Traffic signal dynamic induction control method and device based on large language model

Abstract

The invention discloses a traffic signal dynamic induction control method and device based on a large language model, and the method comprises the steps: collecting the traffic flow information of a target region in real time, and calculating a comprehensive control excitation value of a current phase according to the traffic flow information, the traffic flow information is used for indicating traffic information of a plurality of objects passing through the target area, and the current phase is used for indicating the current stage of traffic signal control; performing signal control effect prediction according to the comprehensive control excitation value through a signal control effect prediction large model to obtain weighted vehicle average delay of a plurality of signal control actions, the plurality of signal control actions including phase jump switching, green light early cut-off, green light extension and green light invariability; a target signal control action is decided from the multiple signal control actions according to the multiple weighted vehicle average delays, the target signal control action is issued to a signal machine, the signal machine is arranged in the target area and used for outputting traffic signals, and the multiple weighted vehicle average delays correspond to the multiple signal control actions in a one-to-one mode.

Description

Technical Field

[0001] This application relates to the field of transportation, and more specifically, to a method and apparatus for dynamic sensing control of traffic signals based on a large language model. Background Technology

[0002] Traffic signal sensing control is a mature traffic control technology. This technology uses induction coils or cameras to identify vehicle arrivals and determines whether the current green light should be extended or ended earlier based on simple headway logic. However, traditional sensing control is only suitable for low-saturation scenarios. In medium-to-high saturation scenarios or scenarios with multiple objectives and strategies in conflict, such as public transport priority or emergency vehicle requests, sensing control requires the addition of a large number of complex logic rules, making it difficult to develop and maintain. Furthermore, the model parameters cannot be automatically adjusted, which can lead to sensing control failure.

[0003] In recent years, large language models have developed rapidly. Although they are mainly used for text processing, their powerful sequence modeling, contextual reasoning and multi-source information fusion capabilities have provided new ideas for traffic signal sensing control.

[0004] In related technologies, some traffic signal sensing control methods based on large language models have been proposed. However, these existing technologies are based on simple sensing control rules and can only be used at intersections with low saturation. In medium to high saturation conditions, vehicles arrive continuously in almost every phase. According to the logic of sensing control, when issuing green lights for each phase sequentially, the continuous arrival of vehicles tends to lead to the execution of the maximum green light each time. In this scenario, it is foreseeable that this proposal will result in long queues in multiple directions, with extremely poor control performance, even worse than manually configured timing schemes.

[0005] In complex traffic scenarios, traditional sensing control methods struggle to effectively integrate multiple policy requirements such as pedestrian and public transport priority, and rapid response of emergency vehicles, leading to poor control performance and a lack of transparency in the decision-making process. No effective solution has yet been proposed. Summary of the Invention

[0006] This application provides a traffic signal dynamic sensing control method and apparatus based on a large language model, which at least solves the problem in related technologies that, in complex traffic scenarios, traditional sensing control methods are difficult to effectively integrate multiple strategy requirements such as pedestrian, public transport priority, and rapid response of emergency vehicles, resulting in poor control performance and a lack of transparency in the decision-making process.

[0007] According to one embodiment of this application, a traffic signal dynamic sensing control method based on a large language model is provided, comprising: real-time acquisition of traffic flow information of a target area, and calculation of a comprehensive control incentive value of the current phase based on the traffic flow information, wherein the traffic flow information is used to indicate traffic information of multiple objects passing through the target area, and the current phase is used to indicate the current stage of traffic signal control; predicting the signal control effect based on the comprehensive control incentive value using a large signal control effect prediction model to obtain a weighted average vehicle delay of multiple signal control actions, wherein the multiple signal control actions include: phase switching, early green light termination, extended green light, and unchanged green light; determining a target signal control action from the multiple signal control actions based on the multiple weighted average vehicle delays, and issuing the target signal control action to the traffic signal controller, wherein the traffic signal controller is set in the target area, the traffic signal controller is used to output traffic signals, and the multiple weighted average vehicle delays correspond one-to-one with the multiple signal control actions.

[0008] In an exemplary embodiment, real-time acquisition of traffic flow information in a target area includes: real-time acquisition of vehicle arrival information through sensing devices deployed in the target area, wherein the vehicle arrival information includes: the average vehicle arrival rate of multiple lanes in the current phase, the arrival time of the target vehicle in each lane in the current phase, and the saturation of the time window [t-Δt,t]; the target vehicle is the last arriving vehicle in the current phase, and the saturation is calculated based on traffic flow; detecting pedestrian crossing data in the current green light phase, wherein the pedestrian crossing data is used to indicate whether there are pedestrians who have not completed crossing the street in the current green light phase; detecting whether there is a bus priority request in the current phase and obtaining bus priority request data; detecting whether there is an emergency vehicle passage request in the current phase and obtaining emergency passage request data, wherein the traffic flow information includes: the vehicle arrival information, the pedestrian crossing data, the bus priority request data, and the emergency passage request data.

[0009] In an exemplary embodiment, calculating the comprehensive control incentive value for the current phase based on the traffic flow information includes: weighting and summing the saturation, the average vehicle arrival rate, and the window duration according to the traffic flow weight of the saturation, the vehicle arrival rate weight of the average vehicle arrival rate, and the window penalty weight of the window duration to obtain a first incentive value, wherein the window duration is the time difference between the current time and the arrival time, and the window penalty weight is negative; calculating a second incentive value based on the pedestrian crossing data, the pedestrian incentive gain corresponding to the pedestrian crossing data, the bus priority request data, the bus incentive gain corresponding to the bus priority request data, the emergency passage request data, and the emergency vehicle incentive gain corresponding to the emergency passage request data; and calculating the comprehensive control incentive value for the current phase i based on the first incentive value and the second incentive value.

[0010] In an exemplary embodiment, before predicting the signal control effect based on the comprehensive control incentive value using the large-scale signal control effect prediction model, the method further includes: collecting initial training data, wherein the initial training data includes: traffic flow data of multiple phases in the target area, comprehensive control incentive value, signal control scheme data for the current period, and initial signal control actions; processing the initial training data using the large-scale initial signal control effect prediction model to obtain the initial weighted average vehicle delay corresponding to the initial signal control action; calculating a training loss based on the initial weighted average vehicle delay and the actual weighted average vehicle delay corresponding to the initial signal control action, and adjusting the model parameters of the large-scale initial signal control effect prediction model based on the training loss to obtain the large-scale signal control effect prediction model.

[0011] In an exemplary embodiment, determining a target traffic control action from a plurality of traffic control actions based on a plurality of weighted average vehicle delays includes: normalizing the plurality of weighted average vehicle delays to obtain a plurality of normalized weighted average vehicle delays; calculating an expected effect score for each traffic control action based on the plurality of normalized weighted average vehicle delays using a sigmoid function; calculating a utility score for each of the plurality of traffic control actions based on the plurality of expected effect scores using a softmax function, wherein the plurality of expected effect scores correspond one-to-one with the plurality of traffic control actions; and determining the target traffic control action from the plurality of traffic control actions based on the plurality of utility scores, wherein the plurality of utility scores correspond one-to-one with the plurality of traffic control actions.

[0012] In an exemplary embodiment, determining the target signal control action from the plurality of signal control actions based on a plurality of utility scores includes: sorting the plurality of signal control actions according to the plurality of utility scores to obtain a plurality of sorted signal control actions; determining whether the utility scores of a plurality of first signal control actions among the sorted plurality of signal control actions are greater than corresponding confidence thresholds, wherein the plurality of first signal control actions correspond one-to-one with a plurality of confidence thresholds, and the plurality of first signal control actions include: phase switching, early green light termination, and green light extension; if it is determined that the utility score of at least one second signal control action among the plurality of first signal control actions is greater than the corresponding confidence threshold, determining the third signal control action among the at least one second signal control action as the target signal control action, wherein the utility score of the third signal control action is ranked first among the at least one second signal control action; if it is determined that the utility scores of all the plurality of first signal control actions are less than or equal to the corresponding confidence thresholds, determining the fourth signal control action among the plurality of signal control actions as the target signal control action, wherein the fourth signal control action is the green light remaining unchanged.

[0013] In an exemplary embodiment, before sending the target signal control action to the traffic signal, the method further includes: performing an execution verification on the target signal control action to determine whether the green light duration of the traffic signal after executing the target signal control action is within a preset range; if it is determined that the target signal control action fails the execution verification, updating the target signal control action to a fourth signal control action and sending the updated target signal control action to the traffic signal, wherein the fourth signal control action is that the green light remains unchanged; if it is determined that the target signal control action passes the execution verification, sending the target signal control action to the traffic signal.

[0014] According to another embodiment of this application, a traffic signal dynamic sensing control device based on a large language model is also provided, comprising: a calculation module, used to collect traffic flow information of a target area in real time, and calculate a comprehensive control incentive value of the current phase based on the traffic flow information, wherein the traffic flow information is used to indicate traffic information of multiple objects passing through the target area, and the current phase is used to indicate the current stage of traffic signal control; a prediction module, used to predict the signal control effect based on the comprehensive control incentive value using a large signal control effect prediction model, and obtain a weighted average vehicle delay of multiple signal control actions, wherein the multiple signal control actions include: phase switching, early green light termination, extended green light, and unchanged green light; and a decision module, used to decide a target signal control action from the multiple signal control actions based on the multiple weighted average vehicle delays, and send the target signal control action to the traffic signal controller, wherein the traffic signal controller is set in the target area, the traffic signal controller is used to output traffic signals, and the multiple weighted average vehicle delays correspond one-to-one with the multiple signal control actions.

[0015] According to another aspect of the embodiments of this application, a computer-readable storage medium is also provided, wherein a computer program is stored in the computer program, and the computer program is configured to execute the above-described traffic signal dynamic sensing control method based on a large language model when it is run.

[0016] According to another aspect of the embodiments of this application, an electronic device is also provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the above-described traffic signal dynamic sensing control method based on a large language model through the computer program.

[0017] This application proposes a dynamic sensing control method for traffic signals. Its core lies in using a large language model to handle complex traffic scenarios. The specific process includes: acquiring real-time traffic flow information within the target area; dynamically calculating the comprehensive control incentive value for the current traffic phase based on this information; then predicting the weighted average vehicle delay after implementing different signal control actions using a large signal control effect prediction model, thereby evaluating the effect of each action; finally, selecting the target signal control action that best reduces delay based on the predicted weighted average vehicle delay. This decision-making process ensures optimal traffic flow efficiency. After selecting the action, the instruction is transmitted to the signal controller located within the target area to adjust the traffic signal to implement the optimal control strategy. Using the above scheme, this application achieves intelligent, efficient, and flexible urban traffic signal control by integrating real-time traffic data, defining a comprehensive incentive value, and utilizing a large model for predictive decision-making. It can respond to traffic changes in real time, significantly improving traffic efficiency and safety, and is particularly suitable for complex traffic environments with multiple strategies coexisting. Furthermore, it solves the problem in related technologies where, in complex traffic scenarios, traditional sensing control methods struggle to effectively integrate multiple strategy requirements such as pedestrian, public transport priority, and rapid emergency vehicle response, leading to poor control effects and a lack of transparency in the decision-making process. Attached Figure Description

[0018] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0019] Figure 1 This is a hardware structure block diagram of a computer terminal for an optional traffic signal dynamic sensing control method based on a large language model, according to an embodiment of this application.

[0020] Figure 2 This is a flowchart of an optional traffic signal dynamic sensing control method based on a large language model, according to an embodiment of this application.

[0021] Figure 3 This is a flowchart illustrating an optional traffic signal dynamic sensing control method based on a large language model, according to an embodiment of this application.

[0022] Figure 4 This is a schematic diagram illustrating the calculation result of an optional expected effect score according to an embodiment of this application;

[0023] Figure 5 This is a structural block diagram of a traffic signal dynamic sensing control device based on a large language model, according to an embodiment of this application. Detailed Implementation

[0024] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0025] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0026] The methods and embodiments provided in this application can be executed on a computer terminal or similar computing system. Taking running on a computer terminal as an example, Figure 1 This is a hardware structure block diagram of a computer terminal for a traffic signal dynamic sensing control method based on a large language model, according to an embodiment of this application. Figure 1 As shown, a computer terminal may include one or more ( Figure 1 Only one is shown in the diagram. A processor 102 (which may include, but is not limited to, a microprocessor MCU or a programmable logic device FPGA, etc.) and a memory 104 for storing data are also shown. In one exemplary embodiment, the computer terminal may further include a transmission device 106 for communication functions and an input / output device 108. Those skilled in the art will understand that... Figure 1 The structure shown is for illustrative purposes only and does not limit the structure of the computer terminal described above. For example, the computer terminal may also include components that are more complex than those described above. Figure 1 The more or fewer components shown, or having the same Figure 1 Equivalent functions or ratios shown Figure 1 The functions shown have more different configurations.

[0027] The memory 104 can be used to store computer programs, such as application software programs and modules, like the computer program corresponding to the traffic signal dynamic sensing control method based on a large language model in this embodiment. The processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, thereby implementing the above-described method. The memory 104 may include high-speed random access memory and may also include non-volatile memory, such as one or more magnetic storage systems, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include memory remotely located relative to the processor 102, and these remote memories can be connected to a secure text network via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0028] The transmission device 106 is used to receive or send data via a network. Specific examples of the network described above may include a wireless network provided by a communication provider for the computer terminal. In one example, the transmission system 106 includes a Network Interface Controller (NIC), which can connect to other network devices via a base station to communicate with the Internet.

[0029] This embodiment provides a traffic signal dynamic sensing control method based on a large language model. Figure 2 This is a flowchart of an optional traffic signal dynamic sensing control method based on a large language model according to an embodiment of this application. The process includes the following steps:

[0030] Step S202: Real-time acquisition of traffic flow information in the target area, and calculation of the comprehensive control excitation value of the current phase based on the traffic flow information, wherein the traffic flow information is used to indicate the traffic information of multiple objects passing through the target area, and the current phase is used to indicate the current stage of traffic signal control;

[0031] It should be noted that in the field of traffic engineering, especially in traffic signal control, "phase" is an important concept used to describe the pattern by which traffic lights allocate green light time to different directions or different types of traffic flows.

[0032] In the embodiments of this application, "phase" is used to distinguish different stages or time periods of traffic signal control. Each phase has its own traffic detection data, saturation, vehicle priority and other information.

[0033] Step S204: The signal control effect prediction model is used to predict the signal control effect based on the comprehensive control excitation value, and the weighted average vehicle delay of multiple signal control actions is obtained. The multiple signal control actions include: phase switching, early green light termination, extended green light, and unchanged green light.

[0034] It should be noted that "green light unchanged" means maintaining the green light duration of this phase in the background scheme without increasing or decreasing it; "green light extended" means extending the green light of the current phase by one unit duration; "green light early termination" means stopping the green light of the current phase and activating the green light for the next phase according to the order of the background scheme; "phase skipping" means stopping the green light of the current phase and skipping the next phase or several phases in the background scheme, and then activating the green light for any other phase.

[0035] Step S206: Based on the multiple weighted average vehicle delays, a target signal control action is determined from the multiple signal control actions, and the target signal control action is sent to the traffic signal controller. The traffic signal controller is located in the target area and is used to output traffic signals. The multiple weighted average vehicle delays correspond one-to-one with the multiple signal control actions.

[0036] The above scheme acquires real-time traffic flow information within the target area and dynamically calculates the comprehensive control incentive value for the current traffic phase based on this information. Then, a large-scale signal control effect prediction model is used to predict the weighted average vehicle delay after implementing different signal control actions, thereby evaluating the effect of each action. Finally, based on the predicted weighted average vehicle delay, the target signal control action that best reduces delay is selected. This decision-making process ensures optimal traffic flow efficiency. After the action is selected, the instruction is transmitted to the traffic signal controller located within the target area to adjust the traffic signal to implement the optimal control strategy. By adopting the above scheme, this application achieves intelligent, efficient, and flexible urban traffic signal control by integrating real-time traffic data, defining comprehensive incentive values, and utilizing large-scale model prediction decisions. It can respond to traffic changes in real time, significantly improving traffic efficiency and safety, and is particularly suitable for complex traffic environments with multiple strategies coexisting. Furthermore, it solves the problem in related technologies where, in complex traffic scenarios, traditional inductive control methods struggle to effectively integrate the needs of multiple strategies such as pedestrian, public transport priority, and rapid response of emergency vehicles, resulting in poor control effects and a lack of transparency in the decision-making process.

[0037] Optionally, real-time collection of traffic flow information in the target area includes: real-time collection of vehicle arrival information through sensing devices deployed in the target area, wherein the vehicle arrival information includes: the average vehicle arrival rate of multiple lanes in the current phase, the arrival time of the target vehicle in each lane in the current phase, and the saturation of the time window [t-Δt,t]; the target vehicle is the last arriving vehicle in the current phase, and the saturation is calculated based on traffic flow; detecting pedestrian crossing data in the current green light phase, wherein the pedestrian crossing data is used to indicate whether there are pedestrians who have not completed crossing the street in the current green light phase; detecting whether there is a bus priority request in the current phase and obtaining bus priority request data; detecting whether there is an emergency vehicle passage request in the current phase and obtaining emergency passage request data, wherein the traffic flow information includes: the vehicle arrival information, the pedestrian crossing data, the bus priority request data, and the emergency passage request data.

[0038] Through advanced sensing devices deployed in the target area, the system continuously captures vehicle dynamics, specifically including the average arrival rate of each lane within the phase, the timestamp of the last arriving vehicle, and saturation indicators based on recent traffic flow. At the same time, the system monitors whether there are pedestrians who have not completed crossing the street during the green light period, and detects whether there are requests for bus priority and emergency vehicle passage. These data together constitute traffic flow information.

[0039] Specifically, the process of collecting traffic flow information includes:

[0040] 1. Collection of social vehicle data:

[0041] By utilizing sensing devices (such as video detectors, radar, and radar-video integrated units) deployed at each approach lane of the intersection, real-time vehicle arrival information is collected, including:

[0042] (1) Vehicle arrival rate Within the time window [t-Δt,t], the average vehicle arrival rate for each lane at phase i is calculated.

[0043] (2) Arrival time of the last detected arriving vehicle in each lane of phase i

[0044] (3) The current green light has lasted for a period of time.

[0045] (4) Saturation Within the [t-Δt,t] time window, the saturation is calculated from the traffic flow indicators (e.g., queue length, flow rate) of the lanes included in that phase; Δt can be the duration of one cycle or a fixed duration such as 120 seconds.

[0046] 2. Pedestrian crossing data detection: If there is a pedestrian crossing phase during the current green light period, use video to identify whether any pedestrians have not completed crossing (are still crossing the zebra crossing); if so, use the indicator function δ ped (i) = 1 record, otherwise δ ped (i) = 0.

[0047] 3. Bus inspection:

[0048] Buses are identified by traffic signals via onboard GPS or RFID tags. If a bus enters the detection zone (e.g., within 500m of the intersection) in a lane included in phase i and the current light for phase i is not green, a bus priority request is triggered, and the expected arrival time of the bus is calculated based on its location.

[0049] Among all the detected buses belonging to phase i that are expected to arrive, the one with the earliest expected arrival time is selected, that is, the minimum expected arrival time among all bus priority requests, denoted as T. bus (i); and with the indicator function δ bus (i) = 1 indicates a bus priority request; if there is no bus priority request, then δ bus (i) = 0.

[0050] 4. Emergency vehicle inspection:

[0051] Similarly, when an emergency vehicle (such as a fire truck, ambulance, or police car) needs to pass through an intersection, the vehicle sends an emergency vehicle passage request to the traffic signal, denoted by the indicator function δ. emg (i) = 1, and record the expected arrival time T of the earliest emergency vehicle. emg (i).

[0052] This embodiment achieves intelligent and personalized adjustment of traffic signal control through real-time and comprehensive traffic data collection and analysis, effectively improving intersection traffic efficiency, ensuring the right-of-way of special traffic entities, enhancing pedestrian safety, and ultimately promoting the optimization and upgrading of the urban traffic system.

[0053] Optionally, calculating the comprehensive control incentive value for the current phase based on the traffic flow information includes: weighting and summing the saturation, average vehicle arrival rate, and window duration according to the traffic flow weight of the saturation, the vehicle arrival rate weight of the average vehicle arrival rate, and the window penalty weight of the window duration, respectively, to obtain a first incentive value, wherein the window duration is the time difference between the current time and the arrival time, and the window penalty weight is a negative value; calculating a second incentive value based on the pedestrian crossing data, the pedestrian incentive gain corresponding to the pedestrian crossing data, the bus priority request data, the bus incentive gain corresponding to the bus priority request data, the emergency passage request data, and the emergency vehicle incentive gain corresponding to the emergency passage request data; and calculating the comprehensive control incentive value for the current phase i based on the first incentive value and the second incentive value.

[0054] To integrate the above multiple strategies, an Integrated Control Incentive (ICI) is defined to measure whether the current phase is "worth" extending the green light. For a given phase, the higher its saturation, the greater the traffic flow, and the shorter the time since the last vehicle passed, the higher its ICI. The ICI also increases if there are pedestrians who haven't finished crossing, or if there are upcoming bus priority requests or emergency vehicle requests. The formula for calculating the ICI is:

[0055]

[0056] Wherein, ω1, ω2, and -ω3 are the traffic flow weight, vehicle arrival rate weight, and empty window penalty weight, respectively;

[0057] β ped For pedestrian excitation gain, β bus For the bus excitation gain, β emg For emergency vehicle excitation gain;

[0058] λ bus λ is used to control the bus excitation decay rate. emg Used to control the excitation decay rate of emergency vehicles.

[0059] Of the three types of hyperparameters (ω, β, θ) mentioned above, only the θ-type hyperparameters are nonlinear. Therefore, λ bus and λ emg The values ​​are manually specified, with suggested values ​​of 0.2 and 0.05 respectively (this setting allows the incentive value for buses to decay more quickly as arrival time increases, avoiding wasting green lights for buses that are too far away; at the same time, the incentive value for emergency vehicles will not decay too quickly, ensuring that emergency vehicles that are far from the intersection can also get green lights as soon as possible).

[0060] Both ω and β are linear parameters and do not need to be manually specified. The above formula can be used as a pre-processor network to participate in the next step of training the larger model, training the parameters of ω and β through backpropagation.

[0061] By calculating the comprehensive control excitation value defined by the formula through pre-network training, the dimensionality of multi-policy sensing control input can be effectively reduced, and an interpretable input can be provided for large language models.

[0062] Optionally, before predicting the signal control effect using the large-scale signal control effect prediction model based on the comprehensive control incentive value, the method further includes: collecting initial training data, wherein the initial training data includes: traffic flow data of multiple phases in the target area, comprehensive control incentive value, signal control scheme data for the current period, and initial signal control actions; processing the initial training data using the large-scale initial signal control effect prediction model to obtain the initial weighted average vehicle delay corresponding to the initial signal control action; calculating the training loss based on the initial weighted average vehicle delay and the actual weighted average vehicle delay corresponding to the initial signal control action, and adjusting the model parameters of the large-scale initial signal control effect prediction model based on the training loss to obtain the large-scale signal control effect prediction model.

[0063] One week before the formal implementation of the large-scale model for predicting traffic signal control effectiveness as described in this application, the intersection can be controlled using rule-based traditional sensor control, and data can be collected as the initial training dataset for the large-scale model. On this training dataset middle:

[0064] The input consists of: traffic detection data, integrated control excitation values, channelization data, current cycle signal control scheme data, background scheme data, and sensor control data for each phase of the intersection, which are combined to form context data; and signal control action 'a' is provided to the large model.

[0065] The output is the predicted average vehicle delay Z at the intersection. This average vehicle delay is a weighted average of the delays of private vehicles, buses, and emergency vehicles.

[0066] Z = γ soc Z soc +γ bus Z bus +γ emg Z emg , where γ soc γ bus γ emg Delay weights are assigned to private vehicles, buses, and emergency vehicles, respectively; these weights are specified based on human experience to suit the needs of specific traffic application scenarios. soc Z bus Z emgThe average delays were for private vehicles, buses, and emergency vehicles, respectively.

[0067] When implementing inductive control, the model makes signal control action decisions based on traffic data. The control action space is a∈A={α0=green light unchanged, a1=green light extended, a2=green light ended early, a3=phase switching}, where “green light unchanged” means maintaining the green light duration of the current phase in the background plan without increasing or decreasing it; “green light extended” means extending the green light duration of the current phase by one unit; “green light ended early” means stopping the green light of the current phase and turning on the green light for the next phase according to the order of the background plan; “phase switching” means stopping the green light of the current phase and skipping the next one or several phases in the background plan, and then turning on the green light for any other phase.

[0068] After each action decision by the model, vehicle delay data for the next cycle after the action is performed is collected, and the weighted average vehicle delay Z is calculated, thus generating an input-output data point, which is then added to the system.

[0069] Obtain a large model training dataset Subsequently, a large-scale traffic control performance prediction model was trained using fine-tuning. This model can predict the weighted average vehicle delay at intersections based on context data. The loss function for fine-tuning the large-scale model is defined by a logarithmic function, as shown in the following equation:

[0070]

[0071] After training a large-scale model for predicting traffic signal control performance at the intersection and implementing the sensor control in this embodiment according to subsequent steps, new training data of {context; a→Z} is continuously collected and added to the dataset used for training the large-scale model. At the same time, the large model is fine-tuned using scheduled tasks (e.g., weekly or monthly) to train a larger model with higher accuracy.

[0072] In an exemplary embodiment, determining a target traffic control action from a plurality of traffic control actions based on a plurality of weighted average vehicle delays includes: normalizing the plurality of weighted average vehicle delays to obtain a plurality of normalized weighted average vehicle delays; calculating an expected effect score for each traffic control action based on the plurality of normalized weighted average vehicle delays using a sigmoid function; calculating a utility score for each of the plurality of traffic control actions based on the plurality of expected effect scores using a softmax function, wherein the plurality of expected effect scores correspond one-to-one with the plurality of traffic control actions; and determining the target traffic control action from the plurality of traffic control actions based on the plurality of utility scores, wherein the plurality of utility scores correspond one-to-one with the plurality of traffic control actions.

[0073] Using the aforementioned large-scale signal control effect prediction model, we can obtain the weighted average vehicle delay under different signal control actions in the current context. The weighted average vehicle delay Z is normalized using the following formula and converted into a score LLM_Score in the form of a sigmoid function to evaluate the expected effect of the corresponding signal control actions.

[0074]

[0075] In an exemplary embodiment, determining the target signal control action from the plurality of signal control actions based on a plurality of utility scores includes: sorting the plurality of signal control actions according to the plurality of utility scores to obtain a plurality of sorted signal control actions; determining whether the utility scores of a plurality of first signal control actions among the sorted plurality of signal control actions are greater than corresponding confidence thresholds, wherein the plurality of first signal control actions correspond one-to-one with a plurality of confidence thresholds, and the plurality of first signal control actions include: phase switching, early green light termination, and green light extension; if it is determined that the utility score of at least one second signal control action among the plurality of first signal control actions is greater than the corresponding confidence threshold, determining the third signal control action among the at least one second signal control action as the target signal control action, wherein the utility score of the third signal control action is ranked first among the at least one second signal control action; if it is determined that the utility scores of all the plurality of first signal control actions are less than or equal to the corresponding confidence thresholds, determining the fourth signal control action among the plurality of signal control actions as the target signal control action, wherein the fourth signal control action is the green light remaining unchanged.

[0076] Because the detector data is updated every second, the traffic flow state perceived by the model also changes in real time. Correspondingly, the integrated control incentive values ​​and delay prediction values ​​in the above steps are also constantly changing. This may cause the "optimal action" based on the expected effect to jump around repeatedly in a short period of time, resulting in high volatility. Therefore, this embodiment uses a softmax approach to calculate the utility of each action, denoted as C(a), and uses a threshold to determine whether to take an action. In other words, an action (except for maintaining a green light) will only be taken when its utility is significantly higher than other actions.

[0077]

[0078] Among the four optional actions, they are arranged in descending order of their radicalness in changing the current plan: "phase switching, early green light termination, extended green light, and unchanged green light". Decision confidence thresholds are set for the first three actions: τ3, τ2, and τ1. These threshold values ​​can be manually specified and adjusted to control the model's sensitivity, for example, 0.7.

[0079] C(ak )>τ k , for k in 1,2,3.

[0080] If the above conditions are met, then take the corresponding action a. k If the above conditions are not met, the default action is to take action a0, which means keeping the green light unchanged in the current phase.

[0081] Optionally, before sending the target signal control action to the traffic signal, the method further includes: performing an execution verification on the target signal control action to determine whether the green light duration of the traffic signal after executing the target signal control action is within a preset range; if it is determined that the target signal control action fails the execution verification, updating the target signal control action to a fourth signal control action and sending the updated target signal control action to the traffic signal, wherein the fourth signal control action is that the green light remains unchanged; if it is determined that the target signal control action passes the execution verification, sending the target signal control action to the traffic signal.

[0082] Based on the final signal control action decision, the green light duration of the current phase is checked: if the green light duration of the current phase does not meet the requirements of maximum or minimum green light due to green light extension / early termination / forced handover, the signal control action is rejected, action a0 is executed, and the green light duration of the current phase remains unchanged.

[0083] If the signal control action decision is a m If the verification is successful, the data is sent to the signal controller; simultaneously, the corresponding large-scale signal control performance prediction LLM_Score(a) is generated. m ) and reasoning process I t (a m The output explains the reasoning process behind the decision to the customer.

[0084] In an optional embodiment, this application combines Figure 3 The implementation process of the above-mentioned traffic signal dynamic sensing control method based on a large language model will be further explained, such as... Figure 3 As shown, the method includes the following steps:

[0085] 1. Traffic data collection;

[0086] Traffic data collection specifically requires the collection of: social vehicle data (including vehicle arrival rate, vehicle arrival time, saturation, etc.), pedestrian crossing data (used to indicate whether there are pedestrians who have not completed crossing the street), bus detection data (to determine whether there are bus priority requests), and emergency vehicle detection data (to determine whether there are emergency vehicle passage requests).

[0087] 2. Establish a front-end network – a multi-strategy integrated control of incentive values;

[0088] The comprehensive control incentive value is calculated to measure whether the current phase is worth extending the green light. For a phase, the higher its saturation, the greater the traffic flow, and the shorter the time since the last vehicle passed, the greater its comprehensive control incentive value. If there are pedestrians who have not completed crossing the street, or if there are upcoming bus priority requests or emergency vehicle requests, the comprehensive control incentive value will also increase.

[0089] 3. Training of a large-scale model for predicting the effectiveness of credit control;

[0090] 4. Calculate the decision confidence level;

[0091] Using the aforementioned large-scale signal control effect prediction model, we can obtain the weighted average vehicle delay (ATR) under different signal control actions in the current context. The ATR Z is normalized and converted into an expected effect score using a sigmoid function to evaluate the expected effect of the corresponding signal control actions. The calculation results are as follows: Figure 4 As shown.

[0092] On the other hand, because the detector data is updated every second, the traffic flow state perceived by the model also changes in real time. Correspondingly, the integrated control incentive value and delay prediction value in the above steps are also constantly changing, which may cause the "optimal action" based on the expected effect to jump repeatedly in a short period of time, resulting in high volatility. Therefore, a softmax approach is used to calculate the utility of each signal control action, and a threshold (equivalent to the confidence threshold mentioned above) is used to decide whether to take an action. In other words, an action (except for keeping the green light unchanged) will only be taken when its utility is significantly higher than other actions.

[0093] 5. Credit control decision verification and output;

[0094] 6. Repeat steps 2-5.

[0095] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods according to the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods of the various embodiments of this application.

[0096] Figure 5 This is a structural block diagram of a traffic signal dynamic sensing control device based on a large language model, according to an embodiment of this application; as shown... Figure 5As shown, the device includes:

[0097] The calculation module 52 is used to collect traffic flow information of the target area in real time and calculate the comprehensive control excitation value of the current phase based on the traffic flow information. The traffic flow information is used to indicate the traffic information of multiple objects passing through the target area, and the current phase is used to indicate the current stage of traffic signal control.

[0098] Prediction module 54 is used to predict the signal control effect based on the comprehensive control excitation value through the large model of signal control effect prediction, and obtain the weighted average vehicle delay of multiple signal control actions, wherein the multiple signal control actions include: phase switching, early green light termination, green light extension, and unchanged green light.

[0099] The decision module 56 is used to determine a target traffic signal action from the multiple traffic signal actions based on the multiple weighted average vehicle delays, and to send the target traffic signal action to the traffic signal controller. The traffic signal controller is located in the target area and is used to output traffic signals. The multiple weighted average vehicle delays correspond one-to-one with the multiple traffic signal actions.

[0100] The aforementioned device acquires real-time traffic flow information within the target area and dynamically calculates the comprehensive control incentive value for the current traffic phase based on this information. Then, a large-scale signal control effect prediction model is used to predict the weighted average vehicle delay after implementing different signal control actions, thereby evaluating the effectiveness of each action. Finally, based on the predicted weighted average vehicle delay, the target signal control action that best reduces delay is selected. This decision-making process ensures optimal traffic flow efficiency. After selecting the action, the instruction is transmitted to the traffic signal controller located within the target area to adjust the traffic signal to implement the optimal control strategy. Using the above scheme, this application achieves intelligent, efficient, and flexible urban traffic signal control by integrating real-time traffic data, defining comprehensive incentive values, and utilizing large-scale model prediction decisions. It can respond to traffic changes in real time, significantly improving traffic efficiency and safety, and is particularly suitable for complex traffic environments with multiple strategies coexisting. Furthermore, it solves the problem in related technologies where, in complex traffic scenarios, traditional inductive control methods struggle to effectively integrate multiple strategy requirements such as pedestrian, public transport priority, and rapid emergency vehicle response, leading to poor control effects and a lack of transparency in the decision-making process.

[0101] In an exemplary embodiment, the aforementioned calculation module 52 is further configured to collect vehicle arrival information in real time via sensing devices deployed in the target area, wherein the vehicle arrival information includes: the average vehicle arrival rate of multiple lanes in the current phase, the arrival time of the target vehicle in each lane in the current phase, and the saturation of the time window [t-Δt,t]; the target vehicle is the last arriving vehicle in the current phase, and the saturation is calculated based on traffic flow; detect pedestrian crossing data in the current green light phase, wherein the pedestrian crossing data is used to indicate whether there are pedestrians who have not completed crossing the street in the current green light phase; detect whether there is a bus priority request in the current phase and obtain bus priority request data; detect whether there is an emergency vehicle passage request in the current phase and obtain emergency passage request data, wherein the traffic flow information includes: the vehicle arrival information, the pedestrian crossing data, the bus priority request data, and the emergency passage request data.

[0102] In an exemplary embodiment, the calculation module 52 is further configured to perform a weighted summation of the saturation, the average vehicle arrival rate, and the window duration based on the traffic flow weight of the saturation, the vehicle arrival rate weight of the average vehicle arrival rate, and the window penalty weight of the window duration, respectively, to obtain a first incentive value, wherein the window duration is the time difference between the current time and the arrival time, and the window penalty weight is a negative value; calculate a second incentive value based on the pedestrian crossing data, the pedestrian incentive gain corresponding to the pedestrian crossing data, the bus priority request data, the bus incentive gain corresponding to the bus priority request data, the emergency passage request data, and the emergency vehicle incentive gain corresponding to the emergency passage request data; and calculate the comprehensive control incentive value of the current phase i based on the first incentive value and the second incentive value.

[0103] In an exemplary embodiment, the prediction module 54 is further configured to collect initial training data, wherein the initial training data includes: traffic flow data of multiple phases of the target area, comprehensive control incentive values, signal control scheme data of the current period, and initial signal control actions; process the initial training data through an initial signal control effect prediction model to obtain the initial weighted average vehicle delay corresponding to the initial signal control action; calculate the training loss based on the initial weighted average vehicle delay and the actual weighted average vehicle delay corresponding to the initial signal control action, and adjust the model parameters of the initial signal control effect prediction model based on the training loss to obtain the signal control effect prediction model.

[0104] In an exemplary embodiment, the decision module 56 is further configured to normalize the multiple weighted average vehicle delays to obtain multiple normalized weighted average vehicle delays; calculate the expected effect score of each traffic control action based on the multiple normalized weighted average vehicle delays using the sigmoid function; calculate the utility score of the multiple traffic control actions based on the multiple expected effect scores using the softmax function, wherein the multiple expected effect scores correspond one-to-one with the multiple traffic control actions; and determine the target traffic control action from the multiple traffic control actions based on the multiple utility scores, wherein the multiple utility scores correspond one-to-one with the multiple traffic control actions.

[0105] In an exemplary embodiment, the decision module 56 is further configured to sort the plurality of signal control actions according to the plurality of utility scores to obtain a plurality of sorted signal control actions; determine whether the utility scores of a plurality of first signal control actions among the sorted plurality of signal control actions are greater than the corresponding confidence thresholds, wherein the plurality of first signal control actions correspond one-to-one with the plurality of confidence thresholds, and the plurality of first signal control actions include: the phase switching, the early green light termination, and the green light extension; if it is determined that the utility score of at least one second signal control action among the plurality of first signal control actions is greater than the corresponding confidence threshold, the third signal control action among the at least one second signal control action is determined as the target signal control action, wherein the utility score of the third signal control action is ranked first among the at least one second signal control action; if it is determined that the utility scores of all the plurality of first signal control actions are less than or equal to the corresponding confidence thresholds, the fourth signal control action among the plurality of signal control actions is determined as the target signal control action, wherein the fourth signal control action is the green light remaining unchanged.

[0106] In an exemplary embodiment, the decision module 56 is further configured to perform an execution verification on the target signal control action to determine whether the green light duration of the signal after executing the target signal control action is within a preset range; if it is determined that the target signal control action fails the execution verification, the target signal control action is updated to a fourth signal control action, and the updated target signal control action is sent to the signal, wherein the fourth signal control action is that the green light remains unchanged; if it is determined that the target signal control action passes the execution verification, the target signal control action is sent to the signal.

[0107] Embodiments of this application also provide a storage medium including a stored program, wherein the program executes any of the methods described above when it is run.

[0108] Optionally, in this embodiment, the storage medium may be configured to store program code for performing the following steps:

[0109] S1, real-time acquisition of traffic flow information in the target area, and calculation of the comprehensive control excitation value of the current phase based on the traffic flow information, wherein the traffic flow information is used to indicate the traffic information of multiple objects passing through the target area, and the current phase is used to indicate the current stage of traffic signal control;

[0110] S2, the signal control effect prediction model predicts the signal control effect based on the comprehensive control excitation value, and obtains the weighted average vehicle delay of multiple signal control actions, wherein the multiple signal control actions include: phase switching, early green light termination, green light extension, and unchanged green light.

[0111] S3, based on the multiple weighted vehicle delays, a target signal control action is determined from the multiple signal control actions, and the target signal control action is sent to the traffic signal controller. The traffic signal controller is located in the target area and is used to output traffic signals. The multiple weighted vehicle delays correspond one-to-one with the multiple signal control actions.

[0112] In one exemplary embodiment, the aforementioned computer-readable storage medium may include, but is not limited to, various media capable of storing computer programs, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard disk, magnetic disk, or optical disk.

[0113] Specific examples in this embodiment can be found in the examples described in the above embodiments and exemplary implementations, and will not be repeated here.

[0114] Embodiments of this application also provide an electronic device including a memory and a processor, wherein the memory stores a computer program and the processor is configured to run the computer program to perform the steps in any of the above method embodiments.

[0115] Optionally, in this embodiment, the processor can be configured to perform the following steps via a computer program:

[0116] S1, real-time acquisition of traffic flow information in the target area, and calculation of the comprehensive control excitation value of the current phase based on the traffic flow information, wherein the traffic flow information is used to indicate the traffic information of multiple objects passing through the target area, and the current phase is used to indicate the current stage of traffic signal control;

[0117] S2, the signal control effect prediction model predicts the signal control effect based on the comprehensive control excitation value, and obtains the weighted average vehicle delay of multiple signal control actions, wherein the multiple signal control actions include: phase switching, early green light termination, green light extension, and unchanged green light.

[0118] S3, based on the multiple weighted vehicle delays, a target signal control action is determined from the multiple signal control actions, and the target signal control action is sent to the traffic signal controller. The traffic signal controller is located in the target area and is used to output traffic signals. The multiple weighted vehicle delays correspond one-to-one with the multiple signal control actions.

[0119] Optionally, the electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor and the input / output device is connected to the processor.

[0120] Optionally, specific examples in this embodiment can refer to the examples described in the above embodiments and optional implementations, and will not be repeated here.

[0121] Obviously, those skilled in the art should understand that the modules or steps of this application described above can be implemented using general-purpose computing systems. They can be centralized on a single computing system or distributed across a network of multiple computing systems. Optionally, they can be implemented using program code executable by a computing system, thereby storing them in a storage system for execution by the computing system. In some cases, the steps shown or described can be performed in a different order than those presented here, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, this application is not limited to any particular combination of hardware and software.

[0122] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the principles of this application should be included within the protection scope of this application.

Claims

1. A traffic signal dynamic sensing control method based on a large language model, characterized in that, include: Real-time acquisition of traffic flow information in the target area, and calculation of the comprehensive control excitation value of the current phase based on the traffic flow information, wherein the traffic flow information is used to indicate the traffic information of multiple objects passing through the target area, and the current phase is used to indicate the current stage of traffic signal control; The signal control effect prediction model predicts the signal control effect based on the comprehensive control excitation value, and obtains the weighted average vehicle delay of multiple signal control actions, including: phase switching, early green light termination, extended green light, and unchanged green light. Based on the multiple weighted average vehicle delays, a target traffic signal action is determined from the multiple traffic signal actions, and the target traffic signal action is sent to the traffic signal controller. The traffic signal controller is located in the target area and is used to output traffic signals. The multiple weighted average vehicle delays correspond one-to-one with the multiple traffic signal actions.

2. The method according to claim 1, characterized in that, Real-time collection of traffic flow information in the target area, including: Vehicle arrival information is collected in real time by sensing devices deployed in the target area. The vehicle arrival information includes: the average vehicle arrival rate of multiple lanes in the current phase, the arrival time of the target vehicle in each lane in the current phase, and the saturation of the time window [t-Δt,t]. The target vehicle is the last vehicle to arrive in the current phase, and the saturation is calculated based on traffic flow. Detect pedestrian crossing data during the current green light phase, wherein the pedestrian crossing data is used to indicate whether there are any pedestrians who have not completed crossing the street during the current green light phase; Detect whether there is a bus priority request in the current phase, and obtain bus priority request data; Detect whether there is an emergency vehicle passage request in the current phase to obtain emergency passage request data, wherein the traffic flow information includes: vehicle arrival information, pedestrian crossing data, bus priority request data, and emergency passage request data.

3. The method according to claim 2, characterized in that, The comprehensive control incentive value for the current phase is calculated based on the traffic flow information, including: The saturation, average vehicle arrival rate, and window duration are weighted and summed according to the traffic flow weight of the saturation, the vehicle arrival rate weight of the average vehicle arrival rate, and the window penalty weight of the window duration to obtain the first incentive value. The window duration is the time difference between the current time and the arrival time, and the window penalty weight is a negative value. The second incentive value is calculated based on the pedestrian crossing data, the pedestrian incentive gain corresponding to the pedestrian crossing data, the bus priority request data, the bus incentive gain corresponding to the bus priority request data, the emergency passage request data, and the emergency vehicle incentive gain corresponding to the emergency passage request data. The integrated control excitation value for the current phase is calculated based on the first excitation value and the second excitation value.

4. The method according to claim 1, characterized in that, Before predicting the signal control effect using the large-scale signal control effect prediction model based on the comprehensive control excitation value, the method further includes: Collect initial training data, which includes: traffic flow data of multiple phases of the target area, comprehensive control incentive values, signal control scheme data for the current period, and initial signal control actions; The initial training data is processed by the initial signal control effect prediction large model to obtain the initial weighted average vehicle delay corresponding to the initial signal control action; The training loss is calculated based on the initial weighted average vehicle delay and the actual weighted average vehicle delay corresponding to the initial signal control action, and the model parameters of the initial signal control effect prediction model are adjusted based on the training loss to obtain the signal control effect prediction model.

5. The method according to claim 1, characterized in that, Based on the weighted average vehicle delays, a target signal control action is determined from the multiple signal control actions, including: The weighted average vehicle delays are normalized to obtain multiple normalized weighted average vehicle delays; The expected effect score of each signal control action is calculated using the sigmoid function based on the multiple normalized weighted average vehicle delays. The utility scores of the multiple credit control actions are calculated using the softmax function based on the multiple expected effect scores, wherein each of the multiple expected effect scores corresponds one-to-one with the multiple credit control actions. The target credit control action is determined from the plurality of credit control actions based on the plurality of utility scores, wherein the plurality of utility scores correspond one-to-one with the plurality of credit control actions.

6. The method according to claim 5, characterized in that, The target credit control action is determined from the plurality of credit control actions based on the plurality of utility scores, including: The multiple credit control actions are sorted according to the multiple utility scores to obtain the sorted multiple credit control actions; Each of the sorted signal control actions is determined to have a utility score greater than the corresponding confidence threshold. The multiple first signal control actions correspond one-to-one with the multiple confidence thresholds. The multiple first signal control actions include: phase switching, early green light termination, and green light extension. If it is determined that at least one second information control action among the plurality of first information control actions has a utility score greater than the corresponding confidence threshold, then the third information control action among the at least one second information control action is determined as the target information control action, wherein the utility score of the third information control action is ranked first among the at least one second information control action; If the utility scores of the plurality of first signal control actions are all less than or equal to the corresponding confidence thresholds, the fourth signal control action among the plurality of signal control actions is determined as the target signal control action, wherein the fourth signal control action is that the green light remains unchanged.

7. The method according to claim 1, characterized in that, Before sending the target signal control action to the signal controller, the method further includes: The target signal control action is executed and verified to determine whether the green light duration of the signal is within a preset range after the target signal control action is executed. If it is determined that the target signal control action fails the execution verification, the target signal control action is updated to the fourth signal control action, and the updated target signal control action is sent to the signal controller, wherein the fourth signal control action is that the green light remains unchanged; If the target signal control action passes the execution verification, the target signal control action is sent to the signal controller.

8. A traffic signal dynamic sensing and control device based on a large language model, characterized in that, include: The calculation module is used to collect traffic flow information of the target area in real time and calculate the comprehensive control excitation value of the current phase based on the traffic flow information. The traffic flow information is used to indicate the traffic information of multiple objects passing through the target area, and the current phase is used to indicate the current stage of traffic signal control. The prediction module is used to predict the signal control effect based on the comprehensive control excitation value through the large model of signal control effect prediction, and obtain the weighted average vehicle delay of multiple signal control actions, wherein the multiple signal control actions include: phase switching, early green light termination, green light extension, and unchanged green light. The decision module is used to determine a target traffic signal action from the multiple traffic signal actions based on the multiple weighted average vehicle delays, and to send the target traffic signal action to the traffic signal controller. The traffic signal controller is located in the target area and is used to output traffic signals. The multiple weighted average vehicle delays correspond one-to-one with the multiple traffic signal actions.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored program, wherein the program, when executed, performs the method of any one of claims 1 to 7.

10. An electronic device comprising a memory and a processor, characterized in that, The memory stores a computer program, and the processor is configured to execute the method of any one of claims 1 to 7 through the computer program.

Images