Reinforcement learning traffic signal real-time optimization system and method based on security constraint
By constructing a traffic signal optimization system based on reinforcement learning with safety constraints, the problems of unsafety and instability in existing signal control technologies have been solved, enabling safe, stable, and efficient deployment in real traffic environments and improving intersection traffic efficiency.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-27
- Publication Date
- 2026-04-07
AI Technical Summary
Existing reinforcement learning traffic signal control systems lack hard constraints on traffic safety, their outputs are incompatible with traffic controllers, making it difficult to cope with abnormal situations, leading to system instability and safety risks.
A traffic signal optimization system based on safety constraints and reinforcement learning is constructed, including a data processing module, a reinforcement learning decision-making module, a safety constraint module, a policy verification module, and a control command execution module. Through safety constraint filtering and policy verification degradation mechanisms, the system ensures that the signal control strategy complies with traffic safety rules and operates stably under abnormal conditions.
It achieves safe, stable, and reliable signal control in real traffic environments, improves intersection traffic efficiency, ensures traffic safety and system robustness, and supports the expansion and deployment of various RL algorithms.
Smart Images
Figure CN121811643A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of intelligent traffic control, and in particular to a safety constraint based reinforcement learning traffic signal real-time optimization system and method. BACKGROUND
[0002] Traditional traffic signal control methods include timing control, induction control, adaptive control, etc., wherein the timing control cannot be adjusted according to real-time traffic fluctuations; the induction and adaptive methods are limited by the sensor deployment density and are insufficient in responding to sudden events. In recent years, reinforcement learning (RL) has gradually increased in the field of traffic, and by allowing the agent to interact with the traffic environment to learn the strategy, the signal phase, cycle and green time can be automatically adjusted according to the real-time traffic state. However, the existing RL based traffic signal control still has the following problems: 1. Lack of traffic safety hard constraint mechanism: reinforcement learning aims to maximize rewards, and may learn strategies that violate safety principles during training or online execution, such as not meeting the minimum green time, not guaranteeing the yellow light time, pedestrians not completing the crossing before switching, and not giving sufficient protection interval for high-risk conflict traffic flow; 2. Reinforcement learning strategy output does not match the engineering interface of the traffic controller: RL output is continuous action or discrete selection, and if there is no interface mechanism, it is easy to cause frequent jumps in instructions, resulting in the traffic controller being unable to execute; 3. Lack of safety degradation capability for abnormal situations (sensor failure, sudden congestion, emergency vehicle passage): existing RL methods are mostly based on ideal environments, and when data is abnormal or traffic state changes sharply, the control strategy may fail.
[0003] The above defects result in the inability of reinforcement learning to be safely and stably deployed in actual urban traffic control systems, and therefore, a real-time signal optimization device and method that combines reinforcement learning and safety constraint rules and has engineering deployability are needed. SUMMARY
[0004] The present application aims to provide a safety constraint based reinforcement learning traffic signal real-time optimization system and method, which aims to solve the problems in the prior art that the strategy does not have mandatory safety constraints, there is a risk of traffic safety; the reinforcement learning output is incompatible with the traffic signal controller instructions, making it difficult to deploy in engineering; there is a lack of safety degradation mechanism for sudden events and data anomalies; and the controller response is unstable, which may frequently switch phases to cause vehicle conflicts and delay rebounds.
[0005] The present application provides a safety constraint based reinforcement learning traffic signal real-time optimization system, comprising: The data processing module is connected with the reinforcement learning decision module, is used for acquiring multi-dimensional traffic flow data in real time, and processing the multi-dimensional traffic flow data to build state information, and transmitting the state information to the reinforcement learning decision module; The reinforcement learning decision module is connected with the data processing module, the safety constraint module and the online optimization module, is used for generating a candidate signal control strategy based on the state information, and sending the candidate signal control strategy to the safety constraint module; The safety constraint module is connected with the reinforcement learning decision module and the policy verification module, is used for performing forced safety filtering on the candidate signal control strategy based on preset traffic safety rules, and outputting a safe signal control strategy; The policy verification module is connected with the safety constraint module and the control instruction execution module, is used for performing feasibility verification on the safe signal control strategy, selecting whether to execute a degradation strategy according to a verification result, and obtaining a final signal control strategy; The control instruction execution module is connected with the policy verification module and the online optimization module, is used for receiving the final signal control strategy, and issuing the final signal control strategy to a traffic signal controller for traffic control; The online optimization module is connected with the reinforcement learning decision module and the control instruction execution module, is used for monitoring a traffic control effect in real time, generating a corresponding optimization strategy based on the traffic control effect, and feeding back the optimization strategy to the reinforcement learning decision module for adaptive optimization.
[0006] The embodiment of the application provides a safety constraint-based reinforcement learning traffic signal real-time optimization method, which comprises the following steps: The data processing module is connected with the reinforcement learning decision module, is used for acquiring multi-dimensional traffic flow data in real time, and processing the multi-dimensional traffic flow data to build state information, and transmitting the state information to the reinforcement learning decision module; The reinforcement learning decision module is connected with the data processing module, the safety constraint module and the online optimization module, is used for generating a candidate signal control strategy based on the state information, and sending the candidate signal control strategy to the safety constraint module; The safety constraint module is connected with the reinforcement learning decision module and the policy verification module, is used for performing forced safety filtering on the candidate signal control strategy based on preset traffic safety rules, and outputting a safe signal control strategy; The policy verification module is connected with the safety constraint module and the control instruction execution module, is used for performing feasibility verification on the safe signal control strategy, selecting whether to execute a degradation strategy according to a verification result, and obtaining a final signal control strategy; The control instruction execution module is connected with the policy verification module and the online optimization module, is used for receiving the final signal control strategy, and issuing the final signal control strategy to a traffic signal controller for traffic control; The traffic control effect is monitored in real time by the online optimization module, an optimization strategy is generated based on the traffic control effect, and the optimization strategy is fed back to the reinforcement learning decision module for adaptive optimization.
[0007] The embodiment of the present application also provides an electronic device, comprising a memory, a processor and a computer program stored in the memory and executable on the processor, and the computer program implements the steps of the safety constraint based reinforcement learning traffic signal real-time optimization method when executed by the processor.
[0008] The embodiment of the present application also provides a computer readable storage medium, and the computer readable storage medium stores an information transmission implementation program, and the program implements the steps of the safety constraint based reinforcement learning traffic signal real-time optimization method when executed by a processor.
[0009] The embodiment of the present application can have the following beneficial effects: the embodiment of the present application constructs a closed-loop control system of "reinforcement learning decision-making - safety constraint filtering - strategy verification degradation - online feedback optimization", fundamentally solves the three core pain points of unsafe strategy, engineering difficulty landing and unstable system in the existing intelligent traffic control, not only ensures that the signal control strategy absolutely meets the traffic safety specification by using the safety constraint engine, but also guarantees the robustness of the system in abnormal conditions through the verification degradation mechanism, finally significantly improves the intersection passing efficiency, and provides complete technical support for realizing safe, reliable and efficient deployment of reinforcement learning in real traffic environment. BRIEF DESCRIPTION OF DRAWINGS
[0010] In order to more clearly illustrate the technical solutions in the one or more embodiments of the present application or the prior art, the drawings needed to be used in the embodiment or prior art description will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments described in the present application, and other drawings can be obtained by those skilled in the art without creative labor.
[0011] Figure 1 is a safety constraint based reinforcement learning traffic signal real-time optimization system schematic diagram of the embodiment of the present application; Figure 2 is a safety constraint based reinforcement learning traffic signal real-time optimization method flow chart of the embodiment of the present application. DETAILED DESCRIPTION
[0012] In order to make the person skilled in the art better understand the technical solutions in one or more embodiments of the present specification, the technical solutions in one or more embodiments of the present specification will be clearly and completely described below in conjunction with the drawings in one or more embodiments of the present specification. Obviously, the described embodiments are only a part of the embodiments of the present specification, not all. Based on one or more embodiments of the present specification, all other embodiments obtained by those skilled in the art without creative labor should belong to the protection scope of the present document.
[0013] System embodiment According to the embodiments of the present application, a safety constraint based reinforcement learning traffic signal real-time optimization system is provided, Figure 1 is a safety constraint based reinforcement learning traffic signal real-time optimization system schematic diagram of the embodiments of the present application, as Figure 1 shown, the safety constraint based reinforcement learning traffic signal real-time optimization system according to the embodiments of the present application specifically comprises: The data processing module 10 is connected with the reinforcement learning decision module, used for acquiring multi-dimensional traffic flow data in real time, processing the multi-dimensional traffic flow data, constructing state information, and transmitting the state information to the reinforcement learning decision module. Among them, the multi-dimensional traffic flow data includes traffic flow data from video detector, geomagnetic detector, millimeter wave radar or V2X vehicle networking terminal.
[0014] The reinforcement learning decision module 11 is connected with the data processing module, safety constraint module and online optimization module, used for generating candidate signal control strategy based on the state information, and sending the candidate signal control strategy to the safety constraint module, specifically used for: Generating candidate signal control strategy based on the state information by using action space constrained deep reinforcement learning model; Among them, the deep reinforcement learning model includes deep Q network DQN, proximal policy optimization PPO, deep deterministic policy gradient DDPG, asynchronous advantage action evaluation A3C or actor critic Actor-Critic algorithm; The reinforcement learning decision module further comprises an action smoothing unit, used for smoothing the output candidate signal control strategy, and inhibiting the violent jump of signal instructions in adjacent control periods.
[0015] The safety constraint module 12 is connected with the reinforcement learning decision module and the policy verification module, used for forcibly filtering the candidate signal control strategy based on the preset traffic safety rules, outputting the safe signal control strategy, and specifically used for: The candidate signal control strategy is forcedly filtered according to a preset traffic safety rule through a traffic conflict dynamic judgment model; The traffic safety rule comprises at least one of a minimum green light time constraint, a minimum yellow light time and all-red time constraint, a minimum crossing time guarantee constraint for pedestrians, a conflict traffic flow safety interval protection constraint, a maximum frequency constraint of phase switching per unit time, and an emergency vehicle priority passing trigger rule. The traffic conflict dynamic judgment model is based on intersection channelization information and real-time traffic flow data, constructs a conflict relationship graph among motor vehicles, non-motor vehicles and pedestrians, and dynamically calculates a conflict risk level, and adjusts a filtering rule of the candidate signal control strategy in real time according to the conflict risk level.
[0016] The strategy verification module 13 is connected with the safety constraint module and the control instruction execution module, and is configured to perform feasibility verification on the safe signal control strategy, select whether to execute a degradation strategy according to a verification result, and obtain a final signal control strategy, and is specifically configured to: enable the degradation strategy when the feasibility verification fails; wherein the degradation strategy comprises at least one of switching to a preset fixed timing scheme, switching to an inductive control scheme based on historical data, and enabling an emergency passing phase scheme for a sudden event; The strategy verification module further comprises a data anomaly detection unit configured to actively trigger the degradation strategy when a data source fails, communication is interrupted, or a traffic state deteriorates sharply.
[0017] The control instruction execution module 14 is connected with the strategy verification module and the online optimization module, configured to receive the final signal control strategy, and send the final signal control strategy to a traffic signal controller for traffic control; The online optimization module 15 is connected with the reinforcement learning decision module and the control instruction execution module, configured to monitor a traffic control effect in real time, generate a corresponding optimization strategy based on the traffic control effect, and feed back the optimization strategy to the reinforcement learning decision module for adaptive optimization.
[0018] The above technical solutions of the embodiments of the present application are described in detail in combination with the specific conditions of the safety constraint based reinforcement learning traffic signal real-time optimization system of the embodiments of the present application.
[0019] The embodiments of the present application provide a safety constraint based reinforcement learning traffic signal real-time optimization device, which can be applied to an intelligent networked traffic intersection control system, and mainly comprises: 1. Data acquisition module Traffic flow information is collected from video detectors, geomagnetic sensors, millimeter-wave radar, or V2X vehicle-to-everything (V2X) terminals, including traffic volume, queue length, speed, delays, and pedestrian status.
[0020] 2. State Fusion Module The original data is subjected to feature fusion, denoising, and missing data compensation to construct a state vector suitable for reinforcement learning.
[0021] 3. Enhance the learning decision-making module Candidate timing actions are generated based on an improved deep reinforcement learning model, including phase selection, green light duration adjustment, phase extension / truncation, etc.
[0022] 4. Safety Constraint Engine The RL output is subjected to "strong constraint filtering", including minimum / maximum green light time constraints, minimum yellow light time constraints, pedestrian protection phase constraints, motor vehicle conflict flow protection, non-motor vehicle mixed traffic protection, emergency vehicle priority rules, maximum switching frequency limit, and phase inconsistency command shielding mechanism.
[0023] 5. Strategy Verification and Degradation Module After RL+ security constraints, the generated strategy is validated for executability; if an anomaly is detected, it is automatically downgraded to a fixed timing scheme, a local adaptive scheme, or an emergency dedicated phase scheme.
[0024] 6. Signal control command issuing module The final executable instructions are sent to the controller, supporting interface interfaces for national or international standard protocols such as GB / T20999, NTCIP, and UTMC.
[0025] 7. Online Strategy Evaluation Module The execution results are statistically analyzed and fed back to the RL module for continuous learning and adaptive optimization.
[0026] Specifically, this invention proposes a real-time traffic signal optimization device with safety constraints based on reinforcement learning, comprising: a data acquisition module, a state fusion module, a reinforcement learning decision-making module, a safety constraint engine, a strategy verification and degradation module, a signal control command issuance module, and an online evaluation module. The reinforcement learning decision-making module further includes an action smoothing module before outputting candidate timing actions. This module smooths the action probability distribution based on a sliding window or exponential weighted average to reduce frequent phase switching and improve the executability and stability of signal control commands.
[0027] Preferably, the reinforcement learning model can be one of the following architectures: DQN, DDPG, PPO, A3C, or Actor-Critic.
[0028] The safety constraint engine is implemented based on a safety rule matrix and a traffic conflict judgment model. The safety constraint engine performs mandatory safety filtering on the timing actions output by the reinforcement learning decision module, including at least one of the following safety rules: 1. Minimum green light time constraint; 2. Minimum yellow light time constraint; 3. Pedestrian crossing protection and restraint; 4. Protection constraints for vehicle conflict flow; 5. Phase switching frequency limitation; 6. Emergency vehicle priority rules.
[0029] The timing scheme filtered by security constraints is then tested for executability by the policy verification and degradation module. If it does not meet the requirements, it is automatically downgraded to the preset security scheme.
[0030] Preferably, the safety constraint engine also includes a dynamic safety determination mechanism based on a traffic conflict discrimination model. This model establishes a conflict relationship matrix of motor vehicle flow, non-motor vehicle flow and pedestrian flow at the intersection, and calculates the risk level of the conflict flow in combination with real-time traffic conditions. Thus, when the reinforcement learning module outputs a phase switching command, it dynamically adjusts the set of permitted or prohibited actions according to the risk level.
[0031] The strategy verification and degradation module is equipped with a data anomaly detection unit. When anomalies such as missing data, sensor failure, V2X terminal disconnection, or sudden traffic incidents are detected, it automatically determines that the current reinforcement learning strategy is unusable and switches to the corresponding degradation strategy according to preset rules, including: 1. Operate according to fixed phase timing; 2. Run the adaptive scheme generated based on the historical daily average traffic model; 3. Activate the emergency passage priority phase scheme.
[0032] The degradation strategies in this embodiment of the invention include fixed timing, adaptive timing, or emergency-dedicated timing.
[0033] Example 1: Safety Constraint Signal Control Based on DQN The system collects data such as the number of vehicles, average speed, and queue length from the detectors. This data is then processed by the state fusion module to form a state vector S. The reinforcement learning module here uses the DQN model. The action space includes: A1: Maintain the current phase; A2: Enter the next phase; A3: Extend the current green light; A4: End the green light early.
[0034] The control execution process is as follows: 1. DQN outputs action A; 2. Safety constraint engine judgment: If the current green light is less than the minimum value, then A4 is prohibited; when there are pedestrians crossing the street, phase switching is prohibited; if the phase switching frequency is too high, then the phase is automatically maintained; if the RL output violates the conflict flow rule, then it is replaced with a safety action. 3. The strategy verification module determines whether the strategy is executable; otherwise, it downgrades to a fixed solution. 4. Send to the controller; 5. Feedback on delays and traffic volume will be used as learning rewards for the next round.
[0035] Example 2: V2X-based policy enhancement It supports data from vehicle-to-everything (V2X) terminals, such as road occupancy, vehicle speed, and arrival time, enabling advance prediction of fleet arrival times. It also reinforces learning-based decision-making to provide phase adjustments in advance and ensures the legality of changes through a safety constraint module.
[0036] Method Implementation Examples According to embodiments of the present invention, a reinforcement learning-based real-time traffic signal optimization method based on safety constraints is provided. Figure 2 This is a flowchart of a real-time traffic signal optimization method based on safety constraints using reinforcement learning, as described in an embodiment of the present invention. Figure 2 As shown, the real-time traffic signal optimization method based on safety constraints according to an embodiment of the present invention specifically includes: Step S201: The data processing module acquires multidimensional traffic flow data in real time, processes the multidimensional traffic flow data, constructs state information, and transmits the state information to the reinforcement learning decision module. Step S202: The reinforcement learning decision module generates a candidate signal control strategy based on the state information and sends the candidate signal control strategy to the security constraint module. Step S203: The safety constraint module performs forced safety filtering on the candidate signal control strategy based on preset traffic safety rules, and outputs a safe signal control strategy. Step S204: The feasibility of the secure signal control strategy is verified by the strategy verification module. Based on the verification result, it is selected whether to execute the degradation strategy to obtain the final signal control strategy. Step S205: Receive the final signal control strategy through the control command execution module, and send the final signal control strategy to the traffic signal controller for traffic management. Step S206: The online optimization module monitors the traffic control effect in real time, generates a corresponding optimization strategy based on the traffic control effect, and feeds the optimization strategy back to the reinforcement learning decision module for adaptive optimization.
[0037] The following describes in detail the above-mentioned technical solutions of the present invention with reference to the specific circumstances of the reinforcement learning-based traffic signal real-time optimization method based on safety constraints in the embodiments of the present invention.
[0038] This invention proposes a reinforcement learning-based real-time traffic signal optimization method with safety constraints, comprising: 1. Collect real-time traffic flow data; 2. The states are fused to form a state vector and then input into the reinforcement learning model; 3. Obtain candidate timing actions; 4. Forced filtering is performed through a security constraint engine; 5. Perform executability verification and necessary downgrades on the filtered strategies; 6. Issue the final instruction to the signal controller; 7. Collect feedback for continuous learning.
[0039] The embodiments of the present invention are method embodiments corresponding to the system embodiments described above. The specific operations of each step can be understood by referring to the description of the system embodiments, and will not be repeated here.
[0040] In summary, compared with the prior art, the embodiments of the present invention have the following beneficial effects: 1. It has enforceable security constraints, completely eliminating security vulnerabilities in RL policies; 2. The system stability is guaranteed within the closed-loop framework of strategy generation, constraint filtering, and degradation replacement; 3. Compatible with multiple RL algorithms, facilitating expansion and upgrades; 4. The device structure is easy to deploy and can be seamlessly integrated with existing controllers; 5. Applicable to intelligent connected transportation, can work in conjunction with V2X terminals, and has future expansion capabilities; 6. Improve traffic efficiency and reduce delays while ensuring the safety of pedestrians and vehicles.
[0041] Device Example 1 This invention provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, it performs the steps described in the method embodiment.
[0042] Device Example 2 This invention provides a computer-readable storage medium storing an information transmission implementation program, which, when executed by a processor, performs the steps described in the method embodiment.
[0043] The computer-readable storage media described in this embodiment include, but are not limited to, ROM, RAM, disk, or optical disk.
[0044] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.
Claims
1. A real-time traffic signal optimization system based on safety constraints and reinforcement learning, characterized in that, include: The data processing module, connected to the reinforcement learning decision module, is used to acquire multi-dimensional traffic flow data in real time, process the multi-dimensional traffic flow data, construct state information, and transmit the state information to the reinforcement learning decision module. The reinforcement learning decision module, connected to the data processing module, the security constraint module, and the online optimization module, is used to generate candidate signal control strategies based on the state information and send the candidate signal control strategies to the security constraint module. The safety constraint module, connected to the reinforcement learning decision module and the policy verification module, is used to perform mandatory safety filtering on the candidate signal control strategies based on preset traffic safety rules and output a safe signal control strategy. The strategy verification module, connected to the security constraint module and the control instruction execution module, is used to verify the feasibility of the secure signal control strategy, select whether to execute the degradation strategy based on the verification result, and obtain the final signal control strategy. The control command execution module is connected to the strategy verification module and the online optimization module, and is used to receive the final signal control strategy and send the final signal control strategy to the traffic signal controller for traffic control. The online optimization module, connected to the reinforcement learning decision module and the control command execution module, is used to monitor the traffic control effect in real time, generate corresponding optimization strategies based on the traffic control effect, and feed the optimization strategies back to the reinforcement learning decision module for adaptive optimization.
2. The system according to claim 1, characterized in that, The multidimensional traffic flow data includes traffic flow data from video detectors, geomagnetic detectors, millimeter-wave radar, or V2X vehicle-to-everything (V2X) terminals.
3. The system according to claim 1, characterized in that, The reinforcement learning decision module is specifically used for: Based on the state information, a candidate signal control strategy is generated using a deep reinforcement learning model constrained by the action space. The deep reinforcement learning model includes Deep Q-Network (DQN), Proximal Policy Optimization (PPO), Deep Deterministic Policy Gradient (DDPG), Asynchronous Advantage Action Evaluation (A3C), or the Actor-Critic algorithm.
4. The system according to claim 3, characterized in that, The reinforcement learning decision module also includes an action smoothing unit, which is used to smooth the output candidate signal control strategy and suppress drastic changes in signal commands within adjacent control cycles.
5. The system according to claim 1, characterized in that, The security constraint module is specifically used for: Based on preset traffic safety rules, the candidate signal control strategy is subjected to mandatory safety filtering through a traffic conflict dynamic discrimination model; The traffic safety rules include at least one of the following: minimum green light time constraint, minimum yellow light time and all-red light time constraint, minimum pedestrian crossing time guarantee constraint, conflict traffic flow safety interval protection constraint, maximum phase switching frequency constraint per unit time, and emergency vehicle priority passage triggering rules. The traffic conflict dynamic discrimination model constructs a conflict relationship diagram among motor vehicles, non-motor vehicles, and pedestrians based on intersection channelization information and real-time traffic flow data, dynamically calculates the conflict risk level, and adjusts the filtering rules of the candidate signal control strategy in real time according to the conflict risk level.
6. The system according to claim 1, characterized in that, The strategy verification module is specifically used for: When feasibility verification fails, a degradation strategy is activated; wherein, the degradation strategy includes at least one of the following: switching to a preset fixed timing scheme, switching to a sensor-based control scheme based on historical data, and activating an emergency passage phase scheme for sudden events.
7. The system according to claim 6, characterized in that, The policy verification module also includes a data anomaly detection unit, which is used to actively trigger the degradation policy when a data source failure, communication interruption, or a sharp deterioration in traffic conditions is detected.
8. A real-time traffic signal optimization method based on safety constraints using reinforcement learning, characterized in that, include: The data processing module acquires multidimensional traffic flow data in real time, processes the multidimensional traffic flow data, constructs state information, and transmits the state information to the reinforcement learning decision module. The reinforcement learning decision module generates candidate signal control strategies based on the state information and sends the candidate signal control strategies to the security constraint module. The safety constraint module performs mandatory safety filtering on the candidate signal control strategies based on preset traffic safety rules, and outputs a safe signal control strategy. The feasibility of the secure signal control strategy is verified by the strategy verification module. Based on the verification results, it is selected whether to execute the degradation strategy to obtain the final signal control strategy. The control instruction execution module receives the final signal control strategy and sends the final signal control strategy to the traffic signal controller for traffic management. The online optimization module monitors the traffic control effect in real time, generates corresponding optimization strategies based on the traffic control effect, and feeds the optimization strategies back to the reinforcement learning decision module for adaptive optimization.
9. An electronic device, characterized in that, include: The memory, the processor, and the computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the real-time traffic signal optimization method based on safety constraints as described in claim 8.
10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores an implementation program for information transmission, which, when executed by a processor, implements the steps of the real-time traffic signal optimization method based on security constraints as described in claim 8.