formula is as follows: Formula (3):

[0089] (3);

[0090] k represents the next period. Indicates arrival rate, It is the predicted arrival rate of traffic phase p in the next cycle. Indicates the saturation release rate. It is the predicted saturation release rate of the current traffic flow phase p at the intersection during the duration of the green light signal. Indicates blockage density. It is the predicted congestion density of traffic flow phase p in the next cycle.

[0091] According to embodiments of the present invention, based on the theory of motion waves, and using the shock wave velocity, the queue length at the end of the current cycle, and the duration of the green light signal, sub-models of the queue lengths for various traffic lights are constructed. For example, the queue lengths for various traffic lights can be dynamically established according to the number of vehicles passing through at different signal phase durations.

[0092] According to an embodiment of the present invention, since the current traffic control model of the intersection is constructed using sub-models of the queue lengths of various traffic lights (based on motion wave theory) and sub-models of congestion density, the traffic control model can accurately reflect the dynamic traffic conditions of the intersection.

[0093] According to an embodiment of the present invention, the signal phase sequence is a first red light phase, a green light phase, a yellow light phase, and a second red light phase. The signal phase information also includes the duration of the first red light phase and the duration of the second red light phase. Based on the shock wave velocity, the queue length at the end of the current cycle, and the duration of the green light signal, a sub-model of the queue length for each of the various traffic lights is constructed, including: determining the queue length of the first red light phase based on the shock wave velocity and the duration of the first red light phase, and constructing a sub-model of the cumulative queue length of the first red light phase based on the sum of the queue length of the first red light phase and the queue length at the end of the current cycle, so as to determine the cumulative queue length of the first red light phase at the current intersection in the next cycle. The system calculates the queue length based on the predicted saturation release rate, predicted congestion density, and green light signal duration. It then constructs a sub-model of the remaining queue length for the green light phase based on the cumulative queue length of the first red light phase and the remaining queue length of the green light phase, to determine the remaining queue length of the green light phase at the current intersection in the next cycle. Finally, it calculates the queue length for the second red light phase based on the shock wave velocity and the duration of the second red light phase, and constructs a sub-model of the cumulative queue length for the second red light phase based on the sum of the queue length of the second red light phase and the remaining queue length of the green light phase, to determine the cumulative queue length of the second red light phase at the current intersection in the next cycle.

[0094] According to an embodiment of the present invention, the sub-model of the queue length at the end of the current cycle is shown in formula (4), the sub-model of the cumulative queue length of the first red light phase is shown in formula (5), the sub-model of the remaining queue length of the green light phase is shown in formula (6), and the sub-model of the cumulative queue length of the second red light phase is shown in formula (7).

[0095] (4);

[0096] Queue length at the end of the current period It is the queue length at the end of the previous cycle in the current cycle. .

[0097] (5);

[0098] It is the cumulative queue length of the first red light phase. It is the duration of the first red light phase. (i.e., the duration of the first red light phase) With shock wave velocity The product of ( ) is the queue length for the first red light phase.

[0099] (6);

[0100] It is the remaining queue length for the green light phase. The duration of the green light signal. This refers to the passage length during the green light phase.

[0101] (7);

[0102] It is the cumulative queue length for the second red light phase. The duration of the second red light phase. This refers to the queue length for the second red light phase.

[0103] According to an embodiment of the present invention, determining the first target duration and the second target duration in the next cycle based on the signal phase sequence, the phase loss duration, and the duration of the green light signal includes: determining the first target duration before the start of the green light phase and the second target duration after the end of the green light phase in the next cycle according to the signal phase sequence, the duration of the first red light phase, the duration of the green light signal, the duration of the yellow light phase, and the duration of the second red light phase, wherein the duration of the yellow light phase is the phase loss duration.

[0104] According to an embodiment of the present invention, by predicting the duration of the first red light phase, the duration of the green light signal, the duration of the yellow light phase (phase loss duration), and the signal phase sequence in the next cycle, the first target duration before the start of the green light phase and the second target duration after the end of the green light phase in the next cycle can be determined.

[0105] According to an embodiment of the present invention, a sub-model of the congestion density of the current intersection is constructed based on a first target duration, a second target duration, and the queue lengths of various traffic lights in the next cycle. This includes: determining a first congestion parameter for the first red light phase based on the first target duration, the queue length at the end of the current cycle, and the queue length of the first red light phase; determining a second congestion parameter for the second red light phase based on the second target duration, the remaining queue length of the green light phase, and the queue length of the second red light phase; and constructing a sub-model of congestion density based on the first congestion parameter, the second congestion parameter, and the predicted congestion density.

[0106] According to an embodiment of the present invention, the sub-model of congestion density is shown in formula (8):

[0107] (8);

[0108] , ,Right now Let this be the queue length for the first red light phase of the next cycle k. This represents the queue length for the second red light phase within the next cycle k. The first congestion parameter is... , The first target duration within the next cycle k is [duration]. The second congestion parameter is [parameter]. , The second target duration is within the next cycle k. N represents the number of cycles in the predicted future cycles. P represents the number of traffic flow phases at the current intersection.

[0109] Traffic phase Congestion measurement in the next cycle (k). This indicates the congestion density for the future cycle.

[0110] According to an embodiment of the present invention, updating the duration of the green light signal in the traffic control model based on predicted traffic parameters and signal phase information until the queue length in a future cycle reaches a target value, thereby obtaining the target green light duration and the target cycle duration for the future cycle, includes: inputting predicted traffic parameters and signal phase information into the traffic control model; updating the duration of the green light signal in the traffic control model based on the constraint that the congestion density in the traffic control model is less than or equal to the critical density, until the cumulative queue length of the second red light phase in multiple consecutive preset cycles in the future cycle reaches a target value, thereby obtaining the target green light duration; and determining the target cycle duration based on the target green light duration, the phase loss duration, the duration of the first red light phase, and the duration of the second red light phase.

[0111] Given the predicted arrival rate, predicted saturation release rate, predicted congestion density, signal phase sequence, and phase loss duration (yellow) for each traffic flow phase, the length of this cycle is determined. Duration of green light signals in each phase Because the objective function (Equations (4) to (8)) has a significant impact on... All are linear, resulting in a linear programming problem, and the solution is... and The optimal cycle length and green light duration are naturally determined by the timing. The green, yellow, and red light durations within the cycle length are then arranged on a timeline and triggered by phase, ensuring that "first red light duration + green light duration + phase loss duration + second red light duration = cycle length". After each round of rolling cycle control is completed, new data is read again to initiate the next round of optimal traffic flow control.

[0112] The queuing growth rate was determined using the motion wave theory, and a globally optimal traffic control model was established using the motion wave theory (formulas (4) to (7)) and the flow density map (formula (8)). The green light allocation rules were determined using dynamic programming. Then, the congestion density under the critical phase state was calculated, and the optimal cycle length and the duration of the green light signal for each phase were solved to achieve optimal traffic flow.

[0113] According to an embodiment of the present invention, based on the constraint that the congestion density in the traffic control model is less than or equal to the critical density, the duration of the green light signal in the traffic control model is updated until the cumulative queue length of the second red light phase in N consecutive preset cycles in the future reaches the minimum value (cumulative delay is minimized), thereby triggering the adjustment of the traffic control model by formulas (4) to (8). Solve the problem to obtain the target green light duration, achieve the optimal traffic flow at the current intersection, and thus improve the traffic efficiency of multiple intersections within the area.

[0114] According to an embodiment of the present invention, the prediction model includes a forget gate, an input gate, an output gate, and a memory unit. The prediction model is trained in the following manner: traffic training information of multiple training intersections during the training period is input into the initial prediction model to obtain the initial prediction value of the current training intersection. The initial prediction value includes initial traffic parameters and initial signal phase information. The multiple training intersections include the current training intersection and its upstream training intersection. The model parameters of the initial prediction model are updated based on the error between the initial prediction value and the true value until the error meets a preset condition to obtain the prediction model. The true value includes the true traffic parameters and true signal phase information of the current training intersection.

[0115] According to an embodiment of the present invention, the initial prediction model can be a pre-trained long short-term memory network, which is constructed according to the following formula, including a forget gate, an input gate, an output gate, and memory units:

[0116] (9);

[0117] To input traffic state features into the Long Short-Term Memory network, These are the hidden state features obtained from the last update. It is the Gate of Oblivion. It's an input gate. It's an output gate. It is a memory unit. It is a retained state feature of the memory unit output during the last update. This refers to the retained state characteristics of the memory unit output during this update process. These are the hidden state features obtained in this update. The activation function is usually the sigmoid function, while tanh is the hyperbolic tangent activation function. These are the weights of the forget gate, input gate, output gate, and memory unit, respectively. These are the biases of the forget gate, input gate, output gate, and memory unit, respectively.

[0118] According to an embodiment of the present invention, during the training phase of the prediction model, gradient descent is used to optimize the parameters of the long short-term memory network, so that the error between the initial predicted value and the true value converges (or reaches its minimum). The preset condition can be the initial predicted value... Compared with the true value The error between them converges (or reaches its minimum). The formula is as follows:

[0119] (10);

[0120] According to an embodiment of the present invention, the model parameters include the weights and biases of the forget gate, input gate, output gate, and memory unit, respectively; the model parameters of the initial prediction model are updated based on the error between the initial predicted value and the true value until the error meets a preset condition to obtain the prediction model, including: determining the gradient of the t-th model parameter according to the t-th updated model parameter, the step size coefficient in the gradient descent algorithm, and the t-th error, where t is an integer greater than or equal to 1; updating the t-th model parameter according to the t-th model parameter, the gradient, and the step size coefficient to obtain the t+1-th updated model parameter.

[0121] Update the weights and biases of the forget gate, input gate, output gate, and memory unit according to the following formula.

[0122] (11);

[0123] It is the first The weights of the short-term memory network are updated every time. It is the first The weights of the short-term memory network are updated once more. This is the step size coefficient, which controls the magnitude of the parameter update along the negative gradient direction. for The gradient. For the first The bias of the short-term memory network during each update. It is the first +1 update duration for the short-term memory network bias. for The gradient.

[0124] In the implementation of model training, a Long Short-Term Memory (LSTM) network is established, which includes a forget gate, input gate, output gate, and memory units. The gating mechanism is a crucial element in ensuring model stability. However, manually tuning the parameters during the model training phase is impractical. Therefore, gradient descent is used to optimize the weights and biases of the LSM network to achieve the initial predicted values. Compared with the true value The error between them converges (or reaches its minimum). The model optimization process involves storing the parameters of the weights, biases, and gate states from the Long Short-Term Memory network update process, based on the current error function.

[0125] For example, the area where the current intersection is located was modeled and calibrated in the VISSIM (Verkehr In Städten – Simulation, Urban Traffic Simulation System) simulation environment. The area can include 5 approach lanes and 13 movements, mainly two phase movements: straight and left turn.

[0126] Traffic flow datasets can be annual traffic data for the region from historical years, including indicators such as the number of vehicles passing through intersections, their speed, and vehicle type at different times. To achieve efficient and accurate data collection, traffic simulation software combines actual traffic data to construct an automated data collection and preprocessing workflow, removing outliers and missing values ​​to ensure the continuity and reliability of the dataset.

[0127] Figure 3 Traffic flow heatmaps at different times at the current intersection are shown according to an embodiment of the present invention.

[0128] like Figure 3 As shown, the traffic flow heat at different times in two months intuitively reflects the vehicle distribution characteristics of the intersection at different times, revealing that the vehicle distribution at the intersection exhibits non-linear fluctuations.

[0129] Figure 4 The prediction results at the current intersection are shown according to an embodiment of the present invention.

[0130] like Figure 4 As shown, data from July and August of historical years were used as the training set, and data from September were used for model testing and validation. The solid line (3) represents the actual data, and the dashed line represents the predicted data. Figure 4 The prediction results show that the error between the predicted traffic flow data and the actual data under different cycles is small, which verifies the robustness of the training and testing results based on the Long Short-Term Memory Network. It shows that the predicted trend of the model is highly consistent with the actual traffic flow trend, proving the effectiveness and reliability of the Long Short-Term Memory Network.

[0131] Figure 5ASimulation results for signal timing according to an embodiment of the present invention are shown.

[0132] like Figure 5A As shown, a statistical data set of 4000 cycles is used for prediction. The signal timing parameters obtained from solving the five models are input into the simulation environment. For each test method, 30 simulation experiments are conducted, and each simulation experiment takes 4900 seconds. Figure 5A This diagram illustrates the time distribution of average travel delays under five traffic signal control methods. The horizontal axis represents the average travel delay of vehicles, and the vertical axis represents the frequency of different delay times. Figure 5A The proposed model performs best in traffic signal control, with its average travel delay mainly concentrated in the lower range (approximately 40–55 seconds), exhibiting a narrow distribution and small fluctuations, indicating smoother traffic flow and less delay. In contrast, the traditional Webster and Maxband models perform poorly in terms of delay, with generally higher average travel delays and a wider distribution range. The Webster and Maxband models are ineffective in alleviating traffic congestion during peak hours. The Fuzzy model and the MDAE (Multi-Objective Dynamic Signal Timing) model perform moderately well in delay control, showing improvement compared to the Webster and Maxband models, but their overall performance still falls short of the proposed model.

[0133] Figure 5B A block diagram of a traffic signal control system based on multi-source heterogeneous data according to an embodiment of the present invention is shown.

[0134] like Figure 5B As shown, the traffic signal control system 500 based on multi-source heterogeneous data includes a data acquisition unit 510, a feature extractor 520, a processor 530, a memory 540, and a signal control device 550.

[0135] Data acquisition unit 510 is used to collect multi-source heterogeneous traffic data in real time from roadside sensors, vehicle-mounted sensors, inductive loops, and historical databases. The collected data is then transmitted to feature extractor 520.

[0136] The feature extractor 520 is used to denoise, fill in missing values ​​and detect anomalies in the raw data output by the data collector, calculate and output traffic state features, and its output traffic state feature vector will be transmitted to the processor 530.

[0137] The processor 530 is coupled to the feature extractor 520. The processor 530 is used to train and predict historical traffic state feature sequences online and transmit them to the memory 540 on schedule.

[0138] The memory 540 is coupled to the processor 530. The memory 540 is used to store historical traffic data, model parameters and weights, training samples, prediction results, and optimization intermediates required for system operation. The memory 540 is also used to store traffic state features output by the feature extractor 520.

[0139] The signal control device 550 receives the optimal timing scheme calculated by the processor to drive the traffic lights to switch according to the set sequence. The signal control device can be a traffic controller, traffic lights, etc. Multiple signal control devices can be used.

[0140] Figure 6 A structural block diagram of a traffic signal control device based on multi-source heterogeneous data according to an embodiment of the present invention is shown.

[0141] like Figure 6 As shown, the traffic signal control device 600 based on multi-source heterogeneous data in this embodiment includes an acquisition module 610, an input module 620, an update module 630, and a control module 640.

[0142] The acquisition module 610 is used to acquire historical traffic information of multiple intersections within a historical period, wherein the multiple intersections include the current intersection and its upstream intersections. In one embodiment, the acquisition module 610 can be used to perform the operation S210 described above, which will not be repeated here.

[0143] The input module 620 is used to input historical traffic information into the prediction model to obtain the predicted traffic parameters of multiple traffic flow phases at the current intersection for each future cycle and the signal phase information of various traffic lights. The multiple traffic flow phases represent the directions in which vehicles can travel at the current intersection. In one embodiment, the input module 620 can be used to perform the operation S220 described above, which will not be repeated here.

[0144] The update module 630 is used to update the duration of the green light signal in the traffic control model based on predicted traffic parameters and signal phase information, until the queue length in the future cycle reaches the target value, thus obtaining the target green light duration and the target cycle duration for the future cycle. Here, various traffic lights include green lights, and queue length represents the number of vehicles waiting to pass through the current intersection. The traffic control model is constructed based on the queue lengths and congestion densities of various traffic lights for different traffic flow phases in the future cycle. In one embodiment, the update module 630 can be used to perform the operation S230 described above, which will not be repeated here.

[0145] The control module 640 is used to control various traffic lights according to the target green light duration and the target cycle duration. In one embodiment, the control module 640 can be used to perform the operation S240 described above, which will not be repeated here.

[0146] According to an embodiment of the present invention, the predicted traffic parameters include the predicted arrival rate, the predicted saturation release rate, and the predicted congestion density. The predicted arrival rate represents the number of vehicles approaching the current intersection within a preset time. The signal phase information includes the signal phase sequence and the phase loss duration.

[0147] According to an embodiment of the present invention, the historical cycle includes the current cycle, and the future cycle includes the next cycle of the current cycle. The traffic control model is constructed based on the following: determining the shock wave velocity based on the predicted arrival rate, predicted saturation release rate, and predicted congestion density of the next cycle; constructing sub-models of the queue lengths of various traffic lights based on the shock wave velocity, the queue length at the end of the current cycle, and the duration of the green light signal, so as to determine the queue lengths of various traffic lights at the current intersection in the next cycle; determining the first target duration and the second target duration in the next cycle based on the signal phase sequence, phase loss duration, and the duration of the green light signal; constructing a sub-model of the congestion density of the current intersection based on the first target duration, the second target duration, and the queue lengths of various traffic lights in the next cycle, so as to determine the congestion density of the current intersection in the next cycle; and constructing a traffic control model for the current intersection based on the sub-models of the queue lengths of various traffic lights and the congestion density sub-model.

[0148] According to an embodiment of the present invention, the signal phase sequence is a first red light phase, a green light phase, a yellow light phase, and a second red light phase. The signal phase information also includes the duration of the first red light phase and the duration of the second red light phase. Based on the shock wave velocity, the queue length at the end of the current cycle, and the duration of the green light signal, a sub-model of the queue length for each of the various traffic lights is constructed, including: determining the queue length of the first red light phase based on the shock wave velocity and the duration of the first red light phase, and constructing a sub-model of the cumulative queue length of the first red light phase based on the sum of the queue length of the first red light phase and the queue length at the end of the current cycle, so as to determine the cumulative queue length of the first red light phase at the current intersection in the next cycle. The system calculates the queue length based on the predicted saturation release rate, predicted congestion density, and green light signal duration. It then constructs a sub-model of the remaining queue length for the green light phase based on the cumulative queue length of the first red light phase and the remaining queue length of the green light phase, to determine the remaining queue length of the green light phase at the current intersection in the next cycle. Finally, it calculates the queue length for the second red light phase based on the shock wave velocity and the duration of the second red light phase, and constructs a sub-model of the cumulative queue length for the second red light phase based on the sum of the queue length of the second red light phase and the remaining queue length of the green light phase, to determine the cumulative queue length of the second red light phase at the current intersection in the next cycle.

[0149] According to an embodiment of the present invention, determining the first target duration and the second target duration in the next cycle based on the signal phase sequence, the phase loss duration, and the duration of the green light signal includes: determining the first target duration before the start of the green light phase and the second target duration after the end of the green light phase in the next cycle according to the signal phase sequence, the duration of the first red light phase, the duration of the green light signal, the duration of the yellow light phase, and the duration of the second red light phase, wherein the duration of the yellow light phase is the phase loss duration.

[0150] According to an embodiment of the present invention, a sub-model of the congestion density of the current intersection is constructed based on a first target duration, a second target duration, and the queue lengths of various traffic lights in the next cycle. This includes: determining a first congestion parameter for the first red light phase based on the first target duration, the queue length at the end of the current cycle, and the queue length of the first red light phase; determining a second congestion parameter for the second red light phase based on the second target duration, the remaining queue length of the green light phase, and the queue length of the second red light phase; and constructing a sub-model of congestion density based on the first congestion parameter, the second congestion parameter, and the predicted congestion density.

[0151] According to an embodiment of the present invention, the update module 630 includes an update submodule and a determination submodule. The update submodule is used to input predicted traffic parameters and signal phase information into the traffic control model, and based on the constraint that the congestion density in the traffic control model is less than or equal to the critical density, update the duration of the green light signal in the traffic control model until the cumulative queue length of the second red light phase in multiple consecutive preset cycles in the future cycle reaches the target value, thereby obtaining the target green light duration; the determination submodule is used to determine the target cycle duration based on the target green light duration, the phase loss duration, the duration of the first red light phase, and the duration of the second red light phase.

[0152] According to an embodiment of the present invention, the prediction model includes a forget gate, an input gate, an output gate, and a memory unit. The prediction model is trained in the following manner: traffic training information of multiple training intersections during the training period is input into the initial prediction model to obtain the initial prediction value of the current training intersection. The initial prediction value includes initial traffic parameters and initial signal phase information. The multiple training intersections include the current training intersection and its upstream training intersection. The model parameters of the initial prediction model are updated based on the error between the initial prediction value and the true value until the error meets a preset condition to obtain the prediction model. The true value includes the true traffic parameters and true signal phase information of the current training intersection.

[0153] According to an embodiment of the present invention, the model parameters include the weights and biases of the forget gate, input gate, output gate, and memory unit, respectively; the model parameters of the initial prediction model are updated based on the error between the initial predicted value and the true value until the error meets a preset condition to obtain the prediction model, including: determining the gradient of the t-th model parameter according to the t-th updated model parameter, the step size coefficient in the gradient descent algorithm, and the t-th error, where t is an integer greater than or equal to 1; updating the t-th model parameter according to the t-th model parameter, the gradient, and the step size coefficient to obtain the t+1-th updated model parameter.

[0154] According to an embodiment of the present invention, historical traffic information includes: historical driving speeds of vehicles passing through multiple intersections collected via inductive loop detectors, historical traffic flow at multiple intersections collected via vehicle-mounted locators, and historical queue lengths at multiple intersections collected via roadside cameras.

[0155] According to embodiments of the present invention, any plurality of modules among the acquisition module 610, input module 620, update module 630, and control module 640 may be combined into one module, or any one of these modules may be split into multiple modules. Alternatively, at least a portion of the functionality of one or more of these modules may be combined with at least a portion of the functionality of other modules and implemented in one module. According to embodiments of the present invention, at least one of the acquisition module 610, input module 620, update module 630, and control module 640 may be at least partially implemented as hardware circuitry, such as a field-programmable gate array (FPGA), a programmable logic array (PLA), a system-on-a-chip, a system-on-a-substrate, a system-on-package, an application-specific integrated circuit (ASIC), or any other reasonable means of integrating or packaging circuitry, or implemented in software, hardware, or firmware, or in any suitable combination of any of these three implementation methods. Alternatively, at least one of the acquisition module 610, input module 620, update module 630, and control module 640 may be at least partially implemented as a computer program module, which, when run, can perform corresponding functions.

[0156] Figure 7 A block diagram of an electronic device suitable for implementing a traffic signal control method based on multi-source heterogeneous data according to an embodiment of the present invention is shown.

[0157] like Figure 7 As shown, an electronic device 700 according to an embodiment of the present invention includes a processor 530, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 702 or a program loaded from a storage portion 708 into a random access memory (RAM) 703. The processor 530 may include, for example, a general-purpose microprocessor (e.g., a CPU), an instruction set processor and / or an associated chipset and / or a special-purpose microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. The processor 530 may also include onboard memory for caching purposes. The processor 530 may include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of the present invention.

[0158] RAM 703 stores various programs and data required for the operation of electronic device 700. Processor 701, ROM 702, and RAM 703 are interconnected via bus 704. Processor 530 executes various operations of the method flow according to embodiments of the present invention by executing programs in ROM 702 and / or RAM 703. It should be noted that the programs may also be stored in one or more memories other than ROM 702 and RAM 703. Processor 530 may also execute various operations of the method flow according to embodiments of the present invention by executing programs stored in said one or more memories.

[0159] According to an embodiment of the present invention, the electronic device 700 may further include an input / output (I / O) interface 705, which is also connected to a bus 704. The electronic device 700 may also include one or more of the following components connected to the input / output (I / O) interface 705: an input section 706 including a keyboard, mouse, etc.; an output section 707 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 708 including a hard disk, etc.; and a communication section 709 including a network interface card such as a LAN card, modem, etc. The communication section 709 performs communication processing via a network such as the Internet. A drive 710 is also connected to the input / output (I / O) interface 705 as needed. A removable medium 711, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on the drive 710 as needed so that computer programs read from it can be installed into the storage section 708 as needed.

[0160] The present invention also provides a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiments; or it may exist independently and not assembled into the device / apparatus / system. The computer-readable storage medium carries one or more programs, which, when executed, implement the method according to the embodiments of the present invention.

[0161] According to embodiments of the present invention, a computer-readable storage medium may be a non-volatile computer-readable storage medium, such as including, but not limited to: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. For example, according to embodiments of the present invention, a computer-readable storage medium may include ROM 702 and / or RAM 703 and / or one or more memories other than ROM 702 and RAM 703 described above.

[0162] Embodiments of the present invention also include a computer program product comprising a computer program containing program code for performing the methods shown in the flowchart. When the computer program product is run on a computer system, the program code enables the computer system to implement the traffic signal control method based on multi-source heterogeneous data provided in the embodiments of the present invention.

[0163] When the computer program is executed by the processor 530, it performs the functions defined in the system / apparatus of this invention. According to embodiments of the invention, the systems, apparatuses, modules, units, etc., described above can be implemented by computer program modules.

[0164] In one embodiment, the computer program may rely on a tangible storage medium such as an optical storage device or a magnetic storage device. In another embodiment, the computer program may also be transmitted and distributed in the form of signals over a network medium, and may be downloaded and installed via the communication section 709, and / or installed from a removable medium 711. The program code contained in the computer program can be transmitted using any suitable network medium, including but not limited to: wireless, wired, etc., or any suitable combination thereof.

[0165] In such an embodiment, the computer program can be downloaded and installed from a network via the communication section 709, and / or installed from the removable medium 711. When the computer program is executed by the processor 530, it performs the functions defined in the system of this embodiment of the invention. According to embodiments of the invention, the systems, devices, apparatuses, modules, units, etc., described above can be implemented by computer program modules.

[0166] According to embodiments of the present invention, program code for executing the computer programs provided in the embodiments of the present invention can be written in any combination of one or more programming languages. Specifically, these computational programs can be implemented using high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. Programming languages ​​include, but are not limited to, languages ​​such as Java, C++, Python, "C", or similar programming languages. The program code can be executed entirely on the user's computing device, partially on the user's device, partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0167] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0168] Those skilled in the art will understand that the features described in the various embodiments of the present invention can be combined and / or combined in various ways, even if such combinations or combinations are not explicitly described in the present invention. In particular, the features described in the various embodiments of the present invention can be combined and / or combined in various ways without departing from the spirit and teachings of the present invention. All such combinations and / or combinations fall within the scope of the present invention.

[0169] The embodiments of the present invention have been described above. However, these embodiments are merely illustrative and not intended to limit the scope of the invention. Although various embodiments have been described above, this does not mean that the measures in the various embodiments cannot be used advantageously in combination. Various substitutions and modifications can be made by those skilled in the art without departing from the scope of the invention, and all such substitutions and modifications should fall within the scope of the invention.

Claims

1. A traffic signal control method based on multi-source heterogeneous data, characterized in that, The method includes: Obtain historical traffic information for multiple intersections within a historical period, wherein the multiple intersections include the current intersection and the upstream intersection of the current intersection; The historical traffic information is input into the prediction model to obtain the predicted traffic parameters of each of the multiple traffic flow phases at the current intersection in the future cycle and the signal phase information of multiple traffic lights. The multiple traffic flow phases represent the directions in which vehicles can travel at the current intersection. The signal phase information includes the signal phase sequence, phase loss duration, first red light phase duration and second red light phase duration. The signal phase sequence is the first red light phase, green light phase, yellow light phase and second red light phase. The predicted traffic parameters and the signal phase information are input into the traffic control model. Based on the constraint that the congestion density in the traffic control model is less than or equal to the critical density, the duration of the green light signal in the traffic control model is updated until the cumulative queue length of the second red light phase in multiple consecutive preset cycles in the future cycle reaches the target value and the target green light is obtained. The target value is the minimum value among the cumulative queue lengths of the second red light phase in each of the multiple preset cycles obtained after updating the green light signal duration in the traffic control model multiple times. The target cycle duration is determined based on the target green light duration, the phase loss duration, the first red light phase duration, and the second red light phase duration. The various traffic lights include green lights, the cumulative queue length represents the number of vehicles waiting to pass through the current intersection, and the traffic control model is constructed based on the queue length and congestion density of the various traffic lights in the future cycle for each of the various traffic flow phases. The various traffic lights are controlled according to the target green light duration and the target cycle duration.

2. The method according to claim 1, characterized in that, The predicted traffic parameters include predicted arrival rate, predicted saturation release rate, and predicted congestion density. The predicted arrival rate represents the number of vehicles approaching the current intersection within a preset time.

3. The method according to claim 2, characterized in that, The historical period includes the current period, and the future period includes the period following the current period. The traffic control model is constructed based on the following: The shock wave velocity is determined based on the predicted arrival rate, the predicted saturation release rate, and the predicted congestion density for the next cycle. Based on the shock wave velocity, the queue length at the end of the current cycle, and the duration of the green light signal, a sub-model of the queue length of each of the various traffic lights is constructed to determine the queue length of each of the various traffic lights at the current intersection in the next cycle. Based on the signal phase sequence, the phase loss duration, and the duration of the green light signal, the first target duration and the second target duration in the next cycle are determined. Based on the first target duration, the second target duration, and the queue lengths of the various traffic lights in the next cycle, a sub-model of the congestion density of the current intersection is constructed to determine the congestion density of the current intersection in the next cycle. Based on the sub-models of the queue lengths of the various traffic lights and the sub-model of the congestion density, the traffic control model of the current intersection is constructed.

4. The method according to claim 3, characterized in that, The sub-models for constructing the queue lengths of the various traffic lights, based on the shock wave velocity, the queue length at the end of the current cycle, and the duration of the green light signal, include: Based on the shock wave velocity and the duration of the first red light phase, the queue length of the first red light phase is determined, and a sub-model of the cumulative queue length of the first red light phase is constructed according to the sum of the queue length of the first red light phase and the queue length at the end of the current cycle, so as to determine the cumulative queue length of the first red light phase of the current intersection in the next cycle. Based on the predicted saturation release rate, the predicted congestion density, and the duration of the green light signal, the passage length of the green light phase is determined, and a sub-model of the remaining queue length of the green light phase is constructed according to the cumulative queue length of the first red light phase and the passage length of the green light phase, so as to determine the remaining queue length of the green light phase of the current intersection in the next cycle. Based on the shock wave velocity and the duration of the second red light phase, the queue length of the second red light phase is determined, and a sub-model of the cumulative queue length of the second red light phase is constructed according to the sum of the queue length of the second red light phase and the remaining queue length of the green light phase, so as to determine the cumulative queue length of the second red light phase of the current intersection in the next cycle.

5. The method according to claim 4, characterized in that, The step of determining the first target duration and the second target duration in the next cycle based on the signal phase sequence, the phase loss duration, and the duration of the green light signal includes: Based on the signal phase sequence, the duration of the first red light phase, the duration of the green light signal, the duration of the yellow light phase, and the duration of the second red light phase, the first target duration before the start of the green light phase and the second target duration after the end of the green light phase are determined in the next cycle, and the duration of the yellow light phase is the phase loss duration.

6. The method according to claim 4, characterized in that, The sub-model for constructing the congestion density of the current intersection based on the first target duration, the second target duration, and the queue lengths of the various traffic lights in the next cycle includes: Based on the first target duration, the queue length at the end of the current cycle, and the queue length of the first red light phase, the first congestion parameter of the first red light phase is determined; Based on the second target duration, the remaining queue length of the green light phase, and the queue length of the second red light phase, the second congestion parameter of the second red light phase is determined. Based on the first congestion parameter, the second congestion parameter, and the predicted congestion density, a sub-model of the congestion density is constructed.

7. The method according to claim 1, characterized in that, The prediction model includes a forget gate, an input gate, an output gate, and a memory unit, and is trained in the following manner: Traffic training information from multiple training intersections during the training period is input into the initial prediction model to obtain the initial prediction value of the current training intersection. The initial prediction value includes initial traffic parameters and initial signal phase information. The multiple training intersections include the current training intersection and the upstream training intersection of the current training intersection. The model parameters of the initial prediction model are updated based on the error between the initial predicted value and the actual value until the error meets a preset condition, thereby obtaining the prediction model. The actual value includes the actual traffic parameters and actual signal phase information of the current training intersection.

8. The method according to claim 7, characterized in that, The model parameters include the weights and biases of the forget gate, the input gate, the output gate, and the memory unit, respectively. The step of updating the model parameters of the initial prediction model based on the error between the initial predicted value and the true value until the error meets a preset condition, thereby obtaining the prediction model, includes: The gradient of the t-th model parameter is determined based on the t-th updated model parameter, the step size coefficient in the gradient descent algorithm, and the t-th error, where t is an integer greater than or equal to 1. The t-th model parameters are updated based on the t-th model parameters, the gradient, and the step size coefficient to obtain the (t+1)-th updated model parameters.

9. The method according to claim 1, characterized in that, The historical traffic information includes: the historical driving speed of vehicles passing through the multiple intersections collected by inductive loop detectors, the historical traffic flow of the multiple intersections collected by vehicle locators, and the historical queue length of the multiple intersections collected by roadside cameras.

Images