Full-color LED display screen intelligent centralized monitoring method and system

Abstract

The invention provides an intelligent centralized monitoring method for a full-color LED display screen, and the method comprises the steps: analyzing the dependency relationship and communication delay between nodes based on an adjacent node correlation model, and recognizing a target region with a network segmentation risk; standby communication path detection is carried out on the risk nodes, and paths with communication delay lower than a preset threshold value are screened as priority switching paths; a distributed consistency protocol is deployed for data difference comparison, and a synchronization sequence list is generated according to the data updating frequency and the service priority; dividing to-be-synchronized data into a plurality of data fragments by adopting a fragment transmission mechanism, and transmitting the data fragments to a target node in parallel through multiple paths; starting an automatic repair mechanism for inconsistent nodes, and acquiring backup data from adjacent nodes to execute differentiation correction; and dynamically optimizing a communication path weight and a synchronization period parameter according to a correction result, and completing iterative updating of the system state.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and particularly to an intelligent centralized monitoring system for full-color LED display screens. Background Art

[0002] As an important carrier for modern information display and intelligent interaction, full-color LED display screens play an indispensable role in fields such as urban management, commercial advertising, and public services. Their stability and reliability are directly related to the efficiency of information transmission. However, current management methods mostly rely on centralized control and manual maintenance, which are not only inefficient but also prone to causing the overall paralysis of the system due to single-point failures, especially in large-scale distributed scenarios, the problem is particularly prominent.

[0003] Existing solutions often are limited to simple fault alarms and passive repairs, lacking the ability to adapt to dynamic changes in complex network environments and unable to effectively handle the situation of multiple nodes failing simultaneously. This limitation makes the display screen system often experience slow responses or even data loss when facing sudden problems, seriously affecting the user experience.

[0004] Looking deeper, the core challenge in this field lies first in how to achieve efficient node coordination and management in a distributed environment. Due to the large number of nodes and their dispersed geographical locations, traditional static network structures are difficult to adapt to dynamic changes, resulting in the inability to adjust in a timely manner when some nodes fail, thereby triggering the problem of network partitioning. And network partitioning further exacerbates the difficulty of data synchronization because the data loss of failed nodes will destroy the integrity of overall monitoring. Traditional synchronization methods often require transmitting a large amount of redundant information, increasing the system burden. These interrelated problems make the stable operation of the system in high-load or complex scenarios a major pain point.

[0005] Therefore, how to achieve dynamic node management, rapid fault takeover, and efficient data synchronization in a distributed full-color LED display screen network has become a key issue in improving system reliability and intelligence. Summary of the Invention

[0006] The present invention provides an intelligent centralized monitoring method for full-color LED display screens, mainly including the following steps: Regularly obtain the operating status data of each node through a distributed data acquisition mechanism to generate a node status distribution map including connection status and performance indicators; Analyze the dependency relationship and communication delay between nodes based on a neighboring node association model to identify target areas at risk of network partitioning; Detect alternative communication paths for risk nodes and select paths with communication delays lower than a preset threshold as preferred switching paths; Deploy a distributed consistency protocol for data difference comparison, and generate a synchronization order list according to the data update frequency and business priority; Adopt a sharding transmission mechanism to divide the data to be synchronized into multiple data segments, and transmit them to the target node through multi-path parallel transmission; After receiving the status update information of each node, detect the data integrity through a status consistency verification model, and generate a status consistency report including abnormal nodes; Start an automatic repair mechanism for inconsistent nodes, and obtain backup data from neighboring nodes to perform differential correction; Dynamically optimize the communication path weight and synchronization period parameters according to the correction results to complete the iterative update of the system state.

[0007] Furthermore, when generating the middle node status distribution map, use the weighted round-robin algorithm to calculate the communication quality scores of each node, and the scores include comprehensive indicators of packet loss rate, signal strength, and historical failure rate.

[0008] Furthermore, the neighboring node association model adopts an improved PageRank algorithm to construct a dependency matrix by analyzing the data flow intensity and topological distance between nodes.

[0009] Furthermore, when detecting the alternate communication path, use path probe technology to measure the bandwidth utilization rate and link stability in real time, and establish a path quality evaluation model including delay jitter parameters.

[0010] Furthermore, the distributed consistency protocol adopts an improved version of the Raft algorithm to adapt to network topology changes by dynamically adjusting the election timeout mechanism.

[0011] Furthermore, the sharding transmission mechanism includes the following steps: Determine the sharding granularity reference value according to the business type label of the data to be synchronized; Dynamically adjust the sharding size in combination with the current transmission path quality, and automatically reduce the sharding volume when the network delay is higher than the preset threshold; Allocate high-priority paths for emergency service data and prohibit sharding to ensure transmission atomicity.

[0012] Furthermore, the status consistency verification model uses the Merkle tree structure to compare data digests, and sets a difference tolerance threshold to distinguish repairable anomalies and serious faults.

[0013] Furthermore, the automatic repair mechanism includes a three-level repair strategy: the first-level repair uses the cached data of adjacent nodes, the second-level repair calls the regional backup nodes, and the third-level repair starts the cloud image recovery.

[0014] Furthermore, the dynamic optimization process uses a reinforcement learning algorithm to obtain the optimal path selection strategy through training with a Q-learning model.

[0015] On the other hand, the present invention also provides an intelligent centralized monitoring system for full-color LED display screens, including: A status collection agent module, deployed at each LED display node, for executing the distributed data collection mechanism; A topology analysis server, configured to run the adjacent node association model and the network segmentation risk identification algorithm; A path manager, integrated into the central control unit, responsible for detecting and making switching decisions for backup communication paths; A distributed synchronization engine, including the shard transmission mechanism and the consistency protocol execution unit; A repair executor module, triggering a multi-level repair strategy in response to the status consistency report; Wherein each module conducts data interaction through a message queue to form a closed-loop control link.

[0016] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects: The present invention discloses an intelligent centralized monitoring method for full-color LED display screens. By periodically collecting the operating status data of each LED display screen node, generating a node status distribution map, analyzing the node dependency relationship to identify the network segmentation risk area, detecting backup communication paths, deploying a distributed consistency protocol for data synchronization, using a shard transmission mechanism to transmit data in parallel, verifying data integrity and automatically repairing inconsistent nodes, and dynamically optimizing communication parameters. The present invention uses a weighted round-robin algorithm to evaluate the node communication quality, improves the PageRank algorithm to construct a dependency relationship matrix, uses a path probe technology to evaluate the quality of backup paths, improves the Raft algorithm to adapt to network topology changes, formulates a shard strategy based on data characteristics, uses a Merkle tree structure to verify consistency, sets a three-level repair strategy, and applies reinforcement learning to optimize path selection. The present invention can achieve efficient status synchronization and fault self-healing of a large-scale distributed full-color LED display screen system, improving the reliability and performance of the large-scale distributed full-color LED display screen system Description of the Drawings

[0017] Figure 1 It is a flowchart of an intelligent centralized monitoring method for a full-color LED display screen of the present invention.

[0018] Figure 2 It is a structural diagram of an intelligent centralized monitoring system for a full-color LED display screen of the present invention. Detailed Embodiments

[0019] To make the objectives, technical solutions and advantages of the present invention more clear, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0020] As Figure 1 , a full-color LED display intelligent centralized monitoring method in this embodiment may specifically include: S101. When generating the node status distribution map, the weighted round-robin algorithm is used to calculate the communication quality scores of each node, and the scores include comprehensive indicators of packet loss rate, signal strength, and historical failure rate.

[0021] The core of the weighted round-robin algorithm lies in dynamically allocating node selection weights. In one embodiment, within a large-scale distributed full-color LED display cluster, the packet loss rate of node A is 2%, the signal strength is -70 dBm, and the historical failure rate is 0.5%. The corresponding parameters of node B are 5%, -85 dBm, and 1.2%. The algorithm assigns 40% weight to the packet loss rate, 30% to the signal strength, and 30% to the historical failure rate. Then the score of node A is 2×0.4 + 70×0.3 + 0.5×0.3 = 24.35, and the score of node B is 5×0.4 + 85×0.3 + 1.2×0.3 = 30.86. At this time, the system preferentially selects node A with a better score.

[0022] The packet loss rate index is realized through sliding window statistics. In one embodiment, the gateway in the LED display cluster samples 100 data packets every 10 seconds. If the number of lost packets in three consecutive windows is 3 / 5 / 4 respectively, the dynamic packet loss rate is calculated as (3 + 5 + 4) / 300 = 4%. This index reflects the real-time network congestion situation. A high packet loss rate will trigger a weight reduction to avoid selecting nodes with unreliable transmissions.

[0023] The signal strength adopts a hierarchical quantization strategy. The RSSI value is divided into 6 levels from -60 dBm to -120 dBm, with each 10 dBm as a grade. In one embodiment, the signal strength in the LED display cluster is divided as follows: -65 dBm corresponds to grade 1 (full score 100 points), -93 dBm corresponds to grade 4 (60 points). Through discretization processing, the scoring oscillation caused by signal fluctuations is avoided.

[0024] The historical failure rate introduces a time decay factor. Suppose an edge node has had 3 failures in the past 24 hours, but no failures in the recent 8 hours. Then the effective number of failures is calculated as 3×e^(-8 / 24) = 2.21 times according to the exponential decay formula, and the final failure rate is 2.21 / 24 ≈ 9.2%. This mechanism enables the system to pay more attention to recent stability and avoid the interference of historical data on current decisions.

[0025] The comprehensive indicators eliminate the dimension differences through normalization. In one embodiment, the packet loss rate of 0 - 10% is mapped to 0 - 100 points, the signal strength of -60 to -120 dBm is mapped to 100 - 0 points, and the failure rate of 0 - 5% is mapped to 100 - 0 points, and then they are superimposed according to the weights. This standardization method ensures the comparability of indicators in different dimensions and makes the scoring results have practical guiding significance.

[0026] The dynamic weight adjustment mechanism is used to handle emergencies. When the score of a certain node drops by more than 20% for two consecutive times, the system automatically increases the weight of the packet loss rate to 50%. For example, when there is a sudden electromagnetic interference in a certain node within the LED display screen cluster and the packet loss rate soars from 3% to 15%, the algorithm quickly reduces the priority of this node through weight offset to ensure the transmission reliability of control instructions.

[0027] S102. The adjacent node association model adopts an improved PageRank algorithm to construct a dependency matrix by analyzing the data flow intensity and topological distance between nodes.

[0028] The traditional PageRank algorithm calculates weights through web link relationships, while the improved version takes the data flow intensity (such as the number of bytes transmitted per second) and topological distance (such as the number of hops) as input parameters.

[0029] In one embodiment, node A transmits 1MB of data to node B per second and is two hops away, and node C transmits 500KB to node B but is one hop away. The algorithm will comprehensively consider the traffic size and the distance attenuation factor (such as a 20% attenuation per hop). Eventually, the dependency weight of B on A may be 0.64 (1MB × 0.8²), and on C may be 0.5 (500KB × 0.8), thereby constructing an asymmetric dependency matrix.

[0030] The specific measurement method of data flow intensity: The system statistically calculates the average traffic between nodes for 5 minutes through historical transmission logs and eliminates burst noise. In one embodiment, nodes X and Y record traffic of 800KB / s, 1.2MB / s, and 900KB / s respectively in three time windows. After sliding average processing, 950KB / s is taken as the stable flow intensity to ensure that the matrix parameters reflect the real business load.

[0031] The dynamic adjustment mechanism of topological distance: When the new node D shortens the original A - B path from two hops to one hop, the system triggers a topology update event and recalculates the dependency weights. In one embodiment, the new weight from A to B is increased from 0.64 to 0.8 (1MB × 0.8¹), and if D has no direct traffic with A, its weight is 0, and the corresponding elements in the matrix are updated in real time to adapt to network changes.

[0032] Technical effects of the dependency matrix: This matrix can identify key hub nodes (such as nodes with a total weight exceeding the threshold), and preferentially allocate resources during video stream fragmentation. In one embodiment, a certain matrix shows that node E is dependent on 8 nodes and the total weight reaches 5.6. The system will mark it as a high-priority fragmentation receiver, reducing the risk of cascading failures caused by congestion at E. Synergy with path quality assessment. High-weight links in the dependency matrix will be monitored by path probes. For example, for the A-B link with a weight of 0.8, the probe sampling frequency is increased from 10 seconds per time to 2 seconds per time, and the delay jitter data is synchronously fed back to the matrix to correct the weight, forming a closed-loop optimization.

[0033] S103. When detecting the backup communication path, the path probe technology is used to measure the bandwidth utilization rate and link stability in real time, and a path quality assessment model including delay jitter parameters is established.

[0034] Implementation method of the path probe technology: In the detection of the backup communication path, the path probe periodically sends probe data packets (such as sending 100-byte UDP packets every 5 seconds) to the target node, and records the round-trip delay (RTT) and packet loss rate. In one embodiment, when the RTT of the primary path exceeds 50 ms, the probe triggers the backup path assessment, and at the same time collects the response time of each hop in the path to form a link state snapshot. This technology identifies instantaneous jitter by comparing historical data (such as the average value in the past 10 minutes), avoiding misjudging the path quality due to single measurement errors.

[0035] Specific operation of bandwidth utilization measurement: The sliding window statistical method is used to calculate the ratio of link throughput every 30 seconds. In one embodiment, when it is detected that the bandwidth utilization rate of a certain path continuously exceeds 80% (the threshold is configurable) for 3 times, it is marked as a congested path. When measuring, the service priorities are distinguished. For example, video stream probe packets have higher priority than ordinary data packets to ensure the measurement accuracy of critical services.

[0036] Construction of the link stability assessment model: Calculate the link interruption frequency and recovery time based on probe data. In one embodiment, if a certain path has more than 3 delay fluctuations exceeding 200 ms within 1 hour, the stability score will be reduced by 50%. The model introduces an exponentially weighted moving average algorithm to make the weight of recent failure data higher (such as 70% in the past 1 hour), quickly reflecting topological changes.

[0037] Quantification processing of delay jitter parameters: The jitter value (such as standard deviation) is divided into 4 levels: 0 - 10 ms is excellent, 10 - 30 ms is good, 30 - 50 ms is medium, and above 50 ms is poor. The evaluation model combines the jitter level and duration (such as continuous 5 times of jitter being "medium" then degrading) to dynamically adjust the path weight.

[0038] Decision Logic of Path Quality Assessment Model: The total score is generated by comprehensively considering bandwidth (weight 40%), stability (weight 30%), and delay jitter (weight 30%). In one embodiment, when the total score of the primary path is below 60 points, it will automatically switch to the backup path (the total score of the backup path needs to be ≥ 75 points). The model sets a hysteresis threshold (5-point difference) to prevent frequent switching and dynamically optimizes the weight allocation through reinforcement learning. Through real-time probes and multi-dimensional evaluations, the system can identify path degradation faster than traditional heartbeat detection.

[0039] S104. The distributed consensus protocol adopts an improved version of the Raft algorithm and adapts to network topology changes by dynamically adjusting the election timeout mechanism.

[0040] When the distributed consensus protocol adopts an improved version of the Raft algorithm, the core purpose of dynamically adjusting the election timeout mechanism is to solve the election conflicts or delays caused by the fixed timeout time in traditional Raft when the network topology changes frequently. The improved algorithm dynamically calculates the election timeout threshold for each node by real-time monitoring of the communication delay between nodes and the topology changes. In one embodiment, when it is detected that the average round-trip delay between a certain node and the Leader node increases from 100ms to 300ms, the election timeout time of this node will be adjusted from the original range of 150ms - 300ms to 300ms - 600ms to avoid invalid elections triggered by network fluctuations.

[0041] The specific implementation of dynamic adjustment depends on the link quality data periodically collected by the topology awareness module. This module measures the delay jitter and packet loss rate between nodes through heartbeat packets. When the packet loss rate between nodes in a certain area exceeds 5%, it is determined as an unstable topology state. At this time, the Raft algorithm will increase the election timeout baseline value of the affected nodes by 50%. In one embodiment, the average delay between node A and most nodes in the cluster is 200ms ± 50ms. When it is detected that the delay suddenly increases to 500ms ± 200ms and the packet loss rate reaches 8%, its election timeout will be adjusted from the default 150ms - 300ms to 225ms - 450ms, thereby reducing the risk of multiple Leaders caused by network partitioning.

[0042] The improved algorithm also introduces the prediction of historical topology change trends. By analyzing the Markov chain model of the node connection status in the past 10 cycles, the stability probability of the next cycle is predicted. If it is predicted that a certain node has a 70% probability of entering a high-delay state, its election timeout time will be increased by 20% in advance. In one embodiment, node B has experienced delay fluctuations above 200ms in 3 consecutive cycles. The system predicts that the probability of its instability in the next cycle is 75%. Therefore, the timeout time of this node is adjusted from 200ms to 240ms to reduce the election storm caused by out-of-prediction timeouts.

[0043] The technical effects are reflected in the following three aspects: First, dynamic adjustment makes the election timeout match the current network condition, greatly shortening the unavailable time of the cluster. Second, by setting different timeout thresholds for each node, it avoids a large number of nodes initiating votes simultaneously, reducing bandwidth competition. Finally, the prediction mechanism enables the system to be forward-looking about topology changes, reducing ineffective Leader switches.

[0044] S105. The shard transmission mechanism includes the following steps: Determine the shard granularity reference value according to the service type label of the data to be synchronized; Dynamically adjust the shard size in combination with the current transmission path quality, and automatically reduce the shard volume when the network delay is higher than the preset threshold; Allocate a high-priority path for emergency service data and prohibit sharding to ensure transmission atomicity.

[0045] Extract the corresponding service priority and data characteristics according to the service type label of the data to be synchronized, and determine the reference value of the shard granularity. In one embodiment, video stream data adopts a sharding rule with key frames prioritized, sets the reference shard size to 1MB, and control instructions adopt a whole-packet transmission mode with the reference shard size being the size of the complete data packet. By collecting the network parameters of the current transmission path in real time, analyze the network delay and bandwidth utilization rate. In one embodiment, when the packet loss rate exceeds 5% or the signal strength is lower than -80dBm, trigger the shard adjustment mechanism to obtain the dynamic adjustment basis for the shard size. If the network delay is higher than the preset threshold (such as 100ms), trigger the shard volume reduction strategy, adjust the shard size to 50% of the original reference value, calculate the adjusted shard size, and determine a new shard scheme. According to the service type label and priority identifier, screen out emergency service data, such as data marked as real-time control instructions, and obtain its transmission path allocation requirements. Judge the availability of the high-priority path. If the high-priority path meets the stability requirements, allocate the emergency service data to this path, such as selecting the node path with the highest communication quality score to determine the transmission channel. According to the atomicity requirements of the emergency service data, set the sharding prohibition rule, such as directly adopting the whole-packet transmission mode for control instruction data to obtain the transmission mode of the complete data packet.

[0046] S106. The state consistency verification model uses a Merkle tree structure to compare data digests and sets a difference tolerance threshold to distinguish reparable anomalies and serious faults.

[0047] The Merkle tree aggregates data blocks layer by layer into a tree structure through a hashing algorithm, and each non-leaf node stores the combined hash of the hash values of its child nodes. In one embodiment, after dividing 1GB of data into 1024 1MB data blocks, the SHA-256 hash value is calculated for each block. The hash values of two adjacent blocks are concatenated and hashed again, recursively forming the root hash. When node A and node B perform data calibration, only the root hash value needs to be compared to quickly discover differences. If they are inconsistent, the specific different blocks are located layer by layer downward, reducing the complexity of traditional full-scale comparison from O(n) to O(log n).

[0048] Dynamic setting method for the difference tolerance threshold: The system defines two levels of thresholds according to the service type. The tolerance threshold for sensor data of LED display nodes is set to 0 (zero difference), while for streaming media data, a 3% block hash mismatch is allowed. In one embodiment, when 2% of the leaf node hashes are found to be inconsistent during the Merkle tree comparison between nodes, the system determines it as a repairable exception and automatically triggers the data synchronization of adjacent nodes; if the difference exceeds 5%, it is determined as a serious fault and a three-level repair mechanism is started. The threshold parameters are dynamically adjusted through indicators such as the historical network packet loss rate and node stability to avoid misjudgment.

[0049] Determination logic for repairable exceptions: When the Merkle tree detects that the middle-layer hashes of a certain video segment are inconsistent, but the hashes of the branches where the key frames are located match, the system only retransmits the different segments. In one embodiment, the hash value of the left branch of the third layer of the Merkle tree of node A is X1, and the corresponding branch of node B is X2, but the hash values of the branches where the key frames are located for both are Y. At this time, the system marks this segment as "partially damaged", preferentially synchronizes the non-key frame data, and maintains the playback continuity.

[0050] Disposal process for serious faults: If the Merkle tree root hash of a certain node does not match 90% of the nodes in the cluster, and the different blocks are distributed in multiple subtrees, the system determines it as a serious fault. In one embodiment, it is detected that the storage segments 1, 3, and 7 of node C simultaneously have hash breaks. The node is immediately isolated, a complete data mirror is pulled from the regional backup node, and the topology routing table is updated to avoid this node at the same time.

[0051] After adopting the Merkle tree structure, the system can quickly locate differences from a large amount of data in an extremely short time, and the efficiency is greatly improved compared with the traditional CRC check. The difference tolerance threshold mechanism enables the system to ensure strong consistency of node sensor data while ensuring eventual consistency for other non-critical data to reduce resource consumption.

[0052] S107. The automatic repair mechanism includes a three-level repair strategy: the first-level repair uses the cached data of adjacent nodes, the second-level repair calls the regional backup node, and the third-level repair starts the cloud mirror recovery.

[0053] First-level repair uses adjacent node cached data: This strategy is based on the principle of local data redundancy. When data anomalies are detected in node A, the system preferentially obtains the most recently synchronized data copies from the caches of its directly connected nodes B and C. In one embodiment, when node A loses a video stream shard, the system will query the key frame data (such as I frames encoded in H.264) stored in the cache of node B within the most recent 3 seconds, and select the latest version for overwrite repair by comparing timestamps. This stage is designed for millisecond-level response (usually <50ms). Since it only involves single-hop communication and the data volume is small (such as a single key frame is about 5KB), it can avoid cross-region transmission overhead. It can quickly repair high-frequency small-scale anomalies (such as packet loss caused by instantaneous network jitter), while reducing dependence on upper-level nodes.

[0054] Second-level repair invokes regional backup nodes: When the caches of adjacent nodes are incomplete (such as both nodes B and C are missing key frames), the system will send a request to the backup node D in the same region (such as the same control computer room or subnet). For example, in the scenario of video stream shard repair, node D stores the complete data within a time window (such as the GOP group in the most recent 10 seconds), and locates the different shards by comparing the Merkle tree root hash value and then transfers them. The delay in this stage is controlled within 200ms, which is suitable for medium-scale data corruption (such as node A losing 20% of its shards). The regional backup node adopts a periodic snapshot mechanism (such as generating incremental backups every 5 seconds) to balance storage overhead and recovery efficiency. This strategy trades space for time to solve the problem of local multi-node chain failures.

[0055] Third-level repair initiates cloud mirror restoration: For global failures (such as the regional backup node crashing), the system pulls the complete mirror from cloud storage. For example, when node A loses all video stream data due to a hard disk failure, the cloud mirror provides the ability to restore by time point (such as a timestamp index accurate to the second level), and is accelerated by CDN edge nodes for transmission. This stage takes a relatively long time (about 2 - 5 seconds), but can ensure the ultimate data consistency. The cloud mirror adopts a hierarchical storage design, with hot data (the most recent 1 hour) stored in SSDs and cold data archived to object storage. As a fallback solution, the technical value of third-level repair lies in maintaining the robustness of the system when dealing with catastrophic failures.

[0056] Collaborative logic of the three-level strategy: The system triggers strategy upgrades through difference thresholds. For example, when the detected data difference rate <5%, only first-level repair is initiated; when the difference rate is 5% - 30%, second-level repair is enabled; when the difference rate >30%, it directly jumps to third-level repair. This hierarchical mechanism realizes the optimal allocation of resources by dynamically evaluating the scope of the fault impact (such as single node / multi-node / global) and repair costs (bandwidth, latency, computing resources). For example, in a combined LED display system when playing a video, first-level repair ensures lip-sync, second-level repair handles local stuttering, and third-level repair deals with service interruptions, forming a fault-tolerant system covering the entire scenario.

[0057] S108. The dynamic optimization process adopts a reinforcement learning algorithm, and an optimal path selection strategy is obtained through training with a Q-learning model.

[0058] Q-learning is a model-free reinforcement learning algorithm, and its core is to learn the optimal strategy by iteratively updating the Q-value table. In the path selection scenario, the system regards network nodes as states, alternative paths as actions, and metrics such as the delay and bandwidth of the path as rewards. In one embodiment, when node A needs to transmit data to node B, the Q-learning model will evaluate all possible paths (such as A→C→B or A→D→E→B), and calculate the Q-value of each path based on historical data.

[0059] In the initial stage, the Q-values are randomly initialized. The model gradually updates the Q-values by balancing exploration (randomly selecting paths) and exploitation (selecting paths with known high Q-values). In one embodiment, assume that the initial Q-value of the path A→C→B is 0, the measured delay for the first transmission is 50 ms, the reward is set to -50, the learning rate α = 0.1, and the discount factor γ = 0.9. Then the Q-value is updated to: 0 + 0.1×(-50 + 0.9×0 - 0) = -5. After multiple iterations, the model will converge to the optimal path strategy.

[0060] To adapt to the dynamic network environment, the Q-learning model needs to sense the changes in the network state in real time. In one embodiment, when node C is congested, the delay of the path A→C→B increases from 50 ms to 200 ms, and the reward is adjusted to -200. The model will quickly reduce the Q-value of this path and instead select the alternative path A→D→E→B. This process is achieved through periodic detection. For example, the link state data is updated every 5 seconds. In one embodiment, the model introduces the ε-greedy strategy (such as ε = 0.1), selects the current optimal path in 90% of the cases, and randomly explores new paths with a 10% probability to avoid getting stuck in local optima.

[0061] The dynamic optimization ability of Q-learning is significantly better than that of static routing algorithms. For example, in the scenario of bursty traffic, the traditional OSPF protocol needs to synchronize the link state database across the network, and the convergence time may reach the second level, while the Q-learning model only needs local Q-value updates and can complete path switching in milliseconds. At the same time, the model can flexibly optimize different metrics through the design of the reward function. For example, by setting the reward to "-delay + 0.5×bandwidth", the requirements for transmission speed and throughput can be balanced.

[0062] such as Figure 2As shown, an embodiment of the present invention also discloses a full-color LED display intelligent centralized monitoring system based on any of the above embodiments, including: A status collection agent module, deployed at each LED display node, for executing the distributed data collection mechanism; A topology analysis server, configured to run the adjacent node association model and the network segmentation risk identification algorithm; A path manager, integrated in the central control unit, responsible for the detection and switching decision of the standby communication path; A distributed synchronization engine, including the shard transmission mechanism and the consistency protocol execution unit; A repair executor module, triggering a multi-level repair strategy in response to the status consistency report; Among them, each module performs data interaction through a message queue to form a closed-loop control link.

[0063] The above are only the preferred embodiments of the present invention, and do not limit the patent scope of the present invention accordingly. Any equivalent structure or equivalent process transformation made by using the content of the specification and drawings of the present invention, or directly or indirectly applied in other related technical fields, shall be equally included in the patent protection scope of the present invention.

Claims

1. An intelligent centralized monitoring method for a full-color LED display screen, characterized in that, It includes the following steps: Regularly obtain the operation status data of each node through a distributed data collection mechanism, and generate a node status distribution map including connection status and performance metrics; Analyze the dependency relationship and communication delay between nodes based on the adjacent node association model, and identify the target areas at risk of network segmentation; Detect alternative communication paths for risk nodes, and select the paths with communication delay lower than the preset threshold as the preferred switching paths; Deploy a distributed consistency protocol to compare data differences, and generate a synchronization sequence list according to the data update frequency and service priority; Use a sharding transmission mechanism to divide the data to be synchronized into multiple data segments, and transmit them to the target node in parallel through multiple paths; After receiving the status update information of each node, detect data integrity through a status consistency verification model, and generate a status consistency report including abnormal nodes; Start an automatic repair mechanism for inconsistent nodes, and obtain backup data from adjacent nodes to perform differential correction; Dynamically optimize the communication path weight and synchronization period parameters according to the correction results to complete the iterative update of the system state.

2. The intelligent centralized monitoring method for full-color LED display screens according to claim 1, wherein When generating the node status distribution map in step S1, a weighted round-robin algorithm is used to calculate the communication quality score of each node, and the score includes comprehensive indicators such as packet loss rate, signal strength, and historical failure rate.

3. The intelligent centralized monitoring method for full-color LED display screens according to claim 1, characterized in that, The adjacent node association model uses an improved PageRank algorithm to construct a dependency relationship matrix by analyzing the data flow intensity and topological distance between nodes.

4. The intelligent centralized monitoring method for full-color LED display screens according to claim 1, wherein When detecting alternative communication paths, path probe technology is used to measure the bandwidth utilization rate and link stability in real time, and a path quality evaluation model including delay jitter parameters is established.

5. The intelligent centralized monitoring method for full-color LED display screens according to claim 1, characterized in that, The distributed consistency protocol uses an improved version of the Raft algorithm to adapt to network topology changes by dynamically adjusting the election timeout mechanism.

6. The intelligent centralized monitoring method for full-color LED display screens according to claim 1, characterized in that, The sharding transmission mechanism includes the following steps: Determine the sharding granularity reference value according to the service type label of the data to be synchronized; Dynamically adjust the sharding size in combination with the current transmission path quality, and automatically reduce the sharding volume when the network delay is higher than the preset threshold; Allocate high-priority paths for emergency service data and prohibit sharding to ensure transmission atomicity.

7. The intelligent centralized monitoring method for full-color LED display screens according to claim 1, wherein The status consistency verification model uses a Merkle tree structure to compare data digests, and sets a difference tolerance threshold to distinguish repairable anomalies and serious faults.

8. The intelligent centralized monitoring method for full-color LED display screens according to claim 7, characterized in that, The automatic repair mechanism includes a three-level repair strategy: The first-level repair uses the cache data of adjacent nodes, the second-level repair calls the regional backup nodes, and the third-level repair starts the cloud mirror recovery.

9. The intelligent centralized monitoring method for full-color LED display screens according to claim 1, characterized in that, The dynamic optimization process uses a reinforcement learning algorithm to obtain the optimal path selection strategy through Q-learning model training.

10. An intelligent centralized monitoring system for a full-color LED display screen based on any one of the methods of claims 1-9, characterized in that, It includes: A status collection agent module, deployed on each LED display node, for executing the distributed data collection mechanism; A topology analysis server, configured to run the adjacent node association model and the network segmentation risk identification algorithm; A path manager, integrated in the central control unit, responsible for the detection and switching decision of alternative communication paths; A distributed synchronization engine, including the sharding transmission mechanism and the consistency protocol execution unit; A repair executor module, triggering a multi-level repair strategy in response to the status consistency report; Among them, each module conducts data interaction through a message queue to form a closed-loop control link.

Images