Intelligent network connection-based traffic safety device digital system

Abstract

This invention relates to the field of intelligent transportation technology, specifically a digital system for traffic safety facilities based on intelligent connected vehicles. The system acquires facility status and traffic environment data in real time through a data acquisition module; a data processing module constructs a heterogeneous, fusion-based, self-healing, and resilient network to achieve dynamic scheduling and fault self-healing of multi-protocol links, generating structured real-time data streams; a strategy output module constructs a traffic facility association map, uses a diffraction chain pre-simulation sandbox to perform risk path deduction, and generates a risk pre-control strategy set; a system optimization module parses and executes strategies through a compensatory control engine, and achieves closed-loop optimization by combining effect quantification evaluation. This invention, by deploying an adaptive control loop of detection input – data link – strategy generation – signal drive – parameter optimization, completes the logical leap of signal control from open-loop response to feedforward control, and from fixed parameters to dynamic generation, significantly enhancing the throughput of road network nodes and the utilization rate of spatiotemporal resources.

Description

Technical Field

[0001] This invention relates to the field of intelligent transportation technology, specifically to a digital system for traffic safety facilities based on intelligent connectivity. Background Technology

[0002] In the construction of smart cities and intelligent transportation systems, utilizing digital and network technologies for intelligent management and control of traffic safety facilities is a key technological direction for improving urban traffic efficiency and safety. Existing technologies generally employ decision-making models based on static rule matching and historical statistical data. This involves setting fixed risk thresholds and combining them with offline traffic big data analysis to monitor the operational status of traffic safety facilities and conduct emergency dispatch. This is currently the mainstream technological approach for achieving traffic risk management.

[0003] However, existing technical solutions have significant shortcomings in the timeliness and network coordination of decision-making, and the decision-making logic they rely on exhibits obvious static and open-loop characteristics. Existing systems mostly rely on preset fixed thresholds for judgment, lacking the ability to perceive the real-time status of facilities in complex electromagnetic and physical environments. They cannot effectively collect and integrate physical signal data streams reflecting the physical status of facilities, topological data streams describing network-related risks, and feedback data streams reflecting the effectiveness of emergency interventions, making it difficult to embed multi-source dynamic data into the real-time decision-making loop. Furthermore, the system only triggers alarms after events such as traffic congestion or accidents occur, failing to predict and warn of the cascading collapse risks that a single fault might trigger, thus amplifying safety risks. Secondly, control measures are isolated; emergency dispatch often targets local, independent events, lacking insight into the topological relationships of the entire transportation network. This can easily lead to unreasonable scheduling of the overall network and secondary congestion due to local optima, making it difficult to adapt to the ever-increasing and dynamically changing traffic demands.

[0004] To address this, a digital system for traffic safety facilities based on intelligent connectivity is proposed. Summary of the Invention

[0005] The purpose of this invention is to provide a digital system for traffic safety facilities based on intelligent connectivity, which aims to achieve closed-loop self-optimization of forward-looking prediction and management decisions for traffic network risks.

[0006] To achieve the above objectives, the present invention provides the following technical solution: A digital system for traffic safety facilities based on intelligent connected vehicles, comprising: The system collects real-time status data of traffic safety facilities and dynamic traffic environment data; it dynamically schedules link resources of communication protocols through the network control unit to construct and run a heterogeneous fusion self-healing elastic network, and performs optimal path selection and fault switching based on link status and data attributes to fuse and transmit the data, generating a structured real-time data stream; it constructs a traffic facility dependency grid and generates a facility association map based on the data stream, uses a diffraction chain pre-simulation sandbox to perform risk path deduction on the map, and generates a risk pre-control strategy set; it constructs and starts a compensatory control engine based on the strategy set, parses the strategies and issues execution control commands, and quantitatively evaluates the command execution effect in combination with newly collected data, feeding the evaluation results back to the strategy output module to achieve closed-loop optimization of system control.

[0007] Preferably, the status data of the traffic safety facility includes: power supply voltage, equipment temperature, enclosure airtightness, and pole vibration amplitude; the dynamic traffic environment data includes: traffic flow at a specific cross section, average driving speed, vehicle headway, and number of pedestrians crossing.

[0008] Preferably, the dynamic scheduling of link resources by the network control unit is implemented as follows: the network control unit first establishes and maintains a link resource pool containing all multi-mode communication terminals and their available communication protocols; then, the network control unit periodically sends probe commands to the terminals to obtain the real-time status of each link; finally, the network control unit selects a specified link from the link resource pool according to the upper-layer service requirements and sends a data transmission scheduling command to the corresponding terminal.

[0009] Preferably, the steps for constructing and operating the heterogeneous fusion self-healing resilient network are as follows: First, the multi-mode communication terminals on each traffic safety facility actively register with the network control unit and report all the communication protocol types they support; then, the network control unit instructs the terminals to establish connections with multiple communication networks and continuously perform status detection; finally, the network control unit maintains the continuous operation of the network based on the detection results and dynamically forms a fusion network topology with multiple paths coexisting and real-time status awareness.

[0010] Preferably, the implementation method of optimal path selection and fault switching based on link status and data attributes is as follows: When selecting the optimal path, the network control unit first parses the service attributes of the data to be transmitted to clarify its requirements, and combines the status parameters of each available link queried in real time, calculates the communication link that best matches the data requirements as the transmission path through a preset weighted matching algorithm; when performing fault switching, the network control unit continuously monitors the communication quality of the current transmission link. Once it falls below a preset threshold, it immediately triggers a rerouting mechanism, excludes the faulty link in the subsequent routing calculation, and re-executes the optimal path selection process, thereby seamlessly switching the data stream to a new optimal path.

[0011] Preferably, the generation of structured real-time data streams is achieved by: performing preprocessing such as timestamp alignment, coordinate system unification, and data format standardization on data collected from different sources and in different formats, and attaching the facility identification identifier to each data point, and finally encapsulating it into a unified format data packet for output to the upper layer.

[0012] Preferably, the detailed steps of the strategy output module in constructing the traffic facility dependency relationship grid and generating the facility association graph are as follows: First, the module parses the structured real-time data stream and, in conjunction with a preset facility basic information database, identifies and quantifies various dependencies between facilities through a relationship reasoning engine to construct the dependency relationship grid. These various dependencies include: functional master-slave relationships defined based on the device topology diagram, spatial adjacency relationships defined based on geographic information, and upstream and downstream influence relationships of traffic flow dynamically calculated based on real-time traffic flow data. Then, the module maps and transforms the dependency relationship grid, abstracting each traffic facility as a node and abstracting the dependencies between facilities as weighted directed edges connecting nodes, thereby generating the facility association graph as the input object for the diffraction chain pre-simulation sandbox.

[0013] Preferably, the diffraction chain simulation sandbox is configured to: receive the facility association graph and a specified initial event node set as input; the diffraction chain simulation sandbox incorporates a graph traversal inference algorithm, which calculates and generates a risk diffraction chain data structure on the facility association graph, starting from the initial event node set, based on the weights and directions of the edges in the graph. This data structure contains the identifiers of affected nodes, risk propagation paths, and time series of the degree of impact; the strategy output module then maps and combines the risk prevention and control strategy set from a preset intervention measure knowledge base based on the risk diffraction chain data structure.

[0014] Preferably, the compensatory control engine built into the system optimization module is configured to: receive the risk pre-control strategy set as input; the engine includes a strategy parser that, based on an instruction mapping rule base, converts each strategy in the risk pre-control strategy set into a device control instruction with explicit execution parameters; the engine is also configured to transmit the generated device control instruction to the target traffic safety facility through the heterogeneous fusion self-healing elastic network.

[0015] Preferably, the compensation control engine is further configured to: continuously acquire status data and environmental data of the affected area from the data acquisition module during and after the execution of the device control command; the engine has a built-in effect evaluation model, which calculates a quantitative performance index by comparing the status and environmental data before and after the command execution; the engine generates a model update command based on the quantitative performance index and transmits the command to the strategy output module to adjust the generation logic of its risk pre-control strategy set.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. This application constructs and operates a heterogeneous, converged, self-healing, resilient network, transforming the transmission method of traffic data from a traditional, rigid link relying on a single communication protocol to a multi-path, adaptive, resilient data network. This network can dynamically select the optimal path and self-heal from faults based on operating costs and service quality. This fundamentally overcomes the shortcomings of the prior art, such as system paralysis due to communication interruptions and high data transmission costs, providing a highly reliable, resilient, and economical data foundation for all subsequent risk prediction and closed-loop control.

[0017] 2. This application transforms traffic risk management from a passive, delayed event response model based on background technologies to a proactive, forward-looking risk prediction model by constructing a network of traffic facility dependencies and a facility association map, and by using a diffraction chain simulation sandbox for deduction. It can quantitatively and visually deduce the cascading impact on the entire transportation network before a single facility failure occurs, elevating safety management from a reactive, post-event remediation approach to a proactive level of pre-event prevention and in-event intervention, significantly enhancing the proactive safety and operational resilience of urban transportation networks.

[0018] 3. By integrating the above-mentioned elements, this application constructs a full-process intelligent control system encompassing "elastic perception, correlation modeling, risk prediction, and closed-loop optimization." Through the quantitative evaluation and feedback mechanism of the compensatory control engine, this system achieves a core transformation from passive alarm to proactive intervention, and from fixed rules to self-evolution. This endows the invention with deep adaptive capabilities, fundamentally solving the shortcomings of existing static decision-making models in handling dynamic and complex traffic environments, and realizing long-term intelligent optimization of traffic safety management. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of a digital system structure for traffic safety facilities based on intelligent network connectivity, according to the present invention. Figure 2 This is a schematic diagram of a digital system for traffic safety facilities based on intelligent connectivity, according to the present invention. Figure 3 This is a core workflow diagram of the network control unit of a digital system for traffic safety facilities based on intelligent connectivity, according to the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Example 1: Reference Figure 1 and Figure 2 This embodiment provides a specific application of the aforementioned digital system for traffic safety facilities based on intelligent connectivity. The technical solution of this invention is applied to a traffic safety and traffic efficiency optimization project on "Binjiang Avenue," a core arterial road in a megacity. This road section experiences massive traffic flow, complex intersections, dense pedestrian traffic, and is adjacent to an international convention and exhibition center, making it susceptible to sudden congestion due to temporary events. The core challenge of this project lies in the fact that traditional passive and isolated traffic management methods are insufficient to address the dynamic and complex traffic risks of this road section, necessitating a digital system capable of proactive prediction, intelligent collaboration, and self-optimization.

[0022] Furthermore, in one specific implementation of this embodiment, the data acquisition module collects data on traffic facilities along "Binjiang Avenue" in the following ways: On each traffic signal pole, status data is collected through built-in sensors, including power supply voltage for monitoring power stability, equipment temperature for reflecting controller health, enclosure airtightness for judging the waterproof performance of the enclosure, and pole vibration amplitude for assessing pole structural risk; dynamic traffic environment data is collected through millimeter-wave radar and high-definition cameras integrated on the pole, including traffic flow statistics at specific sections of each approach lane at the intersection, average vehicle speed, headway between vehicles, and the actual number of pedestrians crossing the crosswalk.

[0023] In this embodiment, by collecting multi-source data from two dimensions—physical state and traffic environment—the system can comprehensively assess the condition of facilities and accurately perceive their operating environment. This overcomes the limitations of traditional technologies that rely on a single data source for traffic statistics, providing a high-quality, high-dimensional data foundation for subsequent refined modeling and accurate prediction.

[0024] Reference Figure 3 Furthermore, in one specific implementation of this embodiment, the dynamic scheduling of link resources by the network control unit is specifically implemented as follows: The network control unit software deployed in the city traffic command center first establishes and maintains in real time a communication link resource pool for all facilities along the "Binjiang Avenue" through network scanning and active device registration. This resource pool records in detail the 5G public network links and Wi-Fi Mesh self-organizing network links possessed by each facility. The network control unit software internally includes: a link resource management module, responsible for establishing and maintaining the link resource pool in the form of a database table; a real-time status monitoring module, responsible for periodically performing detection tasks and updating the status parameters of each link in the resource pool; and a routing scheduling instruction generation module, which responds to upper-layer business requirements by querying the database of the link resource management module, calling the latest data from the real-time status monitoring module, making decisions, and generating the final scheduling instruction. During system operation, the network control unit sends lightweight probe commands to the communication terminals of all facilities at a frequency of one second to continuously obtain the real-time network latency and packet loss rate of each link. When data needs to be transmitted, the network control unit selects a specified link from the resource pool according to the upper-layer service requirements and issues a data transmission scheduling command containing IP address, port and protocol type to the corresponding terminal.

[0025] In this embodiment, by establishing a link resource pool and performing centralized intelligent scheduling, independent communication links are transformed into a unified network resource. This enables the system to match the optimal communication path for high-priority data based on the global state, ensuring the reliability and timeliness of data transmission in complex electromagnetic environments, thereby providing a stable data interaction guarantee for the system.

[0026] Furthermore, in a specific implementation of this embodiment, the step of constructing and running a heterogeneous converged self-healing resilient network is as follows: when a new smart light pole on "Binjiang Avenue" is installed and powered on, its built-in multi-mode communication terminal immediately registers with the server address of the network control unit and reports the 5G and Wi-Fi communication protocol types it supports; after successful authentication, the network control unit immediately instructs the terminal to establish physical connections with the 5G base station and the Wi-Fi module of the adjacent light pole, and begins continuous status detection of it; the terminal and its two communication links are then added to the network topology graph, and the network control unit maintains the continuous operation of the entire network based on the detection results of all nodes, and dynamically forms a converged network topology in which 5G and Wi-Fi paths coexist and the status of each path is known in real time.

[0027] In this embodiment, the automated registration and networking steps simplify the deployment process for adding or replacing facilities, reduce the complexity and error rate of manual configuration, and enable the entire resilient network to have high scalability and maintainability, and adapt to the dynamic updates of urban infrastructure.

[0028] Furthermore, in a specific implementation of this embodiment, the implementation method of optimal path selection and fault switching based on link status and data attributes is as follows: When a traffic accident occurs on "Binjiang Avenue", the video data stream collected by the camera is parsed into a service requirement vector containing specific technical indicators. This vector defines {latency requirement: less than 30ms; bandwidth requirement: greater than 20Mbps; reliability requirement: greater than 99.9%}. At the same time, the network control unit queries the status of the two available links of the terminal where the camera is located in real time. The status of link A (5G) is {latency: 25ms; available bandwidth: 50Mbps; reliability: 99.95%; cost factor: 10}, and the status of link B (Wi-Fi Mesh) is {latency: 80ms; available bandwidth: 35Mbps; reliability: 99.8%; cost factor: 2}. When selecting the optimal path, the network control unit initiates a preset weighted matching algorithm. The decision logic of this algorithm follows a step-by-step evaluation process: 1. Eligibility Screening: The algorithm first compares the real-time status of each link with the service requirement vector of the data, temporarily excluding links that do not meet any hard metric. In this example, link B (Wi-Fi) is excluded because its 80ms latency does not meet the requirement of less than 30ms and will not proceed to the next scoring stage.

[0029] 2. Weighted Scoring: The algorithm first performs dimensionless processing on the filtered link status parameters. For parameters such as latency, where lower values ​​are better, the reciprocal normalization method is used, i.e., the normalized latency value equals the base constant divided by the actual latency value; for parameters such as bandwidth, where higher values ​​are better, the linear proportional method is used, i.e., the normalized bandwidth value equals the actual bandwidth value divided by the system's maximum reference bandwidth. For all candidate links that pass the screening, the algorithm calculates a comprehensive performance score based on the degree to which they meet each indicator, combined with preset weight coefficients associated with data service attributes. This score is equal to the sum of the products of each normalized parameter value and its corresponding weight coefficient. Since the data in this case is emergency video, the weight coefficients for latency and bandwidth are set to the highest, therefore link A obtains a high comprehensive performance score. The weight coefficients are stored in a weight lookup table, which predefines the mapping relationship between different service attributes and a set of specific weight coefficient vectors (e.g., the weight vector for emergency video is {latency weight: 0.6, bandwidth weight: 0.3, reliability weight: 0.1}). When performing weighted scoring, the algorithm first retrieves the corresponding weight vector from the lookup table based on the service attributes of the data to be transmitted, and then uses the coefficient values ​​in the vector to perform weighted summation on the degree to which the candidate links meet the various indicators, and calculates the comprehensive performance score.

[0030] 3. Cost Trade-off: The algorithm applies a penalty adjustment to the overall performance score based on the cost factor of the link itself to arrive at the final matching degree. The specific calculation logic is as follows: the final matching degree equals the overall performance score divided by the cost adjustment coefficient. The cost adjustment coefficient is equal to the product of the link cost factor and the sensitivity constant. Through this division operation, links with higher cost factors will have their final matching degree significantly lowered. This allows for prioritizing lower-cost links (such as Wi-Fi) when performance is similar, and only switching to higher-cost links (such as 5G) when the performance advantage is sufficient to offset the cost penalty.

[0031] Ultimately, the system selects link A (5G) with the highest final matching degree as the optimal path; During failover, the network control unit continuously monitors the transmission of the 5G link. If it detects that the latency has deteriorated to 200ms due to network congestion, exceeding the preset health threshold of 100ms, it immediately triggers a rerouting mechanism. This mechanism first removes the 5G link from the set of available links in the current routing calculation, and then immediately re-executes the aforementioned complete evaluation process of qualification screening, weighted scoring, and cost trade-offs for the video data stream. It calculates a new optimal path (i.e., the Wi-Fi of link B) among the remaining links, thereby seamlessly switching the data stream and ensuring the continuity of video backhaul.

[0032] To achieve transparent and seamless handover at the network layer, the network control unit employs a multipath transmission protocol at the transport layer. All data packets are encapsulated into substreams with global sequence numbers at the sending end and sent through different communication interfaces. When a rerouting mechanism is triggered to switch physical links, the network control unit only changes the sending interface of the substreams, while keeping the transport layer session identifier unchanged. The receiving end reassembles and deduplicates data packets from different links based on the global sequence numbers. Therefore, the connection state perceived by the application layer remains active at all times, thus achieving seamless handover with zero packet loss even when a physical link is interrupted.

[0033] In this embodiment, the mechanism ensures the continuity of critical information transmission. Even in environments with drastic network fluctuations, high-priority data streams can maintain stable transmission through a dynamic rerouting mechanism, providing reliable technical support for emergency response and thus enhancing the overall resilience of the system.

[0034] Furthermore, in one specific implementation of this embodiment, the generation of structured real-time data stream is achieved as follows: the radar point cloud data, video image data, vibration sensor data, etc. collected by the system from various facilities along the "Binjiang Avenue" are first timestamped in the data processing module and uniformly calibrated to standard UTC time; then, all geographical location information is uniformly converted to the city standard CGCS2000 coordinate system; next, all data is uniformly encoded into JSON format, and each piece of data is appended with its facility identification identifier, such as "BJD-LS-01"; finally, this standardized data is encapsulated into a unified format data packet and output to the upper-layer policy output module.

[0035] In this embodiment, the original heterogeneous data is standardized to transform the multi-source data from the front end into unified, high-quality structured information. This step solves the problems of data silos and inconsistent standards in traditional systems, providing a homogeneous and consistent data foundation for subsequent modules to perform correlation analysis, risk simulation, and intelligent decision-making.

[0036] Furthermore, in a specific implementation of this embodiment, the detailed steps of constructing the traffic facility dependency network and generating the facility association map are as follows: The relational reasoning engine of the strategy output module parses the input structured data stream and, in conjunction with the facility basic information database such as GIS map and equipment ledger, identifies and quantifies the dependency relationships between various facilities on "Binjiang Avenue" to construct a dependency network. Specifically, the spatial adjacency relationship between the intersection traffic lights and the upstream camera is defined based on the GIS map, and the upstream and downstream traffic flow impact relationship of the upstream intersection on the downstream intersection is dynamically calculated based on traffic flow data. This module then maps and transforms the network, abstracting each traffic facility (such as "BJD-LS-01") as a node and the dependency relationships between facilities (such as "influence intensity 0.85") as weighted directed edges connecting nodes, thereby generating a facility association map that serves as the input object for the diffraction chain pre-simulation sand table.

[0037] In this embodiment, by constructing a dependency mesh and association graph, the system's analytical dimension is elevated from isolated facility nodes to a networked, interconnected level. This step enables the system to understand the complex physical and functional dependencies between facilities, providing the necessary topological model foundation for achieving system-level prediction and coordinated control.

[0038] The relational reasoning engine also includes: Based on a pre-defined infrastructure information database, a static relationship layer is constructed that defines the physical connections and functional dependencies between facilities. Based on structured real-time data streams, a dynamic relationship layer that quantifies the impact of traffic flow transmission is dynamically calculated and updated through time-series cross-correlation analysis. Based on external event inputs and a built-in microscopic traffic simulation model, changes in the dependencies between facilities under specific future scenarios are predicted, generating a predictive relationship layer. Finally, the engine weighted and fused these three relationship layers to generate a multi-layered dynamic facility association graph.

[0039] The engine first constructs a static relationship layer based on the input pipeline topology map and equipment ledger. This layer clarifies that all traffic lights along "Binjiang Avenue" belong to the same power supply circuit, and that camera 02 is functionally dedicated to monitoring intersection 03. During system operation, the engine continuously analyzes real-time traffic data to construct a dynamic relationship layer. For example, during the morning rush hour, it calculates through time-series cross-correlation analysis that the traffic flow at intersection 01 and the queue length at intersection 02 have a correlation coefficient as high as 0.9, thus establishing a strongly influential directed edge with a weight of 0.9 between these two nodes. Most importantly, when the system receives an external event input from the city's event calendar stating "a large performance will end at the convention center at 8 PM," the engine immediately initiates the construction of the predictive relationship layer. Through its built-in microscopic traffic simulation model, it simulates a scenario where a large flow of people and vehicles surges from the convention center to "Binjiang Avenue" and the surrounding elevated roads, and predictively generates relationship changes for the next hour: it calculates that the influence weight between the convention center exit and entrance 05 of "Binjiang Avenue" will temporarily surge by 300%, and a new, temporary, strong dependency relationship will emerge between "Binjiang Avenue" and the ramps leading to the city expressway. Finally, the engine weighted and merges these three layers to output a multi-layered dynamic facility relationship map that includes fixed physical constraints, reflects the current real traffic flow, and predicts future states, for use in subsequent diffraction chain pre-simulation sandboxes.

[0040] The relation reasoning engine, when constructing the dynamic relation layer, specifically includes feature sliding window and threshold determination steps: The system sets a time sliding window for the traffic data collected from all facilities, with a window length of fifteen minutes. Within each window, the system calculates the Pearson correlation coefficient between the data sequences of two nodes.

[0041] The system sets a correlation threshold of 0.75. If the calculated Pearson correlation coefficient is greater than 0.75, a strong dependency is determined between the two nodes, a connection edge is created in the association graph, and the correlation coefficient value is directly used as the initial weight of the edge; if the correlation coefficient is less than 0.75, no connection edge is created or existing connection edges are pruned. This dynamic threshold determination mechanism ensures that the graph can filter out accidental noise interference and retain only traffic associations with statistical significance.

[0042] By constructing a multi-layered dynamic relationship inference engine, the system's understanding of the transportation network is elevated from a two-dimensional analysis based on history and current status to a three-dimensional insight encompassing future states. This enables the system to scientifically predict how connections will arise under specific events. This forward-looking network relationship insight allows subsequent risk simulations to move beyond static topology simulations and instead be based on a dynamically evolving scenario that more closely approximates future reality, fundamentally improving the accuracy and timeliness of risk prediction.

[0043] Further, in a specific implementation of this embodiment, the step of using the diffraction chain pre-simulation sandbox to perform risk path deduction is as follows: When the traffic lights at the main intersection of "Binjiang Avenue" malfunction, the malfunctioning facility is designated as the initial event node set for deduction; after receiving the instruction, the diffraction chain pre-simulation sandbox performs risk deduction based on the node state propagation logic. The specific steps are: defining the node states in the graph as normal state, affected state, and intervened state. Starting from the initial malfunctioning node, traversing all its outgoing edges. For each edge connecting an upstream node to a downstream node, calculating the risk propagation probability. The calculation logic for this risk propagation probability is: taking the weight of the edge connecting the two nodes (representing the traffic flow influence coefficient), multiplying it by an attenuation coefficient (this coefficient is obtained by subtracting the product of the time attenuation factor and the propagation step size). If the calculated risk propagation probability is greater than a preset risk activation threshold, the state of the downstream node changes from normal state to affected state at the next moment, and it is recorded as the next hop node of the risk diffraction chain. This process iterates until no new nodes are activated, ultimately generating a risk diffraction chain data structure containing all affected state nodes and their activation timestamps. This data structure clearly presents the time series of the propagation path, scope of impact, and degree of impact of the fault spreading upstream and downstream along Binjiang Avenue within 15 minutes, eventually causing traffic paralysis at major intersections and ramps along the route. The strategy output module then matches and combines a risk pre-control strategy set from the intervention measures knowledge base based on the deduction results of this diffraction chain. This strategy set includes specific measures such as temporary traffic control at upstream intersections and issuing detour suggestions to the navigation platform.

[0044] In this embodiment, the risk simulation step transforms risk management from a reactive, reactive model to a proactive, predictive model. By quantifying and predicting the propagation path and scope of chain reactions, it enables managers to develop more targeted intervention plans, preventing local risks from escalating into systemic problems at a lower cost.

[0045] The intervention knowledge base also includes: A modular strategy synthesis engine is established, which has a built-in modular intervention knowledge base. The modular intervention knowledge base stores multiple atomic intervention modules with configurable parameters. The modular strategy synthesis engine is also configured to: receive risk diffraction chain data structure as input, perform optimization calculations through reinforcement learning based on the control objective function, select and combine intervention modules from the knowledge base, configure optimal parameters for each module, and finally synthesize a complete and customized risk prevention and control strategy set.

[0046] When the diffraction chain simulation model predicts a risk diffraction chain that will paralyze five intersections within 15 minutes, the modular strategy synthesis engine is activated. Its objective function is set to control the average queue length growth rate in the predicted affected area to within 20% within 15 minutes, and to minimize the total impact index of the intervention measures. The engine's built-in modular intervention knowledge base stores various atomic intervention modules, such as: an upstream traffic light greening ratio reduction module, a variable message sign information dissemination module, and a navigation platform path guidance module. The engine then initiates a genetic algorithm for rapid optimization. It generates multiple initial strategy populations composed of the aforementioned modules; for example, strategy A contains only a high-intensity traffic light reduction module, while strategy B contains a moderate traffic light reduction module and an information sign dissemination module. By rapidly simulating, cross-referencing, and mutating these strategies in a digital sandbox, the engine iterates to find the optimal solution within seconds: a composite strategy consisting of an upstream traffic light greening ratio reduction module (BJD-LS-02, 45s, 15min) and a variable message sign information dissemination module (VMS-01, template 3). This optimal combination is ultimately encapsulated as the risk prevention and control strategy set for this event.

[0047] The modular strategy synthesis engine employs a real-number encoded genetic algorithm. The specific algorithm configuration and steps are as follows: Chromosome coding: A complete set of risk prevention and control strategies is defined as a chromosome, and each gene locus on the chromosome corresponds to a specific intervention module parameter. For example, the first gene locus represents the green light ratio reduction rate, with a value ranging from 0.1 to 0.5; the second gene locus represents the rate limit value, with a value ranging from 40 to 80; and the third gene locus represents the induction screen display template number, with a value ranging from 1 to 5.

[0048] Population initialization: Randomly generate fifty initial solutions that satisfy the above value range constraints, and use them as the first generation population.

[0049] Fitness function: The fitness value is directly obtained by using the reciprocal of the comprehensive risk index calculated above. That is, the lower the comprehensive risk index, the higher the fitness value of the individual and the greater the probability of being retained.

[0050] Genetic operations: roulette wheel selection was used to select parent individuals; arithmetic crossover was used for crossover operations with a crossover probability of 80%; and non-uniform mutation was used with a mutation probability of 10%.

[0051] Termination condition: The algorithm terminates when the population iteration count reaches 100, or when the fitness value of the best individual no longer increases for five consecutive generations, and outputs the strategy combination corresponding to the current best individual.

[0052] By constructing a modular strategy synthesis engine, the system's emergency response method is upgraded from calling fixed, pre-set solution libraries to a mechanism that dynamically generates customized strategies. This mechanism overcomes the limitation of traditional knowledge bases, which can only deal with known, standard scenarios; through the intelligent combination of atomic intervention measures, it can generate new and optimal response strategies for unexpected and complex risk scenarios. This leap from selecting existing options to generating new solutions endows the system with strong decision-making flexibility and intelligence when dealing with complex and uncertain environments.

[0053] Further, in one specific implementation of this embodiment, the step of constructing the compensatory control engine and executing control commands is as follows: After receiving a risk pre-control strategy set containing "temporary traffic control at the upstream intersection," the system optimization module instantiates and starts the compensatory control engine; the strategy parser within the engine, based on the command mapping rule base, converts this abstract strategy into a device control command with specific execution parameters for the "Binjiang Avenue No. 02" traffic signal, namely, "reduce the cycle duration of the north-south green light signal from 60 seconds to 40 seconds"; the compensatory control engine then accurately transmits the command to the execution unit of the "Binjiang Avenue No. 02" traffic signal through a heterogeneous fusion self-healing elastic network. The command mapping rule base uses a structured text format to store the mapping table. Each mapping rule contains three parts: an abstract strategy template, a device protocol driver, and a parameter conversion formula. For example, for a strategy containing the action type of traffic control and the target object of the upstream intersection, the rule base matches the corresponding device command set, which includes the command code, device type, and parameter value calculation logic. The parameter value calculation logic is described as follows: multiply the current device's signal cycle duration by a preset reduction factor (e.g., 0.6) to obtain the target value. The policy parser reads real-time cycle data, substitutes it into the above logic to calculate the specific value, and encapsulates it into a control frame conforming to the standard communication protocol.

[0054] In this embodiment, this step establishes an automated strategy parsing and instruction conversion mechanism. It can accurately translate high-level, goal-oriented control strategies into executable control instructions with specific parameters for lower-level facilities, ensuring the accurate implementation of top-level decisions and serving as a necessary step in achieving automated closed-loop control.

[0055] The compensation regulation engine also includes: The system optimization module incorporates an online strategy tuning engine. This online strategy tuning engine is configured to: enter a real-time control state when executing an input risk prevention and control strategy; in this state, the online strategy tuning engine continuously fine-tunes the key parameters of the currently executed instruction through an online tuning inner loop, based on high-frequency collected state data, to minimize the deviation between the actual state and the control target; after the control state ends, it quantitatively evaluates the overall performance of this control through a strategy evolution outer loop, and feeds back structured experience data, including the tuning process and the final result, to the strategy output module.

[0056] Upon receiving the risk control strategy of "reducing the green light duration for north-south traffic at intersection 02 upstream to 40 seconds for 15 minutes," the engine immediately executes the instruction and enters a real-time control state aimed at maintaining the queue length at the original faulty intersection below 50 meters. In this state, the engine's online tuning inner loop is activated: it continuously acquires real-time queue length data from the data acquisition module at a high frequency of once every 10 seconds. Five minutes after the control begins, the engine detects that due to an unexpected decrease in traffic flow on a ramp, the actual queue length is only 20 meters, below the 50-meter target. At this point, the inner loop's tuning logic determines that the 40-second green light restriction is too stringent and is causing unnecessary congestion at intersection 02. Therefore, it automatically fine-tunes the green light duration parameter to 45 seconds. Two minutes later, it detects an increasing trend in queue length and adjusts the parameter back to 42 seconds. This online tuning process continues throughout the 15-minute intervention period, ensuring that the control intensity and actual demand maintain a dynamically optimal balance. After the 15-minute intervention period ended, the engine's strategy evolution outer loop began to operate. It not only evaluated the overall performance of the intervention, which reduced the congestion worsening trend by 70%, but also encapsulated the entire parameter tuning process—from 40 seconds to 45 seconds and then to 42 seconds—as valuable structured empirical data. Finally, this empirical data was fed back to the strategy output module.

[0057] By constructing an online policy tuning engine, the closed-loop control of the system is elevated from slow feedback based on inter-event learning to a dual-loop mode combining intra-event optimization and inter-event learning. The online tuning inner loop overcomes the static and lag-based nature of traditional closed-loop systems when executing policies, enabling the system to respond in real-time to minor and unexpected changes during intervention, thus achieving more precise and efficient process control. The policy evolution outer loop solidifies this dynamic tuning experience, rich in process details, into long-term knowledge, achieving a deeper and more insightful model evolution than a single performance score, greatly enhancing the system's adaptability and control accuracy in complex and uncertain environments.

[0058] Furthermore, in one specific implementation of this embodiment, the quantitative evaluation and closed-loop optimization steps are as follows: After the signal at "Binjiang Avenue No. 02" executes the time reduction command, the compensatory control engine continuously obtains traffic data from the downstream (i.e., the original faulty intersection) of the intersection from the data acquisition module; the engine's built-in effect evaluation model calculates the quantitative performance indicator of this control as "congestion deterioration trend reduced by 70%" by comparing the queue length growth rate of the downstream intersection before and after the command execution; the compensatory control engine generates a model update command based on this performance indicator and transmits the command to the strategy output module to increase the weight and confidence of the strategy "reduce the upstream green ratio" in its intervention measure knowledge base, so that it will be called first in similar events in the future.

[0059] In this embodiment, the closed-loop optimization step utilizes quantitative evaluation and feedback methods to construct a mechanism that allows the strategy to optimize itself. This mechanism enables the system to continuously learn from the actual effects of historical interventions, allowing the effectiveness of its control strategy to continuously improve over time. This, in turn, allows it to adapt well to the dynamic changes in the urban traffic environment, achieving long-term intelligent optimization results.

[0060] In summary, this embodiment fully demonstrates the end-to-end application of a digital traffic safety facility system based on intelligent connectivity in the complex urban scenario of "Binjiang Avenue". In this application, the system first collects real-time data on facilities and traffic along the route through a data acquisition module. Then, the data processing module, leveraging a heterogeneous fusion self-healing elastic network, ensures stable transmission of critical data even when local 5G signal congestion occurs due to events at the convention center, and converts the data into structured real-time data streams. Based on this, the strategy output module constructs a facility association map reflecting the traffic transmission patterns of "Binjiang Avenue" using these data streams. Using a diffraction chain simulation model, it successfully predicts the risk that a single traffic light malfunction at an intersection could paralyze the entire main road within 15 minutes, thereby generating a risk pre-control strategy set that includes upstream traffic light flow management. Finally, the system optimization module activates a compensatory control engine to precisely execute the strategy. By quantitatively evaluating the intervention effect, successful experiences are solidified into the strategy model, completing the entire process from risk perception to closed-loop optimization in the actual management of "Binjiang Avenue".

[0061] In this embodiment, the system of the present invention overcomes the core defects of the traditional traffic management model of "Binjiang Avenue," such as weak communication security, delayed risk detection, and rigid congestion control. By deploying a heterogeneous fusion self-healing elastic network, the reliability of data communication is significantly improved under complex conditions such as large passenger flows at the convention center; by utilizing a diffraction chain pre-simulation sand table, the risk response capability of the management unit is shifted from post-event handling to pre-event accurate prediction; and through the closed-loop learning of the compensatory control engine, the control strategy can self-optimize according to the unique traffic tidal characteristics of "Binjiang Avenue." In summary, in the specific application of "Binjiang Avenue," this system transforms a passive, static monitoring system into an intelligent operation and management system with high resilience, foresight, and adaptability, achieving a comprehensive improvement in traffic safety and traffic efficiency in this core road section.

[0062] Example 2: This embodiment provides another specific application of the aforementioned intelligent connected vehicle-based digital traffic safety facility system. The technical solution of this invention is applied to the safety management of the Qinling section of the Beijing-Kunming Expressway, a mountainous highway. The Qinling section of the Beijing-Kunming Expressway is approximately 55 kilometers long, with significant altitude variations, and is prone to severe weather conditions such as fog and icy roads in winter, making it a typical high-accident area. The core objective of this application is to achieve long-term, preventative risk management of this road section through this system, rather than simply providing emergency response to a single accident.

[0063] Furthermore, in this embodiment, to address the issue of fixed network routing decision weights in the data processing module, the method for constructing and optimizing the "weight lookup table" is specifically as follows: The initial weight vector of the lookup table is preset by network experts based on historical experience. After the system is put into operation, a weight optimization agent based on reinforcement learning is deployed. This agent has a one-month learning cycle. Its "state" is the statistical data of all network transmissions collected within the cycle (such as the average latency, packet loss rate, and cost distribution of various services), its "action" is to fine-tune the weight coefficient values ​​in the weight lookup table, and its "reward" is an objective function that aims to maximize the transmission success rate of high-priority data while minimizing the overall network transmission cost. Through continuous "action-reward" loops, the agent can autonomously learn and iterate to find the most suitable weight configuration for the unique network environment along the Qinling section of the Beijing-Kunming Expressway, and then distribute the optimized lookup table to the network control unit, thereby realizing the dynamic adaptation of the routing decision logic. To ensure that the reinforcement learning agent can converge, this embodiment clearly defines the model's state space, action space, and reward function: State Space: A feature vector is constructed. To accommodate the dynamically changing number of links in a resilient network, this vector dimension is set to four times the maximum number of links supported by the system (padded with zeros if necessary). For each valid link, the feature contains four dimensions: the current latency value, the latency variance over a preset time period (e.g., five minutes), the current packet loss rate, and the link congestion flag. This feature vector is normalized and then used as the state input.

[0064] Action Space: A discrete action space is used. For each weight coefficient in the weight lookup table, the action set is defined to contain three discrete values: -0.1, 0, and 0.1. This means that the agent can only make small, incremental adjustments based on the current coefficients to ensure the stability of the network weights and avoid large fluctuations.

[0065] Reward Function: The reward value is calculated as follows: First, the ratio of the 'current average latency' to the 'system's maximum tolerable latency' is calculated. This ratio is then subtracted from the given value and multiplied by a first preset hyperparameter. Next, the total cost of this adjustment is calculated and multiplied by a second preset hyperparameter. Finally, the product of the former and the latter is subtracted to obtain the final reward value. The first and second preset hyperparameters are set to 1.0 and 0.2, respectively. Furthermore, when the network transmission success rate is below 95%, the system will directly impose a penalty of -100 points and force a parameter rollback.

[0066] In this embodiment, by introducing an optimization agent based on reinforcement learning, the setting of the weight lookup table is transformed from a one-time expert definition into a dynamic process capable of self-learning and continuous evolution. This not only solves the problem of "insufficient public disclosure" of weight sources, but also ensures in principle that the network's routing strategy can adapt to environmental changes in the long term and always remain in an optimal state.

[0067] Furthermore, in this embodiment, to address the risk of fog on mountain highways, the "modular strategy synthesis engine" of the strategy output module explicitly defines its objective function when generating strategies as seeking a strategy combination with the lowest overall risk index. The overall risk index is obtained by weighted summation of the sub-indices across three dimensions, as follows: 1. Traffic Efficiency Impact Index: This index reflects the negative impact of intervention measures on traffic efficiency. The system obtains the predicted total vehicle delay time (in seconds) and sets a maximum tolerable delay threshold (e.g., 180 seconds). The index calculation logic is as follows: if the total vehicle delay time is less than the maximum tolerable delay threshold, the index value is equal to the total vehicle delay time divided by the maximum tolerable delay threshold and then multiplied by 100; if the total vehicle delay time is greater than or equal to the maximum tolerable delay threshold, the index value is directly set to 100, indicating that the congestion risk has reached saturation; if the denominator value approaches zero, the system will introduce a very small smoothing constant into the calculation to avoid calculation errors. 2. Active Safety Risk Index: The Active Safety Risk Index assesses the risk of secondary accidents such as rear-end collisions and collisions by analyzing vehicle speed dispersion and headway on the road segment after intervention, and obtains an exponential result. This index is based on a nonlinear mapping of vehicle speed dispersion. The system divides the vehicle speed standard deviation into three intervals: 'low risk, medium risk, and high risk', and uses a piecewise function for calculation: In the low-risk interval (e.g., standard deviation less than 10 km / h), the index value increases slowly and linearly with the standard deviation; once entering the high-risk interval (e.g., standard deviation greater than 20 km / h), the index value increases sharply and exponentially with the standard deviation, to amplify the safety weight under severe weather conditions; 3. Resource Scheduling Cost Index: The resource scheduling cost index is a fixed cost score assigned to each different intervention module, such as variable speed limit signs, tunnel broadcasts, and ramp traffic lights, reflecting the implementation cost of the measure. When summing up the comprehensive risk index, the weight coefficients of each sub-index will vary depending on the scenario; for example, calling 'tunnel broadcasts' is assigned a value of 10, calling 'ramp traffic lights' is assigned a value of 30, and calling 'variable speed limit across the entire line' is assigned a value of 50. When calculating the comprehensive risk index, the system uses the formula: the comprehensive risk index equals the sum of the products of each sub-index and its corresponding scenario weight coefficient. For example, in this case, the scenario is fog on a mountain highway, so the weight of the active safety risk index is set to the highest. The optimization calculation of the strategy synthesis engine essentially involves iteratively finding, among all possible combinations of intervention modules, the combination that minimizes the total value of the weighted comprehensive risk index, and outputting this as the optimal strategy.

[0068] In this embodiment, by clearly defining the objective function and its composition, a clear, quantifiable, and reproducible logical basis is provided for the optimization calculation of the strategy synthesis engine. This transforms the system's strategy generation process from a vague concept into a rigorous, scientific optimization process aimed at multi-dimensional risk management.

[0069] Furthermore, in this embodiment, to address the learning disconnect between the policy output module and the system optimization module, the present invention introduces a "two-layer feedback and co-evolution" mechanism to achieve deep coupling between the "modular policy synthesis engine" and the "online policy tuning engine".

[0070] When a fog warning is issued on the Qinling section of the Beijing-Kunming Expressway, the modular strategy synthesis engine (located in the strategy output module) will first generate a risk pre-control strategy from a macro perspective, such as: {activating the variable speed limit module (setting the target speed to 60km / h) and simultaneously activating the ramp traffic light control module (setting the red light cycle to 30s)}.

[0071] Once the strategy was issued, the "online strategy tuning engine" (located within the system optimization module) immediately started running and entered a real-time control state. During execution, the online tuning inner loop detected that when fog concentration increased again, the originally set target speed of 60 km / h was still somewhat too high. It continuously made minor adjustments to the speed parameter until it was lowered to 55 km / h, thus ensuring that the road's risk index (such as the number of sudden braking incidents) remained within the pre-set target threshold range.

[0072] After the regulatory event concludes, the outer loop of the "online strategy tuning engine" will, on the one hand, feed back the overall performance score of the macro-strategy combination, and on the other hand, provide an additional fine-tuning experience package containing process details. This co-evolutionary mechanism works as follows: 1. Regarding the feedback given by the strategy synthesis engine, its overall performance score will be used as the evaluation value of the "fitness function" in the genetic algorithm, thereby directly guiding its future trend and making it more likely to choose a macro strategy combination such as "variable speed limit plus ramp control" in subsequent situations.

[0073] 2. Based on the feedback from the intervention module knowledge base, the fine-tuning experience package is specifically designed to optimize the atomized intervention modules themselves. For example, the "variable speed limit module" in the knowledge base will learn that in a scenario like "fog in the Qinling Mountains," the initial suggested value for the "target speed" parameter needs to be revised from the general value of 80 km / h to a more targeted 60 km / h.

[0074] In this embodiment, a "two-layer feedback and co-evolution" mechanism is constructed to closely link the two core engines. The policy synthesis engine primarily explores and discovers the macro-optimal combination of "what" and "what." In contrast, the online policy tuning engine focuses on refining the micro-optimal parameters of "how" during actual execution and feeding this valuable experience back to the former, thus guiding its future macro-level decisions. This co-evolutionary model allows the system's learning ability to achieve a complete closed loop from macro to micro and back again, significantly improving the system's long-term evolutionary capability and decision-making accuracy when facing complex problems.

[0075] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A digital system for traffic safety facilities based on intelligent connected vehicles, characterized in that, include: The data acquisition module collects real-time status data of traffic safety facilities and dynamic traffic environment data; The data processing module, through the network control unit, dynamically schedules link resources of two communication protocols to construct and run a heterogeneous fusion self-healing elastic network; it performs optimal path selection and fault switching based on link status and data attributes, and merges the status data and dynamic traffic environment data to generate a structured real-time data stream. The strategy output module, based on the real-time data stream, quantifies the dependencies between facilities through a relational reasoning engine, constructs a transportation facility dependency relationship grid, generates a facility association map through the transportation facility dependency relationship grid, uses a diffraction chain pre-simulation sandbox to perform risk path deduction, and generates a risk pre-control strategy set. The system optimization module, based on the risk pre-control strategy set, constructs a compensation control engine, parses the risk pre-control strategy set, and issues control instructions for execution. By combining newly collected status data and dynamic traffic environment data, the execution of the control commands is quantitatively evaluated, and the evaluation results are fed back to the strategy output module. 2.The intelligent network-based traffic safety device digitalization system of claim 1, wherein, The status data of the traffic safety facilities include: power supply voltage, equipment temperature, enclosure airtightness, and pole vibration amplitude; the dynamic traffic environment data includes: cross-sectional traffic flow, average driving speed, vehicle headway, and number of pedestrians crossing. 3.The intelligent network-based traffic safety device digitalization system of claim 1, wherein, The network control unit dynamically schedules link resources, specifically by: first establishing and maintaining a link resource pool containing all multi-mode communication terminals and available communication protocols; then periodically sending probe commands to the multi-mode communication terminals to obtain the real-time status of each link. The network control unit selects a specified link from the link resource pool according to the upper-layer service requirements and sends a data transmission scheduling instruction to the corresponding terminal. 4.The intelligent network-based traffic safety device digitalization system of claim 1, wherein, The specific steps for constructing the heterogeneous fusion self-healing elastic network are as follows: multi-mode communication terminals on each traffic safety facility actively register with the network control unit and report all supported communication protocol types; the network control unit enables the terminals to establish connections with multiple communication networks and continuously perform status detection; the network control unit maintains the continuous operation of the network based on the detection results, dynamically forming a fusion network topology with multiple paths coexisting and real-time status awareness. 5.The intelligent network-based traffic safety device digitalization system of claim 1, wherein, The implementation method of optimal path selection and fault switching based on link status and data attributes is as follows: When selecting the optimal path, the network control unit parses the service attributes of the data to be transmitted, combines the status parameters of each available link queried in real time, and calculates the communication link that matches the data requirements as the transmission path through weighted matching; when performing fault switching, when the network control unit continuously monitors the communication quality of the current transmission link and it is lower than a preset threshold, it immediately triggers the rerouting mechanism, excludes the faulty link in the subsequent routing calculation, and re-executes the optimal path selection process to seamlessly switch the data stream to the new optimal path.

6. A digital system for traffic safety facilities based on intelligent connected vehicles according to claim 1, characterized in that, The generation of structured real-time data streams is achieved by preprocessing the collected data from different sources and in different formats, including timestamp alignment, coordinate system unification, and data format standardization, and by adding the facility identification identifier to the data, and finally encapsulating it into a unified format data packet for output to the upper layer.

7. A digital system for traffic safety facilities based on intelligent network connectivity as described in claim 1, characterized in that, The detailed steps for generating the facility association graph from the transportation facility dependency relationship grid are as follows: parsing the real-time data stream, combining it with the facility basic information database, quantifying the dependencies between facilities through a relationship reasoning engine, and constructing the transportation facility dependency relationship grid; The dependency mesh is mapped and transformed, each transportation facility is abstracted as a node, and the dependencies between facilities are abstracted as weighted directed edges connecting nodes, generating the facility association graph, which serves as the input object for the diffraction chain pre-simulation sand table.

8. A digital system for traffic safety facilities based on intelligent connected vehicles according to claim 1, characterized in that, The diffraction chain simulation sandbox is configured to: receive the facility association graph and a specified initial event node set as input; the diffraction chain simulation sandbox has a built-in graph traversal and deduction algorithm, which calculates and generates a risk diffraction chain data structure on the facility association graph, starting from the initial event node set, based on the weight and direction of the edges in the graph, including the affected node identifiers, risk propagation paths, and time series of the degree of impact; and based on the risk diffraction chain data structure, maps and combines the risk prevention and control strategy set from the intervention measures knowledge base.

9. A digital system for traffic safety facilities based on intelligent connected vehicles according to claim 1, characterized in that, The compensation control engine is configured to: receive the risk pre-control strategy set as input; the compensation control engine includes a strategy parser, which, based on the instruction mapping rule base, converts the strategies in the risk pre-control strategy set into device control instructions with explicit execution parameters; the compensation control engine also transmits the generated device control instructions to the target traffic safety facility through the heterogeneous fusion self-healing elastic network.

10. A digital system for traffic safety facilities based on intelligent connected vehicles according to claim 1, characterized in that, The compensation control engine continuously acquires status and environmental data of the affected area from the data acquisition module during and after the execution of the equipment control command. The compensation control engine has a built-in evaluation model that calculates quantitative performance indicators by comparing the status and environmental data before and after the command execution. Based on the quantitative performance indicators, the compensation control engine generates model update instructions and transmits the instructions to the strategy output module to adjust the generation logic of the risk pre-control strategy set.

Images