Road traffic signal light control method and system based on vehicle-road-cloud collaboration

By using the vehicle-road-cloud collaborative control system, the LSTM algorithm is used to analyze historical traffic data and adjust the traffic light control strategy in real time. This solves the problem that traditional traffic light systems cannot dynamically respond to changes in traffic flow, and realizes intelligent and efficient traffic management.

CN118692249BActive Publication Date: 2024-11-05SICHUAN HUATI LIGHTING TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411180148.8
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-27
Publication Date
2024-11-05
Estimated Expiration
2044-08-27

AI Technical Summary

Technical Problem

Traditional traffic signal control systems struggle to perceive and respond to dynamic changes in traffic flow in real time, leading to traffic congestion and reduced traffic efficiency.

Method used

By using the vehicle-road-cloud collaborative control system, historical traffic data is analyzed using the Long Short-Term Memory (LSTM) algorithm to construct a spatiotemporal traffic light control algorithm and adjust the traffic light control strategy in real time.

Benefits of technology

It has improved the accuracy and intelligence of traffic management, effectively alleviated urban traffic congestion, reduced the cost of manual intervention, and optimized traffic flow.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118692249B_ABST
    Figure CN118692249B_ABST
Patent Text Reader

Abstract

The embodiment of the present invention discloses a road traffic light control method and system based on vehicle-road-cloud collaboration, which belongs to the field of intelligent transportation technology. The embodiment of the present invention deeply mines the spatiotemporal dynamic information in historical traffic data to construct a spatiotemporal signal light control algorithm that can accurately perceive and predict traffic flow changes. Such a design not only improves the accuracy and intelligence level of traffic management, but also provides a new solution to alleviate urban traffic congestion. At the same time, the embodiment of the present invention also realizes the automation and intelligence of traffic light control, reduces the cost and difficulty of manual intervention, and provides a strong guarantee for the smooth operation of urban traffic.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the field of intelligent transportation technology, and specifically relates to a road traffic signal light control method and system based on vehicle-road-cloud collaboration. Background Art

[0002] In modern urban traffic management, the control strategy of traffic lights plays a vital role in alleviating traffic congestion and improving road capacity. However, traditional traffic light control systems often operate based on fixed time intervals and preset control modes, making it difficult to perceive and respond to dynamic changes in traffic flow in real time. This results in the inability to adjust the control strategy of traffic lights in a timely manner when actual traffic conditions change, which can easily cause traffic congestion and a decrease in traffic efficiency. Summary of the invention

[0003] The present invention provides a road traffic signal light control method and system based on vehicle-road-cloud collaboration, which can solve or partially solve the technical problems involved in the above-mentioned background technology.

[0004] An embodiment of the present invention provides a road traffic signal light control method based on vehicle-road-cloud collaboration, which is applied to a vehicle-road-cloud collaborative control system. The method comprises: obtaining a historical vehicle-road-cloud traffic spatiotemporal data set, wherein the historical vehicle-road-cloud traffic spatiotemporal data set includes a plurality of historical traffic participating units and a plurality of historical road network transmission characteristics, wherein any historical traffic participating unit represents a historical road network flow state data, and any historical road network transmission characteristic represents the road network correlation between two corresponding historical traffic participating units; determining a plurality of historical traffic flow monitoring data subsets based on the historical vehicle-road-cloud traffic spatiotemporal data set, wherein any historical traffic flow monitoring data subset includes a key traffic participating unit and an involved traffic participating unit, and the involved traffic participating unit is associated with the key traffic participating unit through the historical road network transmission characteristics; for any historical traffic flow monitoring data subset, using a long short-term memory algorithm based on the key traffic participating units and the involved traffic participating units included in any historical traffic flow monitoring data subset Unit, determine the spatiotemporal dynamic quantization vector of the key traffic participating unit included in any one of the historical traffic flow monitoring data subsets, the spatiotemporal dynamic quantization vector of the key traffic participating unit included in any one of the historical traffic flow monitoring data subsets includes the spatiotemporal dynamic information of the historical road network flow status data represented by the key traffic participating unit included in any one of the historical traffic flow monitoring data subsets and the spatiotemporal dynamic information of the historical road network flow status data involving the traffic participating unit represented in any one of the historical traffic flow monitoring data subsets; according to the spatiotemporal dynamic quantization vector of the key traffic participating unit included in each historical traffic flow monitoring data subset, debug the long short-term memory algorithm to obtain the spatiotemporal signal light control algorithm, the spatiotemporal signal light control algorithm is used to determine the spatiotemporal dynamic quantization vector of the key traffic participating unit included in the target traffic flow monitoring data subset; and the spatiotemporal dynamic quantization vector of the key traffic participating unit is used to make control decisions on the road traffic signal light relationship network.

[0005] An embodiment of the present invention provides a vehicle-road-cloud collaborative control system, comprising at least one processor and a memory; the memory stores computer-executable instructions; the at least one processor executes the computer-executable instructions stored in the memory, so that the at least one processor executes the above method.

[0006] An embodiment of the present invention provides a readable storage medium, on which a program or instruction is stored. When the program or instruction is executed by a processor, the steps of the above method are implemented.

[0007] In the embodiment of the present invention, by deeply mining the spatiotemporal dynamic information in historical traffic data, a spatiotemporal signal light control algorithm that can accurately perceive and predict traffic flow changes is constructed. Such a design not only improves the accuracy and intelligence level of traffic management, but also provides a new solution for alleviating urban traffic congestion. At the same time, the embodiment of the present invention also realizes the automation and intelligence of traffic light control, reduces the cost and difficulty of manual intervention, and provides a strong guarantee for the smooth operation of urban traffic. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] Figure 1 A flow chart of a road traffic light control method based on vehicle-road-cloud collaboration provided in an embodiment of the present invention.

[0009] Figure 2 A schematic diagram of the structure of a vehicle-road-cloud collaborative control system provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0010] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0011] The terms "first", "second", etc. in the present invention are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments of the present invention can be implemented in an order other than those illustrated or described herein, and the objects distinguished by "first", "second", etc. are generally of the same type, and the number of objects is not limited. For example, the first object can be one or more. In addition, in the present invention, "and / or" means at least one of the connected objects, and the character " / " generally indicates that the objects associated with each other are in an "or" relationship.

[0012] Figure 1 A road traffic light control method based on vehicle-road-cloud collaboration is shown, which is applied to a vehicle-road-cloud collaborative control system. The method includes the following steps 110 to 140.

[0013] The technical solution provided by the embodiment of the present invention realizes comprehensive perception, intelligent decision-making and efficient execution of traffic information by integrating vehicles, road infrastructure and cloud platforms, greatly improving the safety, efficiency and sustainability of the traffic system. The following will introduce in detail how the vehicle-road-cloud collaborative control system implements its technical solution through a complete and specific application scenario to ensure the clarity and coherence of each step.

[0014] Take a typical urban traffic area as an example. This urban traffic area consists of multiple main roads, secondary roads and intersections. The daily traffic volume is large and congestion is prone to occur during peak hours. In order to improve the traffic conditions in this urban traffic area, a vehicle-road-cloud collaborative control system is introduced to collect and analyze traffic data in real time, optimize signal light control strategies, and achieve effective management of traffic flow.

[0015] Step 110: Obtain a historical vehicle-road-cloud traffic spatiotemporal dataset, wherein the historical vehicle-road-cloud traffic spatiotemporal dataset includes multiple historical traffic participating units and multiple historical road network transmission characteristics, wherein any historical traffic participating unit represents a historical road network flow state data, and any historical road network transmission characteristic represents the road network correlation between the corresponding two historical traffic participating units.

[0016] The vehicle-road-cloud collaborative control system first obtains the historical vehicle-road-cloud traffic spatiotemporal dataset containing rich traffic information from the historical database. The historical vehicle-road-cloud traffic spatiotemporal dataset covers multiple historical traffic participants (such as vehicles, pedestrians, and traffic lights) and road network transmission characteristics. Each historical traffic participant represents a specific road network flow state data through its position, speed, acceleration and other parameters. The road network transmission characteristics describe the correlation between different traffic participants, such as the process of a vehicle moving from one intersection to another. This correlation is crucial to understanding the dynamic changes of traffic flow.

[0017] For example, historical traffic participant unit data includes vehicle ID, timestamp, location coordinates (latitude and longitude), speed, acceleration, etc. Historical road network transmission feature data includes intersection number, connecting road number, direction indication, historical average travel time, etc., which are used to describe the traffic flow transmission relationship between different road sections and intersections.

[0018] Step 120: Determine multiple historical traffic flow monitoring data subsets based on the historical vehicle-road-cloud traffic spatiotemporal data set, any historical traffic flow monitoring data subset includes key traffic participating units and involved traffic participating units, and the involved traffic participating units are associated with the key traffic participating units through historical road network transmission characteristics.

[0019] The vehicle-road-cloud collaborative control system further identifies key traffic participants and involved traffic participants based on the historical vehicle-road-cloud traffic spatiotemporal data set, forming multiple historical traffic flow monitoring data subsets. Key traffic participants usually refer to units that have a significant impact on the overall traffic flow, such as vehicles in the main direction of traffic at a large intersection. Involved traffic participants are units that are directly or indirectly associated with key traffic participants through road network transmission characteristics.

[0020] Exemplarily, the data subset construction process is as follows:

[0021] (1) Identify key traffic participants: Based on historical data, the system analyzes which traffic participants have the greatest impact on traffic flow during specific periods (such as morning and evening rush hours), such as vehicles at intersections with large and fast-changing traffic volumes;

[0022] (2) Determine the involved traffic participants: Through the road network transmission characteristics, find the units that have direct or indirect traffic flow transmission relationships with key traffic participants, such as upstream vehicles and downstream vehicles.

[0023] Step 130: For any subset of historical traffic flow monitoring data, determine the spatiotemporal dynamic quantization vector of the key traffic participating units included in any subset of historical traffic flow monitoring data through a long short-term memory algorithm based on the key traffic participating units and involved traffic participating units included in any subset of historical traffic flow monitoring data.

[0024] Among them, the spatiotemporal dynamic quantization vectors of the key traffic participating units included in any one of the historical traffic flow monitoring data subsets include the spatiotemporal dynamic information of the historical road network flow status data represented by the key traffic participating units included in any one of the historical traffic flow monitoring data subsets and the spatiotemporal dynamic information of the historical road network flow status data represented by the involved traffic participating units included in any one of the historical traffic flow monitoring data subsets.

[0025] For each subset of historical traffic flow monitoring data, the vehicle-road-cloud collaborative control system uses the long short-term memory (LSTM) algorithm to combine the historical data of key traffic participants and involved traffic participants to calculate the spatiotemporal dynamic quantization vector of the key traffic participants. The spatiotemporal dynamic quantization vector not only contains the spatiotemporal dynamic information of the historical road network flow status data of the key traffic participants themselves, but also incorporates the relevant information of the involved traffic participants, thereby comprehensively reflecting the dynamic changes of traffic flow.

[0026] The exemplary applications of the LSTM algorithm involved in the embodiments of the present invention are as follows:

[0027] (1) Data preprocessing: sort the historical data of key and involved traffic participating units into time series and perform preprocessing operations such as normalization;

[0028] (2) Model training: Use LSTM network to train historical data. The network structure includes input layer, hidden layer (multi-layer LSTM unit) and output layer. The input layer receives time series data, the hidden layer captures the temporal dependencies in the data, and the output layer outputs the spatiotemporal dynamic quantized vectors of key traffic participating units;

[0029] (3) Vector construction: Each element in the vector represents spatiotemporal dynamic information of different dimensions, such as traffic flow change trends, speed distribution, etc. This information provides a basis for subsequent traffic light control decisions.

[0030] Step 140, based on the spatiotemporal dynamic quantization vectors of the key traffic participants included in each historical traffic flow monitoring data subset, debug the long short-term memory algorithm to obtain a spatiotemporal signal light control algorithm, the spatiotemporal signal light control algorithm is used to determine the spatiotemporal dynamic quantization vectors of the key traffic participants included in the target traffic flow monitoring data subset; and the spatiotemporal dynamic quantization vectors of the key traffic participants are used to make control decisions on the road traffic signal light relationship network.

[0031] Based on the spatiotemporal dynamic quantization vectors calculated from each subset of historical traffic flow monitoring data, the vehicle-road-cloud collaborative control system debugs and optimizes the long-term and short-term memory algorithm, and finally generates a spatiotemporal signal light control algorithm. The spatiotemporal signal light control algorithm can receive the current traffic flow monitoring data in real time, calculate the current spatiotemporal dynamic quantization vectors of key traffic participating units, and adjust the control strategy of road traffic lights accordingly.

[0032] The debugging and optimization process involved in the embodiment of the present invention is exemplified as follows:

[0033] (1) Algorithm evaluation: Use historical data to evaluate the initially generated spatiotemporal traffic light control algorithm, compare the traffic flow status under algorithm control with the actual situation, and identify deficiencies in the algorithm;

[0034] (2) Parameter adjustment: Based on the evaluation results, adjust the parameters of the LSTM model (such as learning rate, number of hidden layer units, number of iterations, etc.) to optimize the algorithm performance;

[0035] (3) Simulation testing: Testing the optimized algorithm in a traffic simulation environment to simulate the signal light control effects in different traffic scenarios to ensure the stability and effectiveness of the algorithm;

[0036] (4) On-site deployment: The debugged spatiotemporal traffic light control algorithm is deployed to the actual traffic area. Traffic data is received in real time through the cloud platform, the traffic light control strategy is calculated, and the control instructions are sent to each traffic light controller for execution through the vehicle-road cooperative system.

[0037] Through steps 110 to 140, the vehicle-road-cloud collaborative control system realizes accurate perception and intelligent control of traffic flow, which can not only reflect the dynamic changes of traffic flow in real time, but also predict future trends based on historical data and current conditions, providing a scientific basis for traffic light control. With the continuous advancement of technology and the in-depth promotion of applications, the vehicle-road-cloud collaborative control system will play an increasingly important role in alleviating traffic congestion, improving road safety, and promoting the development of intelligent transportation systems.

[0038] To improve the readability of the embodiments of the present invention, the above technical solution is introduced below through a complete and smooth application scenario example.

[0039] In the busy urban traffic network, the vehicle-road-cloud collaborative control system is playing an increasingly important role. Take a typical urban traffic area as an example. This urban traffic area consists of multiple main roads, secondary roads and intersections. The daily traffic flow is large and congestion is prone to occur during peak hours. In order to improve this situation, the vehicle-road-cloud collaborative control system is introduced to collect and analyze traffic data in real time, optimize the signal light control strategy, and realize the effective management of traffic flow.

[0040] First, the vehicle-road-cloud collaborative control system obtains spatiotemporal data sets containing rich traffic information from historical databases. These data sets cover multiple historical traffic participants (such as vehicles, pedestrians) and road network transmission characteristics. Each historical traffic participant represents a specific road network flow state data through its position, speed, acceleration and other parameters. The road network transmission characteristics describe the correlation between different traffic participants, such as the process of a vehicle moving from one intersection to another. These data provide a comprehensive historical background of traffic flow for the vehicle-road-cloud collaborative control system.

[0041] Secondly, based on these historical vehicle-road-cloud traffic spatiotemporal data sets, the vehicle-road-cloud collaborative control system further identified key traffic participants and involved traffic participants, forming multiple historical traffic flow monitoring data subsets. Key traffic participants usually refer to units that have a significant impact on the overall traffic flow, such as vehicles in the main direction of traffic at a large intersection. Involved traffic participants are units that are directly or indirectly associated with key traffic participants through road network transmission characteristics. Such data subset construction helps the vehicle-road-cloud collaborative control system focus more on factors that have a significant impact on traffic flow.

[0042] Then, for each subset of historical traffic flow monitoring data, the vehicle-road-cloud collaborative control system uses the long short-term memory (LSTM) algorithm. The algorithm combines the historical data of key traffic participants and involved traffic participants to calculate the spatiotemporal dynamic quantization vector of the key traffic participants. This vector not only contains the spatiotemporal dynamic information of the historical road network flow state data of the key traffic participant itself, but also incorporates the relevant information of the involved traffic participants, thereby fully reflecting the dynamic changes of traffic flow. This provides strong data support for subsequent traffic signal control decisions.

[0043] Finally, based on the spatiotemporal dynamic quantization vectors calculated from each subset of historical traffic flow monitoring data, the vehicle-road-cloud collaborative control system debugged and optimized the long-short-term memory algorithm, and finally generated a spatiotemporal signal light control algorithm. The algorithm can receive the current traffic flow monitoring data in real time, calculate the current spatiotemporal dynamic quantization vectors of key traffic participating units, and adjust the control strategy of road traffic lights accordingly. In this way, the vehicle-road-cloud collaborative control system realizes accurate perception and intelligent control of traffic flow, effectively alleviating traffic congestion problems.

[0044] In summary, the vehicle-road-cloud collaborative control system has successfully achieved effective management of urban traffic flow through a series of steps such as obtaining historical traffic data, identifying key traffic participants, calculating spatiotemporal dynamic quantization vectors, and debugging and optimizing signal light control algorithms. The application of this technology not only improves traffic efficiency, but also lays a solid foundation for the sustainable development of urban transportation.

[0045] It is worth mentioning that step 130 is the core invention of the embodiment of the present invention, which involves a complex and innovative technical solution, which aims to deeply mine the spatiotemporal dynamic information in the historical traffic flow data and provide a scientific basis for the subsequent traffic signal control. Specifically, for any subset of historical traffic flow monitoring data, the system will use the long short-term memory (LSTM) algorithm to determine the spatiotemporal dynamic quantization vector of the key traffic participating units and the information of the involved traffic participating units contained in the data subset.

[0046] First, the system will preprocess the selected subset of historical traffic flow monitoring data, including sorting the historical data of key traffic participants and involved traffic participants in time series, and performing necessary preprocessing operations such as normalization to ensure the accuracy and consistency of the data.

[0047] Secondly, the system will use the powerful time series modeling capabilities of the LSTM algorithm to conduct in-depth analysis and learning of these pre-processed historical data. The structure of the LSTM network includes an input layer, a hidden layer (composed of multiple layers of LSTM units), and an output layer. During the training process, the input layer receives time series data, the hidden layer is responsible for capturing the temporal dependencies in the data, and the output layer ultimately outputs the spatiotemporal dynamic quantized vectors of key traffic participating units.

[0048] The spatiotemporal dynamic quantization vector is a multi-dimensional vector that contains rich spatiotemporal dynamic information. On the one hand, it contains the spatiotemporal dynamic information of the historical road network flow state data represented by the key traffic participants themselves, such as the change trend of flow, speed distribution, etc. On the other hand, it also incorporates the relevant information of the traffic participants involved, thus fully reflecting the dynamic changes of traffic flow. Such a design enables the system to more accurately grasp the complexity and dynamics of traffic flow.

[0049] It can be seen that step 130 achieves deep mining and effective use of historical traffic flow data through the introduction of LSTM algorithm and spatiotemporal dynamic quantization vector. This provides strong data support for subsequent traffic light control, enabling the system to more accurately perceive the dynamic changes of traffic flow and make scientific control decisions accordingly.

[0050] In addition, step 140 is also the core invention of the present invention, which focuses on using the spatiotemporal dynamic quantization vectors of key traffic participating units in the historical traffic flow monitoring data subset to fine-tune the long short-term memory algorithm, and then obtain the spatiotemporal signal light control algorithm. The core value of this algorithm is that it can accurately determine the spatiotemporal dynamic quantization vectors of key traffic participating units in the target traffic flow monitoring data subset, and this vector is the key basis for making road traffic signal light relationship network control decisions.

[0051] First, the system collects the spatiotemporal dynamic quantitative vectors of key traffic participating units that have been calculated in each subset of historical traffic flow monitoring data. These vectors contain rich spatiotemporal dynamic information of traffic flow and are valuable resources for debugging long-term and short-term memory algorithms and optimizing traffic light control strategies.

[0052] Secondly, the system will use these spatiotemporal dynamic quantization vectors to deeply debug and optimize the long short-term memory algorithm. This process involves multiple aspects such as the adjustment of algorithm parameters and the optimization of model structure, aiming to make the algorithm better adapt to the changes in actual traffic flow and more accurately predict and respond to the dynamic situation of traffic flow.

[0053] Then, the debugged and optimized LSTM algorithm is further transformed into a spatiotemporal signal light control algorithm, which can receive the input of the target traffic flow monitoring data subset, calculate the spatiotemporal dynamic quantization vector of key traffic participating units in real time, and make scientific signal light control decisions based on the vector.

[0054] Ultimately, the spatiotemporal signal light control algorithm will be applied to the actual road traffic signal light relationship network. By sensing the dynamic changes of traffic flow in real time and adjusting the signal light control strategy accordingly, the system can effectively alleviate traffic congestion, improve road capacity, and provide strong guarantees for the smooth operation of urban traffic.

[0055] With such a design, step 140 obtains a spatiotemporal traffic light control algorithm by fine-tuning the long short-term memory algorithm. The spatiotemporal traffic light control algorithm can accurately calculate the spatiotemporal dynamic quantization vectors of key traffic participants and provide a scientific basis for the control decision of the road traffic signal network, thereby realizing the intelligent management and optimization of urban traffic.

[0056] In summary, by deeply mining the historical vehicle-road cloud traffic spatiotemporal data set, an innovative traffic light control technology based on spatiotemporal dynamic quantization vector is constructed, which can not only fully perceive the spatiotemporal dynamic changes of traffic flow, but also accurately predict and respond to the real-time status of traffic flow through the long short-term memory algorithm, thereby realizing the intelligent control of the road traffic light relationship network. In detail, by introducing the spatiotemporal dynamic quantization vector, the embodiment of the present invention can more accurately grasp the complexity and dynamics of traffic flow and provide a scientific basis for the control of traffic lights. By using the long short-term memory algorithm to conduct deep learning and prediction of traffic flow data, the embodiment of the present invention can automatically adjust the signal light control strategy according to the real-time traffic conditions and improve the road traffic capacity. By perceiving and responding to the dynamic changes of traffic flow in real time, the embodiment of the present invention can more effectively optimize the signal light control, reduce traffic congestion, and improve the overall operation efficiency of urban traffic, thereby realizing the automation and intelligence of traffic light control, reducing the cost and difficulty of manual intervention, and improving the overall efficiency of traffic management.

[0057] In an optional embodiment, the acquisition of a historical vehicle-road-cloud traffic spatiotemporal dataset includes: acquiring multiple traffic participation events, any one of which is used to represent an event in which a historical vehicle activates a historical navigation route; generating an original vehicle-road-cloud traffic spatiotemporal dataset based on the multiple traffic participation events, the original vehicle-road-cloud traffic spatiotemporal dataset including multiple original traffic participation units and multiple original road network transfer features, any one of which represents a historical road network flow state data, and any one of which represents the road network correlation between two corresponding original traffic participation units; removing the original road network transfer features whose road network correlation meets the first correlation requirement from the original vehicle-road-cloud traffic spatiotemporal dataset to obtain the historical vehicle-road-cloud traffic spatiotemporal dataset.

[0058] It can be understood that this embodiment goes deep into the acquisition and processing of historical vehicle-road-cloud traffic spatiotemporal data sets, which is the cornerstone of building an efficient and accurate traffic light control system.

[0059] First, multiple traffic participation events need to be obtained. These events are microscopic records of traffic activities, and each event represents a historical vehicle activating a historical navigation route. In other words, every time a vehicle selects and drives a specific navigation route at a certain point in history, this behavior is recorded as a traffic participation event. These events not only contain vehicle information, such as vehicle model, license plate number, etc., but also detailed information on the navigation route, such as the starting point, end point, passing sections, estimated travel time, etc. Through these events, the driving trajectory of vehicles in historical time and space can be traced, and the distribution and changes of traffic flow can be analyzed.

[0060] Secondly, based on these rich traffic participation events, the system will generate an original vehicle-road cloud traffic spatiotemporal data set. This data set is a multi-dimensional information set, which contains multiple original traffic participation units and multiple original road network transmission characteristics. In an embodiment of the present invention, each original traffic participation unit represents a historical road network traffic status data. It is not just a simple number or symbol, but an entity that contains rich information. For example, it can represent the traffic flow, average speed, vehicle type distribution, etc. of a specific road section at a certain historical time point. This information can fully reflect the traffic status of the road section at that moment.

[0061] At the same time, each original road network transmission feature represents the road network correlation between the corresponding two original traffic participating units. This correlation is formed based on the propagation and mutual influence of traffic flow in the road network. For example, if two road sections are geographically adjacent and vehicles often travel between the two sections, then the road network correlation between them will be strong. This correlation is of great significance for understanding the dynamic changes of traffic flow and predicting the propagation of traffic congestion.

[0062] However, although the original vehicle-road-cloud traffic spatiotemporal dataset contains rich information, it may also contain some redundant data or data that does not meet specific analysis requirements. Therefore, the dataset needs to be further optimized. Specifically, the original road network transfer features whose road network relevance meets the first relevance requirement are removed. The "first relevance requirement" here is a preset threshold or condition used to judge the strength of the road network relevance or whether it meets specific analysis requirements. For example, a relevance threshold can be set, and only when the relevance between two road sections exceeds the threshold, it is considered that there is a significant traffic flow impact between them. Therefore, during the removal process, the road network relevance represented by each original road network transfer feature will be checked. If it meets the first relevance requirement (that is, the relevance is weak or does not meet specific analysis requirements), it will be removed from the dataset.

[0063] After such processing, the resulting historical vehicle-road-cloud traffic spatiotemporal dataset will be more refined and effective. It not only retains the correlation between road sections that have a significant impact on traffic flow, but also removes redundant or non-compliant data. Such a dataset can provide a solid data foundation for subsequent traffic light control algorithms, helping to more accurately perceive and predict the dynamic changes of traffic flow, thereby formulating a more scientific and reasonable traffic light control strategy.

[0064] Next, multiple historical traffic flow monitoring data subsets will be determined based on the carefully processed historical vehicle-road-cloud traffic spatiotemporal data set. Each historical traffic flow monitoring data subset contains key traffic participating units and involved traffic participating units. In an embodiment of the present invention, key traffic participating units are those road sections or intersections that have an important impact on traffic flow, and involved traffic participating units are other road sections or intersections that have significant traffic flow associations with these key road sections or intersections. By incorporating these road sections or intersections into the same historical traffic flow monitoring data subset, the mutual influence and dynamic changes of traffic flows between them can be analyzed more comprehensively.

[0065] Then, for each subset of historical traffic flow monitoring data, the long short-term memory algorithm will be used to deeply mine the spatiotemporal dynamic information therein. The long short-term memory algorithm is a powerful time series modeling tool that can capture long-term dependencies in the data and predict future trends. In an embodiment of the present invention, the historical data of key traffic participating units and involved traffic participating units will be used as input, and through the learning and training of the long short-term memory algorithm, a quantitative vector that can accurately describe the spatiotemporal dynamic changes of these road sections or intersections is obtained. The quantitative vector not only contains the spatiotemporal dynamic information of the historical road network flow status data of the key traffic participating units themselves, but also incorporates the relevant information of the involved traffic participating units. Such a design makes it possible to grasp the complexity and dynamics of traffic flow more comprehensively.

[0066] Finally, the long short-term memory algorithm will be further debugged and optimized based on the spatiotemporal dynamic quantization vectors of key traffic participants obtained from each subset of historical traffic flow monitoring data. The goal of this process is to obtain a spatiotemporal signal light control algorithm that can automatically adjust the signal light control strategy according to real-time traffic conditions to improve road capacity and alleviate urban traffic congestion. Specifically, the spatiotemporal dynamic quantization vector will be used as input, and the optimal signal light control strategy for the target traffic flow monitoring data subset will be obtained through calculation and reasoning of the spatiotemporal signal light control algorithm. This strategy will take into account multiple factors such as the spatiotemporal dynamic changes of traffic flow, the correlation between road sections, and the physical constraints of traffic lights to ensure that the developed control strategy is both scientific and feasible.

[0067] In this way, through multiple steps such as in-depth mining of historical vehicle-road-cloud traffic spatiotemporal data sets, constructing refined and effective data subsets, applying long short-term memory algorithms for spatiotemporal dynamic modeling, and debugging and optimizing spatiotemporal signal light control algorithms, an intelligent traffic light control system that can accurately perceive and predict dynamic changes in traffic flow is realized. This can not only improve the accuracy and intelligence level of traffic management, but also effectively alleviate urban traffic congestion and improve the overall operating efficiency of urban traffic.

[0068] In the following embodiment, the generation of the original vehicle-road-cloud traffic spatiotemporal dataset based on the multiple traffic participation events includes: determining the signal light link data of multiple historical navigation routes and the intersection statistics of every two historical navigation routes based on the multiple traffic participation events, the intersection statistics of any two historical navigation routes being the number of times the any two historical navigation routes are activated by the same historical vehicle; determining the multiple original traffic participation units based on the signal light link data of the multiple historical navigation routes, the historical road network flow state data represented by any one original traffic participation unit being the signal light link data of a historical navigation route; determining the original road network transfer characteristics between the original traffic participation units corresponding to the any two historical navigation routes based on the intersection statistics of the any two historical navigation routes, the intersection statistics of the any two historical navigation routes being the road network correlation represented by the original road network transfer characteristics between the original traffic participation units corresponding to the any two historical navigation routes; determining the original vehicle-road-cloud traffic spatiotemporal dataset based on the multiple original traffic participation units and the original road network transfer characteristics between every two original traffic participation units.

[0069] How does this embodiment generate an original vehicle-road-cloud traffic spatiotemporal dataset based on multiple traffic participation events? This process involves multiple key steps, including determining the signal light link data of the historical navigation route, calculating the intersection statistics, determining the original traffic participation unit, and extracting the original road network transmission characteristics.

[0070] First, based on multiple traffic participation events, it is necessary to determine the signal light link data of multiple historical navigation routes. This process involves in-depth analysis of traffic participation events to extract each historical vehicle-activated navigation route and related signal light status information. Signal light link data refers to the information about the sequence of signal lights and their status (such as red light, green light, and yellow light) that a vehicle passes through in a specific historical navigation route. By collecting this data, a detailed signal light link for each historical navigation route can be constructed, providing a basis for subsequent analysis.

[0071] At the same time, it is also necessary to calculate the cross statistics of every two historical navigation routes. This statistic refers to the number of times any two historical navigation routes are activated by the same historical vehicle. In other words, it reflects the frequency of different historical navigation routes being driven by the same vehicle. This indicator is important for understanding the correlation and similarity of traffic flows between different navigation routes. By calculating the cross statistics, it is possible to find out which historical navigation routes are often driven by the same vehicle, thereby revealing the potential correlation between them.

[0072] Next, based on the signal light link data of multiple historical navigation routes, multiple original traffic participating units can be determined. In the embodiment of the present invention, each original traffic participating unit represents the signal light link data of a historical navigation route. In other words, it encapsulates a specific historical navigation route and its related signal light status information into an independent entity for subsequent analysis and processing. Such a design enables more flexible processing and analysis of traffic flow data of different historical navigation routes.

[0073] Then, it is necessary to determine the original road network transfer characteristics between the original traffic participating units corresponding to any two historical navigation routes based on the cross-statistics of the two historical navigation routes. This transfer characteristic characterizes the road network correlation, that is, the degree of mutual influence and transmission of traffic flows between two original traffic participating units (or two historical navigation routes). Specifically, if two historical navigation routes are often traveled by the same vehicle, the road network correlation between them will be strong. This correlation is of great significance for understanding the dynamic propagation of traffic flow and the diffusion of congestion. By extracting the original road network transfer characteristics, the correlation and dynamic changes of traffic flows between different historical navigation routes can be revealed more deeply.

[0074] Finally, based on multiple original traffic participating units and the original road network transmission characteristics between each two original traffic participating units, the original vehicle-road-cloud traffic spatiotemporal dataset can be determined. This dataset is a multi-dimensional information collection, which contains rich traffic flow data, signal light status information, and road network correlation information. By constructing such a dataset, a solid data foundation can be provided for the subsequent traffic light control algorithm. In this dataset, each original traffic participating unit represents a specific historical navigation route and its related traffic flow data, while the original road network transmission characteristics reveal the traffic flow correlation and mutual influence between different historical navigation routes. Such a design enables a more comprehensive understanding and analysis of the dynamic changes of urban traffic flow, providing strong support for the subsequent formulation of signal light control strategies.

[0075] In this way, by deeply mining the information of multiple traffic events, an original vehicle-road-cloud traffic spatiotemporal dataset containing rich traffic flow data, signal light status information, and road network correlation information was constructed. This dataset provides a solid data foundation for the subsequent traffic light control algorithm, enabling more accurate perception and prediction of the dynamic changes of traffic flow, thereby formulating a more scientific and reasonable signal light control strategy.

[0076] In some other possible embodiments, the method of determining the spatiotemporal dynamic quantization vector of the key traffic participating units included in any one of the historical traffic flow monitoring data subsets based on the key traffic participating units and the involved traffic participating units included in any one of the historical traffic flow monitoring data subsets by using the long short-term memory algorithm comprises: determining the original spatiotemporal dynamic vector of the key traffic participating units included in any one of the historical traffic flow monitoring data subsets and the original spatiotemporal dynamic vector of the involved traffic participating units included in any one of the historical traffic flow monitoring data subsets by using the long short-term memory algorithm, the original spatiotemporal dynamic vector of the key traffic participating units included in any one of the historical traffic flow monitoring data subsets being used to record the original spatiotemporal dynamic vector of the key traffic participating units included in any one of the historical traffic flow monitoring data subsets. The spatiotemporal dynamic information of the historical road network flow status data represented by the key traffic participating units included in the historical traffic flow monitoring data subset, the original spatiotemporal dynamic vectors of the traffic participating units included in any one of the historical traffic flow monitoring data subsets are used to record the spatiotemporal dynamic information of the historical road network flow status data represented by the traffic participating units included in any one of the historical traffic flow monitoring data subsets; the original spatiotemporal dynamic vectors of the key traffic participating units included in any one of the historical traffic flow monitoring data subsets and the original spatiotemporal dynamic vectors of the traffic participating units included in any one of the historical traffic flow monitoring data subsets are combined to obtain the spatiotemporal dynamic quantization vectors of the key traffic participating units included in any one of the historical traffic flow monitoring data subsets.

[0077] In this embodiment, the focus of the technical solution turns to how to use the long short-term memory algorithm (LSTM) to deeply mine the spatiotemporal dynamic information in the historical traffic flow monitoring data subset, and determine the spatiotemporal dynamic quantization vector of the key traffic participant based on this. This process not only involves the analysis of the historical road network flow status data of the key traffic participant itself, but also covers the relevant data of the involved traffic participants that have significant traffic flow associations with it.

[0078] First, the original spatiotemporal dynamic vector of the key traffic participant in any subset of historical traffic flow monitoring data needs to be determined through the long short-term memory algorithm. This vector is a multi-dimensional data representation that records the spatiotemporal dynamic information of the historical road network flow state data represented by the key traffic participant. In other words, it contains the traffic flow state data of the key traffic participant at different time points, as well as the trends and patterns of how these data change over time. Through such a vector, we can have a more comprehensive understanding of the traffic flow dynamics of the key traffic participant in historical time and space.

[0079] At the same time, it is also necessary to determine the original spatiotemporal dynamic vector of the traffic participating units involved in the same historical traffic flow monitoring data subset. Similar to the original spatiotemporal dynamic vector of the key traffic participating units, this vector also records the spatiotemporal dynamic information of the historical road network flow state data represented by the traffic participating units. It reflects the traffic flow state data of the traffic participating units at different time points and their changing trends and patterns. By extracting this vector, the traffic flow dynamic information of the traffic participating units involved can be taken into consideration, so as to more comprehensively analyze the spatiotemporal dynamics of the key traffic participating units.

[0080] Next, it is necessary to combine the original spatiotemporal dynamic vector of the key traffic participant and the original spatiotemporal dynamic vector of the involved traffic participant. This combination process is achieved through certain mathematical operations or data fusion technology, which aims to integrate the information in the two vectors to form a more comprehensive and rich spatiotemporal dynamic quantitative vector. This new vector not only contains the spatiotemporal dynamic information of the key traffic participant itself, but also incorporates the relevant information of the involved traffic participant. Through such a design, the traffic flow dynamics of the key traffic participant in historical time and space can be described more accurately, while taking into account the influence of the involved traffic participant with significant traffic flow correlation.

[0081] Finally, through the above combination process, the spatiotemporal dynamic quantitative vector of the key traffic participants in any subset of historical traffic flow monitoring data is obtained. This vector is a multi-dimensional data representation, which comprehensively records the traffic flow dynamic information of the key traffic participants in historical time and space, as well as the relevant information of the involved traffic participants with significant traffic flow correlation. Such a design enables a deeper understanding and analysis of the spatiotemporal dynamic characteristics of the key traffic participants, providing more accurate and comprehensive data support for the subsequent formulation of signal light control strategies.

[0082] With this design, the spatiotemporal dynamic information in the subset of historical traffic flow monitoring data is deeply mined through the long short-term memory algorithm, and the spatiotemporal dynamic quantitative vector of the key traffic participants is determined accordingly. This process not only takes into account the historical road network flow status data of the key traffic participants themselves, but also incorporates the relevant information of the involved traffic participants that have significant traffic flow associations with them. Through this design, we can have a more comprehensive understanding of the traffic flow dynamics of the key traffic participants in historical time and space, and provide more accurate and comprehensive data support for the subsequent formulation of signal light control strategies.

[0083] In other preferred embodiments, the determining of the original spatiotemporal dynamic vectors of the key traffic participating units included in any one of the historical traffic flow monitoring data subsets by the long short-term memory algorithm comprises: determining the dense linear features of the key traffic participating units included in any one of the historical traffic flow monitoring data subsets by the long short-term memory algorithm, the dense linear features of the key traffic participating units included in any one of the historical traffic flow monitoring data subsets being the one-hot encoding for recording the historical road network flow state data represented by the key traffic participating units included in any one of the historical traffic flow monitoring data subsets; performing feature downsampling processing on the dense linear features of the key traffic participating units included in any one of the historical traffic flow monitoring data subsets by the long short-term memory algorithm to obtain the dense linear downsampling features of the key traffic participating units included in any one of the historical traffic flow monitoring data subsets; performing integration processing on the dense linear downsampling features of the key traffic participating units included in any one of the historical traffic flow monitoring data subsets by the long short-term memory algorithm to obtain the original spatiotemporal dynamic vectors of the key traffic participating units included in any one of the historical traffic flow monitoring data subsets.

[0084] In this embodiment, the core of the technical solution is how to accurately characterize the original spatiotemporal dynamic vectors of key traffic participants in any subset of historical traffic flow monitoring data with the help of the long short-term memory algorithm (LSTM). This technical path covers the extraction of dense linear features, feature downsampling processing, and final integration processing. Each step aims to deeply explore and refine the unique dynamic information of key traffic participants in historical time and space.

[0085] First of all, it is necessary to clarify that the dense linear features of the key traffic participating units included in any subset of historical traffic flow monitoring data are essentially obtained by one-hot encoding the historical road network flow state data represented by these units. One-hot encoding, as a common data processing method, can convert categorical variables into numerical form, which is convenient for machine learning models to understand and process. In this scenario, one-hot encoding is used to convert the traffic flow state data of key traffic participating units into dense and linear feature representation. Take the traffic lights at a certain intersection as an example. There are three states: red light, green light, and yellow light. Under one-hot encoding, the red light can be represented as [1, 0, 0], the green light as [0, 1, 0], and the yellow light as [0, 0, 1]. This encoding method allows each state to be uniquely and equidistantly represented in the numerical space, which facilitates subsequent analysis and processing.

[0086] Secondly, these dense linear features are downsampled using the long short-term memory algorithm. Feature downsampling, as the name implies, is a technique for reducing feature dimensions. It can reduce data redundancy and computational complexity while retaining the main information of the data set. In this step, the powerful time series processing capabilities of LSTM are used to downsample dense linear features to extract more compact and useful feature representations. Specifically, LSTM learns the temporal dependencies between features and selectively retains the information that is most critical to the prediction task, thereby achieving feature dimensionality reduction.

[0087] Then, the long short-term memory algorithm is used again to integrate the dense linear features that have been downsampled. In short, integration processing is a technology that combines multiple features or models to improve overall performance. In this scenario, the downsampled features are integrated through LSTM to form a comprehensive and rich original spatiotemporal dynamic vector. This vector not only contains the traffic flow status information of key traffic participants at different time points, but also incorporates the trends and patterns of how these status information changes over time. Taking the traffic flow at a certain intersection as an example, after integration processing, a vector of the form [0.1, 0.5, 0.3, ..., 0.2] can be obtained, in which each element represents the traffic characteristics of the intersection at different time points or under different traffic conditions.

[0088] Through the above processing flow, the original spatiotemporal dynamic vector of the key traffic participants in any subset of historical traffic flow monitoring data is finally obtained. This vector is a multi-dimensional data representation, which comprehensively and deeply records the traffic flow dynamic information of the key traffic participants in historical time and space. Compared with traditional traffic flow monitoring data, this vector is more compact, useful, and rich in spatiotemporal dynamic information. It can not only help to more accurately understand the spatiotemporal dynamic characteristics of key traffic participants, but also provide more comprehensive and accurate data support for the subsequent formulation of signal light control strategies.

[0089] In this way, the spatiotemporal dynamic information in the subset of historical traffic flow monitoring data is deeply mined through the long short-term memory algorithm, and the original spatiotemporal dynamic vectors of key traffic participants are determined accordingly. This process not only takes into account the historical road network flow status data of the key traffic participants themselves, but also refines and integrates this information through a series of steps such as dense linear feature extraction, feature downsampling processing, and integrated processing. Such a design enables a more comprehensive and in-depth understanding of the traffic flow dynamics of key traffic participants in historical time and space, providing more accurate and comprehensive data support for the subsequent formulation of signal light control strategies.

[0090] Furthermore, the advantage of this embodiment lies in its powerful time series processing and feature extraction capabilities. As an advanced recurrent neural network, LSTM can effectively process sequence data and capture the time series dependencies in the data. In the field of traffic flow monitoring, this capability is particularly important because changes in traffic flow are often closely related to time. By downsampling and integrating dense linear features through LSTM, the time series change pattern of traffic flow can be captured more accurately, thereby obtaining a more accurate and comprehensive original spatiotemporal dynamic vector.

[0091] In addition, this embodiment has good scalability and flexibility. In practical applications, the model structure and parameters of LSTM can be adjusted according to specific needs and characteristics of the data set to achieve the best feature extraction and integration effect. At the same time, this embodiment can also be combined with other advanced machine learning algorithms, such as deep learning, reinforcement learning, etc., to further improve the accuracy and efficiency of traffic flow monitoring and prediction.

[0092] In summary, the technical solution recorded in this embodiment provides a new and effective idea and method for the spatiotemporal dynamic analysis of key traffic participants in the historical traffic flow monitoring data subset. It can not only help to understand the spatiotemporal dynamic characteristics of traffic flow more deeply, but also provide more accurate and comprehensive data support for the subsequent signal light control strategy formulation. This embodiment is of great significance for improving the accuracy and intelligence level of urban traffic management, and also provides new ideas and methods for alleviating urban traffic congestion.

[0093] In an alternative embodiment, the generating of the original vehicle-road-cloud traffic spatiotemporal dataset based on the multiple traffic participation events includes: determining multiple historical vehicles, multiple historical navigation routes, and the number of times each historical vehicle activates each historical navigation route based on the multiple traffic participation events; determining original traffic participation units based on the multiple historical vehicles and the multiple historical navigation routes, respectively, wherein the historical road network flow state data represented by any original traffic participation unit is the signal light link data of a historical navigation route or the signal light link data of a historical vehicle; for any historical vehicle and any historical navigation route, determining the original road network transfer characteristics between the original traffic participation unit corresponding to the any historical vehicle and the original traffic participation unit corresponding to the any historical navigation route based on the number of times the any historical vehicle activates the any historical navigation route, wherein the number of times the any historical vehicle activates the any historical navigation route and the road network correlation represented by the original road network transfer characteristics between the original traffic participation unit corresponding to the any historical vehicle and the original traffic participation unit corresponding to the any historical navigation route have a mapping relationship; determining the original vehicle-road-cloud traffic spatiotemporal dataset based on the multiple original traffic participation units and the original road network transfer characteristics between every two original traffic participation units.

[0094] It can be understood that the technical solution involved in this embodiment focuses on how to generate an original vehicle-road-cloud traffic spatiotemporal dataset based on multiple traffic participation events. This process involves multiple key steps, including determining historical vehicles, historical navigation routes and their activation times, defining original traffic participation units, determining original road network transmission characteristics, and finally constructing an original vehicle-road-cloud traffic spatiotemporal dataset.

[0095] First, it is necessary to determine multiple historical vehicles and multiple historical navigation routes based on multiple traffic participation events, as well as the number of times each historical vehicle activated each historical navigation route. This process is actually a refinement and deconstruction of traffic participation events, so that we can have a deeper understanding of the interactive relationship between vehicles and navigation routes. Taking the traffic system of a certain city as an example, there are 1,000 traffic participation events, which involve 200 historical vehicles and 50 historical navigation routes. By analyzing these events, we can know the number of times each vehicle activated each route, such as vehicle A activated route B 10 times, vehicle C activated route D 5 times, etc.

[0096] Secondly, it is necessary to determine the original traffic participating units based on multiple historical vehicles and multiple historical navigation routes. Here, the historical road network flow state data represented by any original traffic participating unit is the signal light link data of a historical navigation route or the signal light link data of a historical vehicle. Signal light link data refers to the status information of a series of signal lights that a vehicle passes through during driving, including the location, color, change time, etc. of the signal lights. These data are crucial to understanding the traffic flow status of the road network. For example, vehicle A and route B can be regarded as two original traffic participating units. The signal light link data of vehicle A records the status of all signal lights it passes during driving, while the signal light link data of route B records the status of all signal lights encountered by vehicles passing through this route.

[0097] Then, for any historical vehicle and any historical navigation route, it is necessary to determine the original road network transfer characteristics between the original traffic participating unit corresponding to the historical vehicle and the original traffic participating unit corresponding to the historical navigation route based on the number of times the historical vehicle activated the historical navigation route. The original road network transfer characteristics actually characterize the road network correlation between the vehicle and the route, that is, the vehicle's driving frequency and habits on a certain route. Taking vehicle A and route B as an example, since vehicle A activated route B 10 times, it can be considered that the road network correlation between vehicle A and route B is strong, so the original road network transfer characteristics between them may have a higher value, such as 0.9. On the contrary, if vehicle A rarely or never activates route D, then the original road network transfer characteristics between them may have a lower value, such as 0.1.

[0098] Finally, it is necessary to determine the original vehicle-road-cloud traffic spatiotemporal dataset based on multiple original traffic participating units and the original road network transfer characteristics between each two original traffic participating units. This dataset is actually a multi-dimensional data matrix, in which each row represents an original traffic participating unit, each column represents another original traffic participating unit, and each element in the matrix represents the original road network transfer characteristics between the two original traffic participating units. For example, there are 200 original traffic participating units (including 100 vehicles and 100 routes), then the original vehicle-road-cloud traffic spatiotemporal dataset will be a 200x200 matrix. Each element in the matrix represents the road network correlation between two original traffic participating units, thus providing comprehensive vehicle-road-cloud traffic spatiotemporal information.

[0099] In this way, multiple traffic participation events can be effectively converted into original vehicle-road-cloud traffic spatiotemporal datasets, providing strong data support for subsequent traffic flow monitoring, prediction and optimization. This dataset not only contains detailed information on vehicles and routes, but also reveals the complex interactive relationships between them, thereby providing a deeper and more comprehensive view of the traffic network. In addition, this embodiment also has good flexibility and scalability. In practical applications, the definition of the original traffic participation unit and the calculation method of the original road network transmission characteristics can be adjusted according to specific needs and the characteristics of the data set to achieve the best data generation effect. At the same time, this embodiment can also be combined with other advanced traffic data analysis methods, such as traffic flow prediction, traffic congestion identification, etc., to further improve the intelligence level and efficiency of traffic management.

[0100] In another alternative embodiment, the long short-term memory algorithm includes a first AI algorithm branch and a second AI algorithm branch, and the algorithm configuration weight of the first AI algorithm branch is different from the algorithm configuration weight of the second AI algorithm branch; the determining of the spatiotemporal dynamic quantization vector of the key traffic participating units included in any one of the historical traffic flow monitoring data subsets by the long short-term memory algorithm according to the key traffic participating units and involved traffic participating units included in any one of the historical traffic flow monitoring data subsets includes: in response to the key traffic participating units included in any one of the historical traffic flow monitoring data subsets representing the signal light link data of a historical vehicle, the determining of the spatiotemporal dynamic quantization vector of the key traffic participating units included in the historical traffic flow monitoring data subsets by the first A The I algorithm branch determines the spatiotemporal dynamic quantization vector of the key traffic participating units included in any one of the historical traffic flow monitoring data subsets based on the key traffic participating units and the involved traffic participating units included in any one of the historical traffic flow monitoring data subsets; in response to the key traffic participating units included in any one of the historical traffic flow monitoring data subsets representing the signal light link data of a historical navigation route, the second AI algorithm branch determines the spatiotemporal dynamic quantization vector of the key traffic participating units included in any one of the historical traffic flow monitoring data subsets based on the key traffic participating units and the involved traffic participating units included in any one of the historical traffic flow monitoring data subsets.

[0101] In this embodiment, the branched application of the long short-term memory algorithm (LSTM) is further explored to adapt to the spatiotemporal dynamic quantitative analysis of different types of traffic participating units in the historical traffic flow monitoring data subset. This embodiment envisions that LSTM is not only a single algorithm model, but is differentiated into two AI algorithm branches with different algorithm configuration weights: the first AI algorithm branch and the second AI algorithm branch. The original intention of designing these two branches is to more accurately capture and quantify the spatiotemporal dynamic characteristics of different traffic participating units (i.e., historical vehicles and historical navigation routes).

[0102] First of all, it should be clear that the algorithm configuration weights of the first AI algorithm branch and the second AI algorithm branch are different, which means that they have different focuses and optimization strategies when processing data. This design allows the two branches to conduct more specialized learning and analysis for different types of traffic participating units, thereby improving the accuracy and effectiveness of the spatiotemporal dynamic quantization vector.

[0103] When any subset of historical traffic flow monitoring data includes key traffic participating units that represent the signal light link data of a historical vehicle, the technical solution chooses to analyze it through the first AI algorithm branch. This is because the signal light link data of historical vehicles usually contains rich time series information, such as the vehicle's position, speed, and the status of the passing signal lights at different time points. The first AI algorithm branch, with its specific algorithm configuration weights, is better at capturing complex patterns in such time series data and can generate quantitative vectors that reflect the spatiotemporal dynamic characteristics of historical vehicles. For example, for a historical vehicle that often travels on a specific route during peak hours in the morning and evening, the first AI algorithm branch can learn the periodicity of its driving pattern and reflect it in the spatiotemporal dynamic quantitative vector, which may include key indicators such as driving speed during peak hours and signal light waiting time.

[0104] On the other hand, when any subset of historical traffic flow monitoring data includes key traffic participating units that represent the signal light link data of a historical navigation route, the technical solution turns to the use of the second AI algorithm branch. The signal light link data of the historical navigation route focuses on the traffic flow conditions on the entire route, including the traffic volume at different time points, the frequency of signal light switching, and the resulting traffic congestion. The configuration weights of the second AI algorithm branch make it more suitable for analyzing such macro traffic flow characteristics, and can generate quantitative vectors that reflect the spatiotemporal dynamic characteristics of historical navigation routes. For example, for a historical navigation route connecting the city center and the suburbs, the second AI algorithm branch can capture the significant increase in traffic volume during specific time periods (such as the morning rush hour on weekdays), as well as traffic bottlenecks caused by improper signal light switching. These characteristics will be reflected in the spatiotemporal dynamic quantitative vector.

[0105] Through such a branched design, it is possible to respond more flexibly to different types of traffic participants and generate more accurate and detailed spatiotemporal dynamic quantitative vectors for them. These vectors not only contain the dynamic information of the traffic participants themselves, but also imply the interaction between them and other traffic participants (i.e., involved traffic participants), thereby providing a richer and more comprehensive data basis for subsequent traffic flow monitoring, prediction and optimization. In addition, the above branched design also brings advantages in algorithm optimization and computing efficiency. Since the two AI algorithm branches can conduct specialized learning and analysis for different types of traffic participants, they can be more efficient when processing the data types they are good at, thereby reducing unnecessary computing overhead. At the same time, this design also facilitates further optimization and adjustment of the algorithm to adapt to the dynamic changes in traffic flow and new monitoring needs.

[0106] In some exemplary embodiments, the determining of multiple historical traffic flow monitoring data subsets based on the historical vehicle-road-cloud traffic spatiotemporal dataset includes: for any historical traffic flow monitoring data subset, taking any historical traffic participant unit in the historical vehicle-road-cloud traffic spatiotemporal dataset as a key traffic participant unit; selecting at least one historical traffic participant unit whose road network correlation meets the second correlation requirement from each historical traffic participant unit in the historical vehicle-road-cloud traffic spatiotemporal dataset that is connected to the one key traffic participant unit through the historical road network transmission feature, as an involved traffic participant unit; determining the any historical traffic flow monitoring data subset based on the one key traffic participant unit and the involved traffic participant unit.

[0107] In this exemplary embodiment, the process of determining multiple historical traffic flow monitoring data subsets based on the historical vehicle-road-cloud traffic spatiotemporal data set is described in detail. This process involves the careful screening of historical traffic participating units and the in-depth analysis of road network relevance to ensure that the generated historical traffic flow monitoring data subsets can accurately reflect the spatiotemporal dynamic characteristics of traffic flow.

[0108] First, for any subset of historical traffic flow monitoring data, any historical traffic participant in the historical vehicle-road-cloud traffic spatiotemporal dataset is used as a key traffic participant. The key traffic participant can be a historical vehicle or a historical navigation route, which represents the core analysis object in the data subset. Taking the traffic system of a certain city as an example, the historical vehicle-road-cloud traffic spatiotemporal dataset contains 1,000 historical traffic participants, including both vehicles and navigation routes. When generating a subset of historical traffic flow monitoring data, vehicle A can be selected as the key traffic participant.

[0109] Secondly, it is necessary to select at least one historical traffic participant whose road network relevance meets the second relevance requirement from the various historical traffic participant units in the historical vehicle-road cloud traffic spatiotemporal data set that are connected to the key traffic participant unit through the historical road network transmission characteristics, as the involved traffic participant unit. The historical road network transmission characteristics characterize the interaction relationship between historical traffic participant units and the transmission mode of traffic flow, while the second relevance requirement is a preset threshold used to screen other traffic participant units with sufficiently strong road network relevance to the key traffic participant unit. Continuing with vehicle A as an example, the second relevance requirement is set to 0.5, that is, only historical traffic participant units whose road network relevance with vehicle A is greater than or equal to 0.5 are selected as involved traffic participant units. Through the analysis of the historical vehicle-road cloud traffic spatiotemporal data set, it is found that the road network relevance of vehicle B, vehicle C and navigation route D with vehicle A is 0.6, 0.7 and 0.8 respectively, so they are all selected as involved traffic participant units.

[0110] Then, any subset of historical traffic flow monitoring data will be determined based on the key traffic participant and the selected involved traffic participant. This data subset is actually a spatiotemporal data set that includes the key traffic participant and all involved traffic participants, which comprehensively records the traffic flow status of these traffic participants in a specific time period. Taking vehicle A as an example, its corresponding subset of historical traffic flow monitoring data will include vehicle A's own signal light link data (such as driving speed, number of stops, etc.), as well as vehicle B, vehicle C and navigation route D's related data (such as the number of encounters with vehicle A, driving time on the same road section, etc.).

[0111] With such a design, multiple historical traffic flow monitoring data subsets can be generated, each of which revolves around a key traffic participant unit and contains data of other traffic participants with strong road network association. These data subsets not only provide rich traffic flow information, but also reveal the interactive relationship between different traffic participants and the transmission mode of traffic flow. Taking the data subset of vehicle A as an example, the driving habits of vehicle A, the frequency of encounters with other vehicles, and the traffic flow conditions on different navigation routes can be understood by analyzing the data therein, thereby providing strong data support for monitoring, prediction and optimization of traffic flow. In addition, this embodiment also has good flexibility and scalability. In practical applications, the selection criteria of key traffic participants and involved traffic participants, as well as the numerical value of the second association requirement, can be adjusted according to specific needs and the characteristics of the data set. At the same time, this embodiment can also be combined with other advanced traffic data analysis methods, such as traffic congestion identification, traffic event detection, etc., to further improve the intelligent level and efficiency of traffic management. It can be seen that this exemplary embodiment provides an effective idea and method to determine multiple historical traffic flow monitoring data subsets based on the historical vehicle-road-cloud traffic spatiotemporal data set. It can not only help to gain a deeper understanding of the spatiotemporal dynamic characteristics of traffic flow, but also provide rich data support for subsequent traffic flow monitoring, prediction and optimization.

[0112] In some further exemplary embodiments, the long short-term memory algorithm is debugged based on the spatiotemporal dynamic quantization vectors of the key traffic participating units included in each historical traffic flow monitoring data subset to obtain the spatiotemporal signal light control algorithm, including: for any pair of historical traffic flow monitoring data subsets, based on the spatiotemporal dynamic quantization vectors of the key traffic participating units included in any pair of historical traffic flow monitoring data subsets, determining the traffic dispatch training score between the key traffic participating units included in any pair of historical traffic flow monitoring data subsets, the traffic dispatch training score between the key traffic participating units included in any pair of historical traffic flow monitoring data subsets being obtained by identifying the key traffic participating units included in any pair of historical traffic flow monitoring data subsets. The traffic dispatching score between the historical road network flow status data represented by the key traffic participating units; obtaining the a priori traffic dispatching score between the key traffic participating units included in any pair of historical traffic flow monitoring data subsets, the a priori traffic dispatching score between the key traffic participating units included in any pair of historical traffic flow monitoring data subsets is the traffic dispatching score between the historical road network flow status data represented by the key traffic participating units included in any pair of historical traffic flow monitoring data subsets obtained by annotation; based on the traffic dispatching training score and the a priori traffic dispatching score between the key traffic participating units included in at least one pair of historical traffic flow monitoring data subsets, the long short-term memory algorithm is debugged to obtain the spatiotemporal signal light control algorithm.

[0113] In this exemplary embodiment, the technical solution focuses on how to fine-tune the long short-term memory algorithm based on the spatiotemporal dynamic quantization vectors of key traffic participating units contained in each subset of historical traffic flow monitoring data, and finally obtain the spatiotemporal signal light control algorithm. This process involves multiple links such as in-depth analysis of historical traffic flow data, calculation of traffic dispatch scores, and algorithm debugging, aiming to ensure that the obtained spatiotemporal signal light control algorithm can accurately reflect the spatiotemporal dynamic characteristics of traffic flow and effectively improve the control efficiency of traffic lights.

[0114] First, for any pair of historical traffic flow monitoring data subsets, the traffic dispatch training score between these key traffic participating units is determined based on the spatiotemporal dynamic quantization vectors of the key traffic participating units included in the data subset. The score is obtained through identification, which reflects the traffic dispatch effect between the historical road network flow state data represented by these key traffic participating units. Taking the traffic system of a certain city as an example, there are two historical traffic flow monitoring data subsets, which contain key traffic participating units A and B respectively. By analyzing these two data subsets, the spatiotemporal dynamic quantization vectors of A and B can be obtained, and then the traffic dispatch training score between them can be calculated. This score can be a value between 0 and 1, such as 0.8, which means that the traffic dispatch effect between A and B is better.

[0115] Secondly, it is necessary to obtain the prior traffic dispatch score between any pair of key traffic participating units included in the historical traffic flow monitoring data subset. This score is obtained through annotation, and it also reflects the traffic dispatch effect between the historical road network flow state data represented by these key traffic participating units, but the difference is that it is given in advance based on historical experience or expert knowledge. Continuing with A and B as an example, the prior traffic dispatch score between A and B can be given, such as 0.7, based on historical records or expert evaluations.

[0116] Then, the LSTM algorithm is debugged based on the traffic dispatch training scores and prior traffic dispatch scores between key traffic participants included in at least one pair of historical traffic flow monitoring data subsets. The goal of this debugging process is to enable the LSTM algorithm to learn how to optimize the control strategy of traffic lights based on the spatiotemporal dynamic quantized vectors of key traffic participants, thereby maximizing the traffic dispatch effect. Specifically, the traffic dispatch training scores and prior traffic dispatch scores can be used as the input and expected output of the algorithm, and the algorithm can be repeatedly trained and the parameters of the algorithm adjusted so that the algorithm can accurately predict and optimize the control strategy of traffic lights.

[0117] During the debugging process, various optimization algorithms and techniques can be used to improve the accuracy and efficiency of the spatiotemporal traffic light control algorithm. For example, a gradient descent algorithm can be used to minimize the prediction error, or a cross-validation technique can be used to evaluate the generalization ability of the algorithm. At the same time, the input features of the algorithm can be selected and transformed according to the needs of the actual application scenario and the characteristics of the data set to further improve the performance and stability of the algorithm.

[0118] In this way, a finely tuned spatiotemporal signal light control algorithm can be obtained. This algorithm can optimize the control strategy of traffic lights in real time according to the spatiotemporal dynamic quantization vectors of key traffic participants, thereby effectively alleviating traffic congestion and improving road capacity. Taking A and B as an example, the debugged spatiotemporal signal light control algorithm can dynamically adjust the control strategy of traffic lights according to the real-time traffic flow status of A and B, so that the traffic dispatch effect between A and B is optimized.

[0119] In a subsequent possible embodiment, the obtaining of the prior traffic dispatch score between the key traffic participating units included in any pair of historical traffic flow monitoring data subsets includes: determining the prior traffic dispatch score between the key traffic participating units included in any pair of historical traffic flow monitoring data subsets as a first preset score based on the existence of historical road network transmission characteristics between the key traffic participating units included in any pair of historical traffic flow monitoring data subsets in the historical vehicle-road-cloud traffic spatiotemporal dataset; and determining the prior traffic dispatch score between the key traffic participating units included in any pair of historical traffic flow monitoring data subsets as a second preset score based on the absence of historical road network transmission characteristics between the key traffic participating units included in any pair of historical traffic flow monitoring data subsets in the historical vehicle-road-cloud traffic spatiotemporal dataset.

[0120] In this embodiment, the technical solution further explores and implements how to obtain the prior traffic dispatch score between key traffic participants included in any pair of historical traffic flow monitoring data subsets. This process is mainly based on the historical road network transmission characteristics between key traffic participants in the historical vehicle-road cloud traffic spatiotemporal data set to make judgments and settings, aiming to provide a more accurate and reasonable basis for determining the prior traffic dispatch score.

[0121] First, if any pair of key traffic participants included in the historical traffic flow monitoring data subset has obvious historical road network transmission characteristics in the historical vehicle-road cloud traffic spatiotemporal data set, then the prior traffic dispatch score between the key traffic participants is set to a higher first preset score. The historical road network transmission characteristics here refer to the significant correlation and transmission between key traffic participants in terms of traffic flow, driving trajectory, traffic light status, etc. Taking the traffic system of a certain city as an example, the historical vehicle-road cloud traffic spatiotemporal data set records the driving trajectory, speed, number of stops, and traffic light status of vehicles A and B in multiple time periods in detail. Through in-depth analysis of these data, it is found that vehicles A and B often meet on the same road section, and their traffic status data such as driving speed, number of stops, and traffic light status are highly correlated. This shows that there are obvious historical road network transmission characteristics between vehicles A and B, and the traffic dispatch effect between them is more significant. Therefore, the prior traffic dispatch score between vehicles A and B can be set to a higher first preset score, such as 0.9, to indicate that the traffic dispatch relationship between them is relatively close.

[0122] Secondly, if any pair of key traffic participants included in the historical traffic flow monitoring data subsets do not have obvious historical road network transmission characteristics in the historical vehicle-road cloud traffic spatiotemporal data set, then the prior traffic scheduling score between the pair of key traffic participants is set to a lower second preset score. This means that the traffic scheduling effect between the pair of key traffic participants may not be very significant, or the interaction relationship between them and the traffic flow transmission pattern are not very obvious. Continuing with vehicle A as an example, it is found in the historical vehicle-road cloud traffic spatiotemporal data set that there is no significant correlation between the traffic status data such as the driving trajectory, speed, number of stops, and traffic light status of vehicle A and another vehicle C, that is, there is no obvious historical road network transmission feature between them. Then the prior traffic scheduling score between vehicle A and vehicle C can be set to a lower second preset score, such as 0.3, to indicate that the traffic scheduling relationship between them is relatively weak.

[0123] In this way, the prior traffic dispatch score between any pair of key traffic participants included in the historical traffic flow monitoring data subset can be more reasonably set according to the historical road network transmission characteristics in the historical vehicle-road-cloud traffic spatiotemporal data set. This score reflects the traffic dispatch effect between key traffic participants and provides an important reference for subsequent algorithm debugging and optimization of spatiotemporal signal control strategies. At the same time, this scoring setting method based on historical road network transmission characteristics also has certain flexibility and scalability, and can adapt to changes in different traffic scenarios and needs.

[0124] In practical applications, the specific values ​​of the first preset score and the second preset score can be adjusted according to the characteristics of the traffic scene, the size of the data set, and the needs of algorithm debugging. For example, in urban areas with severe traffic congestion, in order to pay more attention to the traffic dispatching effect between key traffic participants with significant interactive relationships and traffic flow transmission patterns, the first preset score can be set higher, such as 0.95 or 1.0. In suburban or rural areas with relatively good traffic conditions, in order to avoid the algorithm from overly neglecting the traffic dispatching effect between key traffic participants whose interactive relationships are not obvious but may still have a certain impact, the second preset score can be set slightly higher, such as 0.4 or 0.5.

[0125] In addition, this embodiment can also be combined with other advanced traffic data analysis methods and machine learning algorithms to further improve the accuracy and reliability of the prior traffic scheduling score. For example, a clustering algorithm can be used to deeply mine and analyze the historical vehicle-road cloud traffic spatiotemporal data set to discover more potential historical road network transmission characteristics and traffic scheduling patterns. At the same time, the prior traffic scheduling score can also be dynamically adjusted in combination with real-time traffic flow data and historical traffic event data to adapt to the ever-changing traffic conditions and the real-time optimization requirements of the spatiotemporal signal control strategy. For example, if real-time traffic flow data shows that the degree of traffic congestion on a certain section of road has suddenly increased, the prior traffic scheduling score between the key traffic participating units on the section can be correspondingly increased to encourage the algorithm to pay more attention to the traffic scheduling effect in this area.

[0126] In summary, this embodiment provides a more accurate and reasonable idea and method to obtain the prior traffic dispatch score between any key traffic participating units included in the subset of historical traffic flow monitoring data. It can not only help to more accurately judge the traffic dispatch effect between key traffic participating units, but also provide strong data support for the subsequent algorithm debugging and optimization of spatiotemporal signal light control strategy. This embodiment is of great significance for promoting the development of intelligent transportation systems and improving the accuracy and intelligence level of urban traffic management. In future research, it is possible to further explore how to optimize the setting method of prior traffic dispatch scores, how to improve the matching degree between scores and actual application scenarios, and how to apply this embodiment to a wider range of traffic management and control scenarios. For example, it is possible to consider applying this embodiment to multiple fields such as urban traffic planning, traffic light optimization, and traffic congestion management to improve the overall operating efficiency of urban traffic and the travel experience of traffic participants. At the same time, it is also possible to combine the latest traffic science and technology development trends, such as autonomous driving, vehicle networking, etc., to further expand and improve the application scenarios and functions of this embodiment.

[0127] In some other independent embodiments, the method includes: obtaining a target vehicle-road-cloud traffic spatiotemporal data set, wherein the target vehicle-road-cloud traffic spatiotemporal data set includes multiple target traffic participating units and multiple target road network transmission characteristics, any target traffic participating unit represents a target road network flow state data, and any target road network transmission characteristic represents the road network correlation between two corresponding target traffic participating units; determining multiple target traffic flow monitoring data subsets based on the target vehicle-road-cloud traffic spatiotemporal data set, any target traffic flow monitoring data subset includes key traffic participating units and involved traffic participating units, and the involved traffic participating units are associated with the key traffic participating units through the target road network transmission characteristics; for any one The target traffic flow monitoring data subset determines the spatiotemporal dynamic quantization vector of the key traffic participating units included in any one of the target traffic flow monitoring data subsets through a spatiotemporal signal light control algorithm based on the key traffic participating units and involved traffic participating units included in any one of the target traffic flow monitoring data subsets. The spatiotemporal dynamic quantization vector of the key traffic participating units included in any one of the target traffic flow monitoring data subsets includes the spatiotemporal dynamic information of the target road network flow status data represented by the key traffic participating units included in any one of the target traffic flow monitoring data subsets and the spatiotemporal dynamic information of the target road network flow status data represented by the involved traffic participating units included in any one of the target traffic flow monitoring data subsets.

[0128] In this embodiment, the method is mainly carried out around the target vehicle-road-cloud traffic spatiotemporal data set, and through a series of processing steps, the spatiotemporal dynamic quantization vectors of key traffic participating units in each target traffic flow monitoring data subset are finally determined. This technical solution not only deeply explores the correlation between traffic participating units, but also fully utilizes the advantages of the spatiotemporal signal light control algorithm, providing strong support for precise traffic management and control.

[0129] First, it is necessary to obtain the target vehicle-road-cloud traffic spatiotemporal dataset. This dataset is the basis for subsequent analysis. It contains multiple target traffic participating units and multiple target road network transmission characteristics. Here, any target traffic participating unit represents a specific target road network flow state data, which may be the vehicle's driving speed, the state of traffic lights, the degree of congestion of the road section, etc. Any target road network transmission feature represents the road network correlation between the corresponding two target traffic participating units. This correlation may be based on the spatial distance between them, the similarity of driving trajectories, the transmission mode of traffic flow, etc.

[0130] With this data set, we can then use it to determine multiple target traffic flow monitoring data subsets. Each such subset includes key traffic participants and involved traffic participants. The key traffic participants are the core of the subset, while the involved traffic participants are associated with the key traffic participants through the target road network transmission characteristics. This association may mean that they have a certain mutual influence in terms of traffic flow, driving trajectory, traffic light status, etc.

[0131] Then, for any subset of target traffic flow monitoring data, the spatiotemporal dynamic quantization vector of the key traffic participant unit is determined through the spatiotemporal signal control algorithm. This vector contains rich information, which includes not only the spatiotemporal dynamic information of the target road network flow state data represented by the key traffic participant unit itself, but also the spatiotemporal dynamic information of the target road network flow state data represented by the involved traffic participant units associated with the key traffic participant unit. Such a design enables a more comprehensive understanding of the dynamic changes of the key traffic participant units in the spatiotemporal dimension, and how such changes are affected by other traffic participant units.

[0132] Taking the traffic system of a certain city as an example, the target vehicle-road cloud traffic spatiotemporal dataset records 100 target traffic participants, including 80 vehicles and 20 traffic lights. These traffic participants establish a complex association network through the target road network transmission characteristics. One of the key traffic participants is selected: a busy traffic light in the city center, as the key traffic participant of a certain target traffic flow monitoring data subset. The involved traffic participants associated with the traffic light include 10 surrounding vehicles and another 3 traffic lights.

[0133] Through the spatiotemporal signal light control algorithm, the spatiotemporal dynamic quantization vector of this key traffic light can be determined. This vector not only includes the spatiotemporal dynamic information of the traffic light itself, such as its signal state (red light, green light, yellow light) and duration in different time periods, but also includes the spatiotemporal dynamic information such as the driving speed, driving trajectory and number of stops of the 10 vehicles associated with it, as well as the spatiotemporal dynamic information such as the signal state and duration of the other three traffic lights.

[0134] With this spatiotemporal dynamic quantification vector, we can more accurately understand the dynamic changes of this key traffic light in the spatiotemporal dimension, and how this change is affected by surrounding vehicles and other traffic lights. This is of great significance for optimizing the control strategy of traffic lights, improving traffic efficiency, and reducing traffic congestion and traffic accidents.

[0135] In practical applications, this method can be widely used in many fields such as urban traffic management, traffic planning, and traffic light optimization. For example, in urban traffic management, the control strategy of traffic lights can be adjusted in real time by analyzing the spatiotemporal dynamic quantization vectors of key traffic participants to adapt to changes in traffic flow and improve traffic efficiency. In traffic planning, this method can be used to predict future trends in traffic flow and provide a scientific basis for the construction and transformation of traffic infrastructure. In traffic light optimization, the control effect of traffic lights can be improved by optimizing the spatiotemporal dynamic quantization vectors of key traffic participants, thereby reducing traffic congestion and the occurrence of traffic accidents.

[0136] In addition, this method can also be combined with other advanced traffic data analysis methods and machine learning algorithms to further improve the accuracy and efficiency of the analysis. For example, clustering algorithms can be used to preprocess the target vehicle-road-cloud traffic spatiotemporal data set, and traffic participants with similar spatiotemporal dynamic characteristics can be classified into the same category to more efficiently determine the target traffic flow monitoring data subset. At the same time, machine learning algorithms such as neural networks can also be used to train the spatiotemporal signal light control algorithm, so that it can more accurately determine the spatiotemporal dynamic quantitative vectors of key traffic participants.

[0137] In some other independent embodiments, the key traffic participating units included in any pair of target traffic flow monitoring data subsets include target key traffic participating units, and the target key traffic participating units represent the target road network flow status data to be coordinated with traffic lights; the method also includes: determining the traffic dispatching score between the key traffic participating units included in any pair of target traffic flow monitoring data subsets based on the spatiotemporal dynamic quantization vectors of the key traffic participating units included in any pair of target traffic flow monitoring data subsets, the traffic dispatching score between the key traffic participating units included in any pair of target traffic flow monitoring data subsets is the traffic dispatching score between the target road network flow status data represented by the key traffic participating units included in any pair of target traffic flow monitoring data subsets obtained by identification; based on the traffic dispatching score between the key traffic participating units included in any pair of target traffic flow monitoring data subsets being lower than a preset congestion tolerance score, regional signal light coordinated control is performed on the target road network flow status data to be coordinated with traffic lights represented by the target key traffic participating units.

[0138] In this embodiment, the specific application of the subset of traffic flow monitoring data is further explored, especially the coordinated control of traffic lights for key traffic participants. The core of this embodiment is to determine whether regional coordinated control of traffic lights is needed by analyzing and calculating the traffic dispatch scores between key traffic participants, in order to optimize traffic flow and reduce congestion.

[0139] First of all, it should be clear that in any pair of target traffic flow monitoring data subsets, one or more key traffic participating units are included. Among these key traffic participating units, there is a special target key traffic participating unit, which represents the target road network flow state data to be coordinated with signal lights. This target key traffic participating unit may be an important traffic node, such as a busy intersection or traffic hub, and the change of its traffic flow has a significant impact on the operation status of the entire road network.

[0140] Secondly, based on the spatiotemporal dynamic quantitative vectors of the key traffic participants included in any pair of target traffic flow monitoring data subsets, the traffic dispatch score between these key traffic participants can be further determined. This score is obtained by identifying and analyzing the interactions and correlations between the target road network flow state data represented by these key traffic participants. Specifically, the traffic dispatch score may take into account multiple factors, such as the density, speed, direction, etc. of traffic flow, as well as the spatiotemporal dynamic changes between these factors. For example, if the traffic flow density between two key traffic participants is high and the speed is slow, then the traffic dispatch score between them may be low, indicating that the traffic coordination between the two units needs to be further optimized.

[0141] Then, decisions need to be made based on this traffic dispatch score. If the traffic dispatch score between any pair of key traffic participants included in the target traffic flow monitoring data subset is lower than the preset congestion tolerance score, it means that the current traffic conditions may have reached or are close to the critical point of congestion, and timely intervention and optimization are required. At this time, the target road network flow status data represented by the target key traffic participant to be coordinated with the traffic lights can be coordinated with regional traffic lights.

[0142] Regional signal light coordination control is an advanced traffic management strategy that can optimize the traffic flow of the entire region by coordinating the signal light control strategies of different intersections or traffic nodes. For example, if a key traffic participant is a busy intersection and its traffic dispatch score with other key traffic participants is low, then the signal light control strategy of this intersection and surrounding intersections can be adjusted to optimize the distribution of traffic flow and travel speed, thereby reducing congestion and improving traffic efficiency.

[0143] In practical applications, take a busy commercial district in a city as an example. There are multiple key traffic nodes in the area, including major intersections, bus stops, and transportation hubs. By applying the above technical solutions, the traffic dispatch scores between these key traffic participants can be monitored and analyzed in real time. If it is found that the score between a certain intersection and other key nodes is low, the regional signal light collaborative control strategy can be immediately activated to adjust the signal light control strategy of the intersection and surrounding intersections. In this way, not only can the congestion of the intersection be effectively alleviated, but also the traffic flow distribution of the entire commercial district can be optimized, and traffic efficiency and service level can be improved.

[0144] In addition, this method can also be combined with other traffic management strategies and technologies to achieve more comprehensive traffic optimization. For example, this method can be combined with an intelligent transportation system to more accurately calculate traffic dispatch scores and perform coordinated control of traffic lights by collecting and analyzing traffic flow data, traffic light status data and other information in real time. At the same time, big data and machine learning technologies can also be used to explore the potential laws and patterns of traffic flow data, providing a more scientific basis and prediction for coordinated control of traffic lights.

[0145] Further, Figure 2 Schematic diagram of a vehicle-road-cloud collaborative control system 200 provided in an embodiment of the present invention. Figure 2 The vehicle-road-cloud collaborative control system 200 shown includes a processor 210, which can call and run a computer program from a memory to implement the method in the embodiment of the present invention.

[0146] Alternatively, if Figure 2 As shown, the vehicle-road-cloud collaborative control system 200 may further include a memory 230. The processor 210 may call and run a computer program from the memory 230 to implement the method in the embodiment of the present invention.

[0147] The memory 230 may be a separate device independent of the processor 210 , or may be integrated into the processor 210 .

[0148] Alternatively, if Figure 2 As shown, the vehicle-road-cloud collaborative control system 200 may also include a transceiver 220, and the processor 210 may control the transceiver 220 to interact with other devices. Specifically, it may send information or data to other devices, or receive information or data sent by other devices.

[0149] Optionally, the vehicle-road-cloud collaborative control system 200 can implement the corresponding processes corresponding to the storage engine or components in the storage engine (such as a processing module) or a device deployed with a storage engine in each method of the embodiments of the present invention. For the sake of brevity, they will not be repeated here.

[0150] It should be understood that the processor of the embodiment of the present invention may be an integrated circuit chip with signal processing capability.

[0151] It is understood that the memory in the embodiments of the present invention may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memory. It should be noted that the memory of the systems and methods described herein is intended to include but is not limited to suitable types of memory.

[0152] Based on the above, a readable storage medium is provided, on which a program or instruction is stored, and when the program or instruction is executed by a processor, the steps of the above method are implemented.

[0153] It should be noted that, in this article, the terms "comprise", "include" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the presence of other identical elements in the process, method, article or device including the element. In addition, it should be pointed out that the scope of the method and device in the embodiment of the present invention is not limited to performing functions in the order shown or discussed, and may also include performing functions in a substantially simultaneous manner or in reverse order according to the functions involved, for example, the described method may be performed in an order different from that described, and various steps may also be added, omitted, or combined. In addition, the features described with reference to certain examples may be combined in other examples.

[0154] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a computer software product, which is stored in a storage medium (such as ROM / RAM, a magnetic disk, or an optical disk), and includes a number of instructions for enabling a terminal (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present invention.

[0155] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present invention, ordinary technicians in this field can also make many forms without departing from the purpose of the present invention and the scope of protection of the present invention, all of which are within the protection of the present invention.

Claims

1. A road traffic signal light control method based on vehicle-road-cloud collaboration, characterized in that: Applied to a vehicle-road-cloud collaborative control system, the method includes: Obtain a historical vehicle-road-cloud traffic spatiotemporal data set, wherein the historical vehicle-road-cloud traffic spatiotemporal data set includes a plurality of historical traffic participating units and a plurality of historical road network transmission characteristics, wherein any historical traffic participating unit represents a historical road network flow state data, and any historical road network transmission characteristic represents the road network correlation between two corresponding historical traffic participating units; Determine a plurality of historical traffic flow monitoring data subsets based on the historical vehicle-road-cloud traffic spatiotemporal data set, any of which includes key traffic participant units and involved traffic participant units, and the involved traffic participant units are associated with the key traffic participant units through historical road network transmission characteristics; For any subset of historical traffic flow monitoring data, determine the spatiotemporal dynamic quantization vector of the key traffic participating units included in any subset of historical traffic flow monitoring data based on the key traffic participating units and involved traffic participating units included in any subset of historical traffic flow monitoring data by using a long short-term memory algorithm, the spatiotemporal dynamic quantization vector of the key traffic participating units included in any subset of historical traffic flow monitoring data includes the spatiotemporal dynamic information of the historical road network flow state data represented by the key traffic participating units included in any subset of historical traffic flow monitoring data and the spatiotemporal dynamic information of the historical road network flow state data represented by the involved traffic participating units included in any subset of historical traffic flow monitoring data; According to the spatiotemporal dynamic quantization vectors of the key traffic participating units included in each historical traffic flow monitoring data subset, the long short-term memory algorithm is debugged to obtain a spatiotemporal signal light control algorithm, the spatiotemporal signal light control algorithm is used to determine the spatiotemporal dynamic quantization vectors of the key traffic participating units included in the target traffic flow monitoring data subset; and the spatiotemporal dynamic quantization vectors of the key traffic participating units are used to make control decisions on the road traffic signal light relationship network; The long short-term memory algorithm is debugged based on the spatiotemporal dynamic quantization vectors of key traffic participating units included in each historical traffic flow monitoring data subset to obtain the spatiotemporal signal light control algorithm, including: For any pair of historical traffic flow monitoring data subsets, determining the traffic dispatch training score between the key traffic participating units included in any pair of historical traffic flow monitoring data subsets based on the spatiotemporal dynamic quantization vectors of the key traffic participating units included in any pair of historical traffic flow monitoring data subsets, wherein the traffic dispatch training score between the key traffic participating units included in any pair of historical traffic flow monitoring data subsets is obtained by identifying the traffic dispatch scores between the historical road network flow state data represented by the key traffic participating units included in any pair of historical traffic flow monitoring data subsets; Obtaining a priori traffic dispatching scores between key traffic participating units included in any pair of historical traffic flow monitoring data subsets, wherein the a priori traffic dispatching scores between key traffic participating units included in any pair of historical traffic flow monitoring data subsets are obtained by annotation of traffic dispatching scores between historical road network flow state data represented by key traffic participating units included in any pair of historical traffic flow monitoring data subsets; Based on the traffic dispatch training scores and prior traffic dispatch scores between key traffic participating units included in at least one pair of historical traffic flow monitoring data subsets, the long short-term memory algorithm is debugged to obtain a spatiotemporal signal light control algorithm; The obtaining of the prior traffic dispatch score between the key traffic participating units included in any pair of historical traffic flow monitoring data subsets includes: According to the existence of historical road network transmission characteristics between the key traffic participating units included in any pair of historical traffic flow monitoring data subsets in the historical vehicle-road-cloud traffic spatiotemporal data set, the priori traffic dispatch score between the key traffic participating units included in any pair of historical traffic flow monitoring data subsets is determined to be a first preset score; Based on the fact that there is no historical road network transmission feature between the key traffic participating units included in any pair of historical traffic flow monitoring data subsets in the historical vehicle-road-cloud traffic spatiotemporal dataset, the prior traffic scheduling score between the key traffic participating units included in any pair of historical traffic flow monitoring data subsets is determined to be the second preset score.

2. The method according to claim 1, characterized in that: The acquisition of the historical vehicle-road-cloud traffic spatiotemporal data set includes: Obtain multiple traffic participation events, where any traffic participation event is used to represent an event in which a historical vehicle activates a historical navigation route; Generate an original vehicle-road cloud traffic spatiotemporal data set based on the multiple traffic participation events, the original vehicle-road cloud traffic spatiotemporal data set includes multiple original traffic participation units and multiple original road network transfer features, any one of the original traffic participation units represents a historical road network flow state data, and any one of the original road network transfer features represents the road network correlation between the corresponding two original traffic participation units; The original road network transfer features whose road network correlation meets the first correlation requirement in the original vehicle-road-cloud traffic spatiotemporal dataset are removed to obtain a historical vehicle-road-cloud traffic spatiotemporal dataset.

3. The method according to claim 2, characterized in that The generating of the original vehicle-road-cloud traffic spatiotemporal dataset according to the multiple traffic participation events includes: Determine the signal light link data of a plurality of historical navigation routes and the intersection statistics of every two historical navigation routes according to the plurality of traffic participation events, wherein the intersection statistics of any two historical navigation routes is the number of times the two historical navigation routes are activated by the same historical vehicle; Determine the plurality of original traffic participating units according to the signal light link data of the plurality of historical navigation routes, wherein the historical road network flow state data represented by any one of the original traffic participating units is the signal light link data of a historical navigation route; Determine, based on the cross-statistics of the arbitrary two historical navigation routes, the original road network transfer characteristics between the original traffic participating units corresponding to the arbitrary two historical navigation routes, wherein the cross-statistics of the arbitrary two historical navigation routes are the road network relevance characterized by the original road network transfer characteristics between the original traffic participating units corresponding to the arbitrary two historical navigation routes; The original vehicle-road-cloud traffic spatiotemporal data set is determined based on the multiple original traffic participating units and the original road network transfer characteristics between every two original traffic participating units.

4. The method according to claim 3, characterized in that The method of determining the spatiotemporal dynamic quantization vector of the key traffic participating units included in any one of the historical traffic flow monitoring data subsets by using the long short-term memory algorithm according to the key traffic participating units and involved traffic participating units included in any one of the historical traffic flow monitoring data subsets includes: Determine the original spatiotemporal dynamic vectors of the key traffic participating units included in any one of the historical traffic flow monitoring data subsets and the original spatiotemporal dynamic vectors of the involved traffic participating units included in any one of the historical traffic flow monitoring data subsets by using a long short-term memory algorithm, the original spatiotemporal dynamic vectors of the key traffic participating units included in any one of the historical traffic flow monitoring data subsets are used to record the spatiotemporal dynamic information of the historical road network flow state data represented by the key traffic participating units included in any one of the historical traffic flow monitoring data subsets, and the original spatiotemporal dynamic vectors of the involved traffic participating units included in any one of the historical traffic flow monitoring data subsets are used to record the spatiotemporal dynamic information of the historical road network flow state data represented by the involved traffic participating units included in any one of the historical traffic flow monitoring data subsets; The original spatiotemporal dynamic vectors of the key traffic participating units included in any one of the historical traffic flow monitoring data subsets and the original spatiotemporal dynamic vectors of the involved traffic participating units included in any one of the historical traffic flow monitoring data subsets are combined to obtain the spatiotemporal dynamic quantization vectors of the key traffic participating units included in any one of the historical traffic flow monitoring data subsets.

5. The method according to claim 4, characterized in that The method of determining the original spatiotemporal dynamic vector of the key traffic participating units included in any one subset of historical traffic flow monitoring data by using the long short-term memory algorithm includes: Determine the dense linear features of the key traffic participating units included in any one of the historical traffic flow monitoring data subsets by a long short-term memory algorithm, wherein the dense linear features of the key traffic participating units included in any one of the historical traffic flow monitoring data subsets are used to record the unique hot encoding of the historical road network flow state data represented by the key traffic participating units included in any one of the historical traffic flow monitoring data subsets; Performing feature downsampling processing on the dense linear features of the key traffic participating units included in any one of the historical traffic flow monitoring data subsets by using the long short-term memory algorithm to obtain the dense linear downsampling features of the key traffic participating units included in any one of the historical traffic flow monitoring data subsets; The dense linear down-sampling features of the key traffic participating units included in any subset of the historical traffic flow monitoring data are integrated and processed by the long short-term memory algorithm to obtain the original spatiotemporal dynamic vectors of the key traffic participating units included in any subset of the historical traffic flow monitoring data.

6. The method according to claim 2, characterized in that The generating of the original vehicle-road-cloud traffic spatiotemporal dataset according to the multiple traffic participation events includes: Determining a plurality of historical vehicles, a plurality of historical navigation routes, and a number of times each historical vehicle activated each historical navigation route according to the plurality of traffic participation events; Determine original traffic participating units respectively according to the multiple historical vehicles and the multiple historical navigation routes, the historical road network flow state data represented by any original traffic participating unit is the signal light link data of a historical navigation route or the signal light link data of a historical vehicle; For any historical vehicle and any historical navigation route, according to the number of times the any historical vehicle activates the any historical navigation route, an original road network transfer feature between an original traffic participating unit corresponding to the any historical vehicle and an original traffic participating unit corresponding to the any historical navigation route is determined, and the number of times the any historical vehicle activates the any historical navigation route and the road network association represented by the original road network transfer feature between the original traffic participating unit corresponding to the any historical vehicle and the original traffic participating unit corresponding to the any historical navigation route have a mapping relationship; The original vehicle-road-cloud traffic spatiotemporal data set is determined based on the multiple original traffic participating units and the original road network transfer characteristics between every two original traffic participating units.

7. The method according to claim 6, characterized in that The long short-term memory algorithm includes a first AI algorithm branch and a second AI algorithm branch, and the algorithm configuration weight of the first AI algorithm branch is different from the algorithm configuration weight of the second AI algorithm branch; the determining of the spatiotemporal dynamic quantization vector of the key traffic participating units included in any one of the historical traffic flow monitoring data subsets by the long short-term memory algorithm based on the key traffic participating units and involved traffic participating units included in any one of the historical traffic flow monitoring data subsets includes: In response to the key traffic participant unit included in any one of the historical traffic flow monitoring data subsets representing the signal light link data of a historical vehicle, the first AI algorithm branch determines the spatiotemporal dynamic quantization vector of the key traffic participant unit included in any one of the historical traffic flow monitoring data subsets according to the key traffic participant unit and involved traffic participant units included in any one of the historical traffic flow monitoring data subsets; In response to the key traffic participating units included in any one of the historical traffic flow monitoring data subsets representing the signal light link data of a historical navigation route, the second AI algorithm branch is used to determine the spatiotemporal dynamic quantization vector of the key traffic participating units included in any one of the historical traffic flow monitoring data subsets based on the key traffic participating units and involved traffic participating units included in any one of the historical traffic flow monitoring data subsets.

8. The method according to claim 1, characterized in that Determining a plurality of historical traffic flow monitoring data subsets based on the historical vehicle-road-cloud traffic spatiotemporal data set includes: For any subset of historical traffic flow monitoring data, any historical traffic participating unit in the historical vehicle-road-cloud traffic spatiotemporal data set is taken as a key traffic participating unit; Select at least one historical traffic participant unit whose road network relevance meets the second relevance requirement from each historical traffic participant unit in the historical vehicle-road cloud traffic spatiotemporal data set that is connected to the one key traffic participant unit through the historical road network transmission feature as the involved traffic participant unit; According to the one key traffic participant unit and the involved traffic participant units, determine any one subset of historical traffic flow monitoring data.

9. A vehicle-road-cloud collaborative control system, characterized in that: The method comprises at least one processor and a memory; the memory stores computer-executable instructions; the at least one processor executes the computer-executable instructions stored in the memory, so that the at least one processor executes the method according to any one of claims 1 to 8.