A cloud-edge collaboration based intelligent conference screen real-time collaboration system and method

Abstract

The application discloses a kind of wisdom meeting screen real-time cooperation systems and methods based on cloud edge cooperation, it is related to wisdom meeting equipment technical field, the method includes: when edge node detects that cooperation interruption occurs, the duration grade of pre-judgment and the grade of cooperative urgency are judged, and interruption pre-judgment matrix is generated;Dynamic allocation of cache resource is executed and data preheating for rapid recovery is executed, and preheating strategy label containing the executed strategy identification is generated;Cloud end receives the interruption pre-judgment matrix and preheating strategy label of each edge node, constructs global cooperative stress thermodynamic diagram, and issues resource coordination and behavior optimization instruction to relevant node;When interruption recovery condition is triggered, edge node is positioned based on preheating strategy label local cache analysis state, quickly compares local cache analysis state with the latest global state in cloud end, generates state difference influence report, and executes fusion strategy.

Description

Technical Field

[0001] This invention relates to the field of smart meeting equipment technology, specifically a real-time collaboration system and method for smart meeting screens based on cloud-edge collaboration. Background Technology

[0002] With the widespread adoption of smart conference screens, real-time collaboration based on cloud-edge integration has become the mainstream model. However, in complex scenarios such as unstable wireless networks and frequent device switching, collaborative sessions between edge nodes (conference screens) and the cloud or other nodes are prone to interruption. Traditional interruption recovery methods have a single recovery strategy, typically employing simple data retransmission or full synchronization, failing to differentiate based on the cause of the interruption and user scenario, resulting in low recovery efficiency or disrupted user experience. Furthermore, edge nodes and the cloud lack intelligent collaboration during the recovery process; the edge often passively waits for cloud instructions, while the cloud struggles to promptly perceive the specific interruption status and recovery urgency of each edge node, hindering global resource coordination and easily triggering network congestion or data conflicts during recovery. Existing solutions often focus on data consistency, neglecting the continuity of user interaction during the recovery process; users often have to wait for data synchronization to complete before continuing, disrupting the flow of thought during the meeting. Summary of the Invention

[0003] The purpose of this invention is to provide a real-time collaboration system and method for smart conference screens based on cloud-edge collaboration, so as to solve the problems raised in the prior art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a real-time collaboration method for smart meeting screens based on cloud-edge collaboration, the method comprising: S1. When an edge node detects a collaboration interruption, it performs parallel predictions on the expected duration of the interruption and the collaboration urgency level after the interruption is restored, and generates an interruption prediction matrix. S2. Based on the interruption prediction matrix, the edge node performs dynamic allocation of cache resources and data preheating for rapid recovery, and generates a preheating strategy label containing the identifier of the executed strategy. S3. The cloud receives the interruption prediction matrix and the preheating strategy label from each edge node, constructs a global collaborative pressure heat map, and issues resource coordination and behavior optimization instructions to relevant nodes. S4. When the interruption recovery condition is triggered, the edge node locates the analysis status of the local cache based on the preheating strategy label, quickly compares the analysis status of its local cache with the latest global status in the cloud, and generates a status difference impact report. S5. The edge node executes the fusion strategy based on the coordination urgency level in the interruption prediction matrix and the state difference impact report.

[0005] According to the above scheme, step S1 includes: The edge node determines the expected duration of the interruption as short-term, medium-term, or long-term based on the trigger type of the interruption event and the associated running status parameters. The short-term level indicates that the interruption is expected to recover within a first time threshold, the medium-term level indicates that the expected recovery time is between the first and second time thresholds, and the long-term level indicates that the expected recovery time exceeds the second time threshold. Based on the user interaction state and collaborative task state before the interruption, the edge node predicts the urgency of collaboration after the interruption is restored as high urgency, medium urgency, or low urgency. The high urgency level indicates that the user was in an active, high-frequency interaction state or undertaking a key collaborative task before the interruption; the medium urgency level indicates that the user was in a moderately active interaction or task state before the interruption; and the low urgency level indicates that the user was in a passive viewing or low-activity state before the interruption. The interruption prediction matrix is ​​composed of the expected duration level and the coordination urgency level.

[0006] According to the above scheme, the triggering types of the interruption event include network connection abnormality, device state switching, or application abnormality; the operating status parameters include network signal attenuation rate, historical average device switching time, or application process abnormal stack depth; the network connection abnormality specifically refers to the wireless network signal strength being lower than a preset threshold or the transmission control protocol connection timing out continuously; the device state switching specifically refers to user operation causing the conference screen to switch from the current display and computing main device to a slave device; the application abnormality specifically refers to the core process of the smart conference application crashing or service becoming unresponsive; The user interaction status includes whether the edge node was executing a collaborative task issued by the cloud before the interruption; the collaborative task issued by the cloud includes, but is not limited to, real-time annotation of specified document areas, transcription of the voice content of a specific speaker, or verification of data conclusions submitted by other nodes. The collaborative task status includes the priority tags of the tasks processed by the edge nodes before the interruption. The priority tags are assigned by the cloud during the collaboration process and are used to identify the criticality and processing order of the tasks in the global collaborative process.

[0007] According to the above scheme, step S2 includes: The edge node selects and executes a target caching strategy from the caching strategies based on the interruption prediction matrix. The identifier of the target caching strategy constitutes the first part of the warm-up strategy label. The caching strategy defines the retention rules for collaborative data in local memory and storage space. The identifier is used to uniquely distinguish different caching strategies. The edge node performs a data preheating action based on the interruption prediction matrix. The type identifier of the data preheating action constitutes the second part of the preheating strategy label. The data preheating action is to prepare the data required for recovery or to establish a connection channel in advance during the interruption.

[0008] According to the above scheme, when the interruption prediction matrix corresponds to a combination of short-term interruption and high urgency, the edge node selects the extreme retention strategy; when the interruption prediction matrix corresponds to a combination of long-term interruption and high urgency, the edge node selects the key anchor point caching strategy; for other combinations, the edge node selects the intelligent summarization caching strategy. The extreme retention strategy refers to prioritizing and completely retaining the user's interaction sequence at the last moment before the interruption and related incomplete analysis intermediate states within a limited cache space; the key anchor point caching strategy refers to caching only the user's original, non-reproducible logical conclusions and core annotation semantics, abandoning reconstructable complete data; the intelligent summarization caching strategy refers to extracting and compressing features from collaborative content and analysis status, retaining only structured summaries and key feature vectors. When the expected duration level is predicted to be long-term, the data preheating action includes sending a request to the cloud, pre-generating a state difference data packet, and attaching the preheating strategy label to the request. The state difference data packet refers to the incremental data set that the cloud predicts the possible changes during the interruption based on the preheating strategy label and pre-calculates, which can be quickly applied when the edge node recovers.

[0009] According to the above scheme, step S3 includes: The cloud receives interruption status information from each edge node. The interruption status information includes the expected duration level of the interruption, the coordination urgency level, and the coordination relationship between the nodes. The coordination relationship is defined by the coordination task dependency graph, which represents the dependency between nodes on data flow or control flow. The cloud analyzes the interruption prediction matrix of each edge node. If multiple nodes with the same high urgency level and located on the same collaborative link are detected, the link is marked as a high-pressure collaborative region. The cloud analyzes the preheating strategy labels of each edge node. If a node's strategy is detected as an extreme retention strategy, the pressure level of the high-pressure collaborative region corresponding to that node is increased. High-pressure collaborative regions and high-pressure recovery points are identified globally, and a global collaborative pressure heatmap is constructed based on the identification results. A high-pressure recovery point refers to an edge node in the high-pressure collaborative region whose task is blocked due to its reliance on the output of the interrupted node. The global collaborative pressure heatmap, in the form of visualization or data structure, represents the congestion risk and resource demand intensity faced by different collaborative links during the recovery phase. Based on the global collaborative pressure heatmap, the cloud sends resource coordination instructions to the interrupted nodes. These instructions include instructions for caching resource allocation adaptation and instructions for reserving the necessary computing power and bandwidth resources for recovery. It also sends behavior optimization instructions to edge nodes that are collaboratively associated with the interrupted nodes. These instructions include instructions for suspending modifications to non-core collaborative data and instructions for controlling the frequency of collaborative data updates. The update frequency is dynamically adjusted based on the preheating strategy label of the associated node. Dynamic adjustment includes reducing the update frequency if the preheating strategy label of the associated node indicates that it adopts a maximum retention strategy; and allowing or increasing the update frequency if it indicates that it adopts a key anchor or intelligent summary strategy.

[0010] According to the above scheme, step S4 includes: When the interruption recovery condition is triggered, the edge node determines the storage structure and integrity of the local cache analysis status based on the preheating strategy label; the storage structure includes whether the cached data is organized in complete time sequence, logical summary or key anchor mode; the integrity includes the degree to which the cached data covers the original collaborative context and logical structure. The edge node sends a request to the cloud to obtain the latest global status of the cloud. The latest global status of the cloud includes the global collaborative analysis status and the current collaborative status of each associated edge node. The global collaborative analysis status includes a unified view of the collaborative progress of all online nodes maintained by the cloud. The current collaborative status refers to the latest task progress, data version and interaction results of each node. Based on the storage structure, the edge node determines the comparison granularity and focus range for comparing with the latest global state in the cloud; based on the completeness, it determines the confidence assessment of the difference range when generating the state difference impact report. The comparison granularity and focus range are determined according to the storage structure. If it is a limit-retention structure, a fine-grained complete comparison is performed; if it is a summary or anchor structure, a key logic consistency comparison is performed. The confidence assessment is used to quantify the credibility of the state difference impact report.

[0011] According to the above scheme, step S5 includes: The edge node acquires the collaboration urgency level and extracts the difference range and impact degree from the state difference impact report. Based on the collaboration urgency level, difference range, and impact degree, a target fusion strategy is selected from the fusion strategies. The difference range refers to the breadth of data or logical modules where state differences occur. The impact degree refers to the potential damage level caused by the difference to the continuity and correctness of the collaborative task after recovery. The edge nodes perform a fusion operation on the differences between the local state and the cloud state based on the target fusion strategy.

[0012] According to the above scheme, selecting the target fusion strategy from the fusion strategies includes: When the collaboration urgency level is high urgency level, if the degree of difference is lower than a first difference threshold, a first fusion mode is selected and executed; if the degree of difference is higher than or equal to the first difference threshold, a second fusion mode is selected and executed; when the collaboration urgency level is medium urgency level, the granularity and order of fusion are determined based on the amount of difference data or conflict complexity associated with the degree of difference, and a third fusion mode is selected and executed; when the collaboration urgency level is low urgency level, a fourth fusion mode is selected and executed; the first difference threshold includes: if the target caching strategy is a limit retention strategy, the first difference threshold is set to a first value; if the target caching strategy is a key anchor point caching strategy, the first difference threshold is set to a second value; if the target caching strategy is an intelligent summarization caching strategy, the first difference threshold is set to a third value; wherein, the first value is greater than the second value, and the second value is greater than the third value; The more aggressive the caching strategy adopted in the early stage, the higher the tolerance for differences encountered during recovery; the more streamlined the caching strategy in the early stage, the more sensitive it is to differences, and the more likely it is to require user intervention for confirmation. The first fusion mode aims for real-time interactive continuity, allowing the local state to take effect immediately as the interaction baseline, and completing the asynchronous fusion of difference data in the background; the second fusion mode aims to quickly resolve key conflicts, by presenting a merged view focusing on core differences to the user and obtaining confirmation, and performing selective fusion; the third fusion mode aims to restore efficiency, by fusion of difference data in batches and progressively according to preset business logic priorities; the fourth fusion mode aims to ensure data integrity and consistency, by uniformly updating the local state after ensuring that all difference data has been synchronized and verified. Asynchronous fusion includes synchronizing and merging data in a background thread without interrupting the user's current interface operation; selective fusion includes merging only the core differences confirmed by the user, while the rest are processed according to preset rules; progressive fusion includes merging the differences into the local state in batches according to the importance or dependency of the data, so that users can work in advance based on the partially merged context.

[0013] A real-time collaboration system for smart meeting screens based on cloud-edge collaboration, comprising: an interruption prediction module, a warm-up strategy module, a cloud scheduling module, a report generation module, and a fusion execution module; The interruption prediction module is deployed on edge nodes. When a collaboration interruption is detected, it generates an expected duration level and a collaboration urgency level based on the interruption event trigger type, associated running status parameters, user interaction status, and collaboration task status, and constructs an interruption prediction matrix. The preheating strategy module is deployed at the edge node and connected to the interruption prediction module. It is used to receive the interruption prediction matrix and select a target caching strategy according to a specific combination of duration and urgency in the interruption prediction matrix; execute the corresponding data preheating action according to the duration level, and encode the identifier of the target caching strategy and the data preheating action into a preheating strategy label. The cloud scheduling module, deployed on a cloud server, communicates with each edge node. It receives and analyzes the interruption prediction matrix and preheating strategy labels of each node. The analysis specifically includes identifying high-urgency collaborative links as high-pressure areas based on the prediction matrix, adjusting the pressure level of the high-pressure areas according to the strategy identifier in the preheating strategy label, constructing a global collaborative pressure heat map based on this, generating and issuing resource coordination instructions and behavior optimization instructions to relevant nodes. The parameters of the behavior optimization instructions are dynamically adjusted according to the preheating strategy labels of the associated nodes. The report generation module is deployed on the edge node and connected to the preheating strategy module. It is used to determine the comparison granularity and confidence level based on the caching strategy and data integrity indicated by the preheating strategy label during interruption recovery, compare the local cache status with the global status in the cloud, and generate a status difference impact report. The fusion execution module is deployed on an edge node and is connected to the interruption prediction module and the report generation module, respectively. It is used to extract the collaborative urgency level in the interruption prediction matrix and analyze the degree of difference in the state difference impact report. The first difference threshold used for decision-making is dynamically set according to the target caching strategy type selected by the preheating strategy label generation unit, and the corresponding fusion mode is selected and executed based on the urgency level, the degree of difference and the dynamic threshold.

[0014] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention uses a dual-dimensional parallel prediction mechanism to intelligently assess the interruption duration and recovery urgency when an interruption occurs, and dynamically selects differentiated caching, preheating, and fusion strategies based on this. While ensuring data consistency, it prioritizes the continuity of user interaction in high-urgency scenarios, greatly shortens the effective recovery time, and improves the smoothness of collaboration. 2. This invention actively reports semantically rich interruption prediction matrix and preheating strategy labels through edge nodes. Based on this, the cloud constructs a global collaborative pressure heat map, realizing the accurate reservation and allocation of recovery resources, avoiding resource contention and network storms during the recovery period, and improving the overall stability and efficiency of the system in large-scale collaborative scenarios. 3. This invention achieves differentiated fusion through interruption recovery by dynamically setting thresholds based on collaboration urgency, degree of difference, and caching strategy, and adaptively selecting the corresponding fusion mode. This ensures both the continuity of real-time interaction in high-urgency scenarios and the integrity and consistency of data in low-urgency scenarios, thus adapting to the multi-scenario collaboration needs of smart meeting screens. Attached Figure Description

[0015] Figure 1 The present invention provides a step-by-step flowchart of a real-time collaboration method for a smart meeting screen based on cloud-edge collaboration. Figure 2 This is a schematic diagram of the structure of a real-time collaboration system for a smart meeting screen based on cloud-edge collaboration according to the present invention. Detailed Implementation

[0016] 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.

[0017] Example: Figures 1-2 As shown, this invention provides a technical solution: a real-time collaboration method for smart meeting screens based on cloud-edge collaboration. The method includes the following steps: S1. When an edge node detects a collaboration interruption, it performs parallel predictions on the expected duration of the interruption and the collaboration urgency level after the interruption is restored, and generates an interruption prediction matrix. S2. Based on the interruption prediction matrix, the edge node performs dynamic allocation of cache resources and data preheating for rapid recovery, and generates a preheating strategy label containing the identifier of the executed strategy. S3. The cloud receives the interruption prediction matrix and the preheating strategy label from each edge node, constructs a global collaborative pressure heat map, and issues resource coordination and behavior optimization instructions to relevant nodes. S4. When the interruption recovery condition is triggered, the edge node locates the analysis status of the local cache based on the preheating strategy label, quickly compares the analysis status of its local cache with the latest global status in the cloud, and generates a status difference impact report. S5. The edge node executes the fusion strategy based on the coordination urgency level in the interruption prediction matrix and the state difference impact report.

[0018] In this embodiment, taking the edge node smart conference screen as an example, a network connection anomaly is detected at time t=10:00:00, triggering a collaboration interruption, where t represents the absolute time point at which the interruption is detected; Specifically, step S1 includes: The edge node determines the expected duration of the interruption as short-term, medium-term, or long-term based on the trigger type of the interruption event and the associated running status parameters. The short-term level indicates that the interruption is expected to recover within a first time threshold, the medium-term level indicates that the expected recovery time is between the first and second time thresholds, and the long-term level indicates that the expected recovery time exceeds the second time threshold. Specifically, an abnormal network connection is detected, which is manifested by the Wi-Fi signal strength RSSI dropping sharply from -50dBm to -85dBm within 3 seconds and remaining below the preset threshold Th_net=-80dBm; where Th_net represents the signal strength threshold that triggers the network abnormality judgment. Network signal attenuation rate v; calculate the Wi-Fi signal strength RSSI change in the last second v=(RSSI_t-RSSI_(t-1)) / Δt; where RSSI_t represents the signal strength at the current moment, RSSI_(t-1) represents the signal strength at the previous moment, and Δt represents the time interval; In this embodiment, the prediction logic is as follows: The system records the recovery time distribution of the past 100 similar interruptions; the calculated v=-25dBm / s falls into the historical data, and the probability of the attenuation rate range [-30,-15]dBm / s for a medium-duration interrupt is 70%; therefore, the predicted duration of this interruption is predicted to be of medium duration. The preset time thresholds for short, medium, and long duration are: T_short=5 seconds, T_long=30 seconds; T_short and T_long are two time thresholds that distinguish between short and medium duration and medium and long duration, respectively; based on network and device reliability engineering data and user experience analysis, such as T_short being set as the limit time when the user does not perceive obvious lag, and T_long being set as the time threshold when the user begins to feel obvious waiting and may perform other operations, this is only an example and is not a limitation. Based on the user interaction state and collaborative task state before the interruption, the edge node predicts the urgency of collaboration after the interruption is restored as high urgency, medium urgency, or low urgency. The high urgency level indicates that the user was in an active, high-frequency interaction state or undertaking a key collaborative task before the interruption; the medium urgency level indicates that the user was in a moderately active interaction or task state before the interruption; and the low urgency level indicates that the user was in a passive viewing or low-activity state before the interruption. Specifically, within the 5-second time window before the interruption, user operation events such as touch, annotation, and page turning were recorded at a frequency of f_op = 8 times / second, which is higher than the preset high-frequency interaction threshold Th_high = 5 times / second. Th_high represents the lower limit threshold for high-frequency interaction. The high-frequency interaction threshold is used to distinguish whether a user is in a high-frequency interaction state and is obtained by analyzing a large amount of user behavior data from historical meetings. Before the interruption, the node was processing a real-time annotation core chart task assigned by the cloud. Its priority label P_task is assigned by the cloud based on the task's criticality in the meeting, and its value range is [0,1]. In this embodiment, P_task=0.9, indicating high priority. In this embodiment, a weighted scoring method is used. The urgency scoring formula is: S_urgency = α × (f_op / Th_high) + β × P_task; where α and β are weight coefficients, and α + β = 1; α represents the influence weight of user operation frequency on urgency, and β represents the influence weight of task priority on urgency; the calculated S_urgency ≈ 1.28; the mapping rule is determined by analyzing the urgency of subsequent user behavior corresponding to different S_urgency intervals in historical collaboration data; for example, the numerical range of S_urgency is divided into three equal parts, or a dividing point is set according to experience, such as S_urgency > 1.0 is a high urgency level, 0.5 ≤ S_urgency ≤ 1.0 is a medium urgency level, and S_urgency < 0.5 is a low urgency level; therefore, the predicted urgency level of collaboration after the interruption is high urgency level; The interruption prediction matrix is ​​composed of the expected duration level and the coordination urgency level; Specifically, the expected duration level and the coordination urgency level results are combined to generate an interruption prediction matrix M_interrupt. The interruption prediction matrix is ​​represented as an ordered pair or feature vector. In this embodiment, M_interrupt=(Medium, High); or encoded as a numeric vector [1, 2], with medium duration encoded as 1 and high urgency encoded as 2. Furthermore, the triggering types of the interruption events include network connection abnormalities, device state switching, or application abnormalities; the operating status parameters include network signal attenuation rate, historical average device switching time, or application process abnormal stack depth; the network connection abnormality specifically refers to the wireless network signal strength being lower than a preset threshold or the transmission control protocol connection timing out continuously; the device state switching specifically refers to user operation causing the conference screen to switch from the current display and computing entity to a subordinate device; the application abnormality specifically refers to the core process of the smart conferencing application crashing or service becoming unresponsive; The user interaction status includes whether the edge node was executing a collaborative task issued by the cloud before the interruption; the collaborative task issued by the cloud includes, but is not limited to, real-time annotation of specified document areas, transcription of the voice content of a specific speaker, or verification of data conclusions submitted by other nodes. The collaborative task status includes the priority tags of the tasks processed by the edge nodes before the interruption. The priority tags are assigned by the cloud during the collaboration process and are used to identify the criticality and processing order of the tasks in the global collaborative process.

[0019] Specifically, step S2 includes: The edge node selects and executes a target caching strategy from the caching strategies based on the interruption prediction matrix. The identifier of the target caching strategy constitutes the first part of the warm-up strategy label. The caching strategy defines the retention rules for collaborative data in local memory and storage space. The identifier is used to uniquely distinguish different caching strategies. Furthermore, when the interruption prediction matrix corresponds to a combination of short-term interruption and high urgency, the edge node selects a limit retention strategy; when the interruption prediction matrix corresponds to a combination of long-term interruption and high urgency, the edge node selects a key anchor point caching strategy; for other combinations, the edge node selects an intelligent summarization caching strategy. The limit retention strategy refers to prioritizing and completely retaining the user's interaction sequence at the last moment before the interruption and related incomplete analysis intermediate states within a limited cache space; the key anchor point caching strategy refers to caching only the user's original, non-reproducible logical conclusions and core annotation semantics, abandoning reconstructable complete data; the intelligent summarization caching strategy refers to extracting and compressing features from collaborative content and analysis status, retaining only structured summaries and key feature vectors. Specifically, based on the interruption prediction matrix M_interrupt=(Medium, High), the preset strategy mapping rules are queried; the rules stipulate that for other combinations that are not short-term-high urgency or long-term-high urgency, an intelligent summary caching strategy is executed. In this embodiment, the intelligent summarization caching strategy employs a key feature extraction algorithm; it analyzes the collaborative content on the current screen, such as whiteboard handwriting, document paragraphs, and topic lists; for handwriting, the Douglas-Peucker algorithm is used to simplify the trajectory, retaining key turning points and compressing the data volume by 70%; for text, the TF-IDF algorithm is used to extract keywords from each paragraph to form a semantic summary; all summaries and intermediate results of the current analysis state are cached together; this strategy is identified as Strategy_Summary. The edge node performs a data preheating action based on the interruption prediction matrix. The type identifier of the data preheating action constitutes the second part of the preheating strategy label. The data preheating action is to prepare the data required for recovery or establish a connection channel in advance during the interruption. Furthermore, when the expected duration level is predicted to be a long duration level, the data preheating action includes sending a request to the cloud, pre-generating a state difference data packet, and attaching the preheating strategy label to the request. The state difference data packet refers to the set of incremental data that the cloud predicts may occur during the interruption based on the preheating strategy label and pre-calculates, which can be quickly applied when the edge node recovers. Specifically, in this embodiment, since the expected duration level is medium, the action of requesting the pre-generation of differential data packets is not triggered according to the rules; however, the system still performs basic warm-up, that is, maintains the active heartbeat with the cloud control channel, and adds the object ID of the currently collaboratively edited document Doc_A: ObjID_DocA and the last known version number Ver_123 to the warm-up task queue; the type identifier of this warm-up action is denoted as WarmUp_Basic; this is only an example and is not a limitation.

[0020] Specifically, step S3 includes: The cloud receives interruption status information from each edge node. The interruption status information includes the expected duration level of the interruption, the coordination urgency level, and the coordination relationship between the nodes. The coordination relationship is defined by the coordination task dependency graph, which represents the dependency between nodes on data flow or control flow. Specifically, the cloud receives the interruption status information of the edge node, including M_interrupt, L_warm, and its position in the collaborative task dependency graph. Cloud analysis reveals that two nodes on this Task_Chain_X have reported a high urgency level. According to the rules, this task chain is marked as a high-pressure collaborative zone H_zone_X. Further analysis of L_warm by the cloud reveals that its caching strategy is Strategy_Summary, which is non-limited retention, therefore it does not increase the pressure level of H_zone_X. The cloud analyzes the interruption prediction matrix of each edge node. If multiple nodes with the same high urgency level and located on the same collaborative link are detected, the link is marked as a high-pressure collaborative region. The cloud analyzes the preheating strategy labels of each edge node. If a node's strategy is detected as an extreme retention strategy, the pressure level of the high-pressure collaborative region corresponding to that node is increased. High-pressure collaborative regions and high-pressure recovery points are identified globally, and a global collaborative pressure heatmap is constructed based on the identification results. A high-pressure recovery point refers to an edge node in the high-pressure collaborative region whose task is blocked due to its reliance on the output of the interrupted node. The global collaborative pressure heatmap, in the form of visualization or data structure, represents the congestion risk and resource demand intensity faced by different collaborative links during the recovery phase. Based on the global collaborative pressure heatmap, the cloud sends resource coordination instructions to the interrupted nodes. These instructions include cache resource allocation adaptation instructions and instructions for reserving the necessary computing power and bandwidth resources for recovery. It also sends behavior optimization instructions to edge nodes that are collaboratively associated with the interrupted nodes. These instructions include instructions to postpone modifications to non-core collaborative data and instructions to control the frequency of collaborative data updates. The update frequency is dynamically adjusted based on the preheating strategy label of the associated node. Dynamic adjustment includes reducing the update frequency if the preheating strategy label of the associated node indicates that it adopts a maximum retention strategy; and allowing or increasing the update frequency if it indicates that it adopts a key anchor or intelligent summary strategy. Specifically, a global collaborative pressure heatmap (HeatMap_Global) is constructed by integrating information from all nodes. In this embodiment, the global collaborative pressure heatmap is a weighted graph data structure, where nodes represent edge nodes and edges represent collaborative relationships. When constructing the global collaborative pressure heatmap (HeatMap_Global), for any collaborative edge e_{ij} connecting two edge nodes i and j, its weight W_{ij} is calculated by the formula: W_{ij}=γ×U_{ij}+δ×P_{ij}; where W_{ij} represents the weight of edge e_{ij}. The larger this value, the greater the pressure of the collaborative relationship between node i and node j under the current interruption recovery scenario. The higher the force or criticality, the more core the value represented by the heatmap; γ represents the urgency influence coefficient; a non-negative real constant used to adjust the contribution ratio of collaborative urgency to edge weights. The larger the γ value, the more the system values ​​the urgency of the task itself when assessing collaborative pressure; U_{ij} represents the collaborative urgency score, quantifying the urgency of the joint task participated in by nodes i and j; its calculation can be based on the urgency level in the two-node interruption prediction matrix, such as: U_{ij}=max(Urgency_Level_i, Urgency_Level_j), mapping the urgency level to a numerical value; or U_{ij}=(S_urgency_i+ S_urgency_j) / 2, using the mean of the original urgency scores S_urgency of the two nodes; δ represents the caching strategy penalty coefficient, used to adjust the contribution ratio of the additional risk brought about by the aggressive caching strategy adopted by the dependent party to the edge weight; P_{ij} represents the caching strategy penalty term, the core logic of which is: if the strategy of node j indicates that it has extremely high requirements for data consistency, then it will be more sensitive to the recovery delay of node i. Based on HeatMap_Global, the cloud sends a resource coordination instruction to the interrupted node: reserving an additional 10% computing resource quota for it in the cloud and marking its data synchronization recovery channel as priority level P1; At the same time, behavior tuning instructions were issued to online nodes that have a direct dependency on this node: it was suggested that they reduce the frequency of submitting non-critical intermediate results to the shared data zone Data_Zone_X from once per second to once every 3 seconds; the reduction ratio of update frequency in the instruction parameters was partly based on the fact that the L_warm of the associated interrupted node does not contain an extreme retention policy, so there is no need to perform extreme rate limiting.

[0021] In this embodiment, the network is restored at t=10:00:15, and the interruption recovery condition is triggered as an example; Specifically, step S4 includes: When the interruption recovery condition is triggered, the edge node determines the storage structure and integrity of the local cache analysis status based on the preheating strategy label; the storage structure includes whether the cached data is organized in complete time sequence, logical summary or key anchor mode; the integrity includes the degree to which the cached data covers the original collaborative context and logical structure. Specifically, based on L_warm.Cache=Strategy_Summary, the edge node knows that the locally cached data is a summary; therefore, it determines that the storage structure is a logical summary and the integrity is medium. The edge node sends a request to the cloud to obtain the latest global status of the cloud. The latest global status of the cloud includes the global collaborative analysis status and the current collaborative status of each associated edge node. The global collaborative analysis status includes a unified view of the collaborative progress of all online nodes maintained by the cloud. The current collaborative status refers to the latest task progress, data version and interaction results of each node. Specifically, the node sends a request to the cloud to obtain the latest global state S_cloud of ObjID_DocA after Ver_123; Since the local data is a summary, the comparison focuses on key logical consistency. The TF-IDF keyword summary vector V_local cached locally is compared with the corresponding summary vector V_cloud of the latest document in the cloud. In this embodiment, the cosine similarity algorithm is used, with the formula: Similarity=(V_local·V_cloud) / (||V_local||·||V_cloud||); where · represents the dot product operation of vectors, and ||V|| represents the Euclidean norm of vector V; the calculated Similarity=0.85, and the Similarity range is [0,1], with values ​​closer to 1 indicating greater similarity. Meanwhile, comparing the local topic list with the cloud topic list, it was found that a new topic risk discussion had been added in the cloud. Based on the storage structure, the edge node determines the comparison granularity and focus range for comparing with the latest global state in the cloud; based on the completeness, it determines the confidence assessment of the difference range when generating the state difference impact report. The comparison granularity and focus range are determined according to the storage structure. If it is a limit-retention structure, a fine-grained complete comparison is performed; if it is a summary or anchor structure, a key logic consistency comparison is performed. The confidence assessment is used to quantify the credibility of the state difference impact report. Specifically, a report R_diff is generated based on the comparison results; the report states: Difference type: New content added, minor semantic changes to content; Scope of differences: This includes certain paragraphs and the overall list of topics in document Doc_A; Impact level: High; because the new issues may change the direction of the discussion, and the current situation is one of high urgency. Confidence assessment: Since it is based on summary comparison, the confidence level is rated as 0.7, which is moderately high, and the range is [0,1]. This is only for illustrative purposes and is not a limitation.

[0022] Specifically, step S5 includes: The edge node acquires the collaboration urgency level and extracts the difference range and impact degree from the state difference impact report. Based on the collaboration urgency level, difference range, and impact degree, a target fusion strategy is selected from the fusion strategies. The difference range refers to the breadth of data or logical modules where state differences occur. The impact degree refers to the potential damage level caused by the difference to the continuity and correctness of the collaborative task after recovery. Furthermore, the selection of the target fusion strategy from the fusion strategies includes: When the collaboration urgency level is high urgency level, if the degree of difference is lower than a first difference threshold, a first fusion mode is selected and executed; if the degree of difference is higher than or equal to the first difference threshold, a second fusion mode is selected and executed; when the collaboration urgency level is medium urgency level, the granularity and order of fusion are determined based on the amount of difference data or conflict complexity associated with the degree of difference, and a third fusion mode is selected and executed; when the collaboration urgency level is low urgency level, a fourth fusion mode is selected and executed; the first difference threshold includes: if the target caching strategy is a limit retention strategy, the first difference threshold is set to a first value; if the target caching strategy is a key anchor point caching strategy, the first difference threshold is set to a second value; if the target caching strategy is an intelligent summarization caching strategy, the first difference threshold is set to a third value; wherein, the first value is greater than the second value, and the second value is greater than the third value; The more aggressive the caching strategy adopted in the early stage, the higher the tolerance for differences encountered during recovery; the more streamlined the caching strategy in the early stage, the more sensitive it is to differences, and the more likely it is to require user intervention for confirmation. Specifically, the first, second, and third values ​​are dynamically acquired and adjusted based on an adaptive learning mechanism that utilizes historical recovery performance. Maintain a historical recovery log library, recording tuple data (CST, CDM, UIF) for each interruption recovery event. CST represents the caching strategy type used by the edge node associated with this recovery during the interruption; CDM represents a comprehensive difference metric D calculated based on the state difference impact report during recovery, where D is the complement of the Similarity calculated in step S4; and UIF is a boolean flag indicating whether user intervention is ultimately required for this recovery, i.e., whether the second fusion mode was executed; True indicates user intervention is required, and False indicates no user intervention is required. For each caching strategy type s, s∈{Limited Retention, Key Anchor, Smart Summary}, its corresponding difference threshold Th_s, i.e. the first, second and third values, are maintained and updated independently. Once a recovery event is completed and its UIF is determined, the event's data (s, D, Flag) is added to the log database. Periodically analyze historical data under this strategy s; the goal is to find the threshold Th_s that best distinguishes between cases requiring user intervention and cases without user intervention. Specifically, for strategy s, its historical data is divided into two categories according to UIFl: Category C0 (Flag=False): the set of difference metrics that do not require user intervention {D_i}; Category C1 (Flag=True): the set of difference metrics that require user intervention {D_j}. Calculate the empirical probability distributions for the two categories; in this embodiment, the comprehensive difference measure D follows a Gaussian distribution, and its mean μ0 and μ1 and variance σ0 are estimated using data from category C0 and category C1, respectively. 2 and σ1 2 Find a threshold Th_s that minimizes the classification error rate for the two classes of data. For example, take the weighted midpoint of the means of two distributions and consider their variance: Th_s = (σ1×μ0 + σ0×μ1) / (σ0 + σ1). That is, in the overlapping area of ​​the two data distributions, take a point whose distance from the center of the two distributions is inversely proportional to the width of the distribution. The class with larger variance receives less weight, and the threshold will be more biased towards the class with smaller variance. Then update the calculated Th_s to the current difference threshold of the strategy s. When the system is initially running, an empirical default threshold is set for each strategy s, such as limit preservation: 0.5, key anchor: 0.3, and smart summary: 0.15, and it converges quickly through the above learning process; this is only an example and is not a limitation. The first fusion mode aims for real-time interactive continuity, allowing the local state to take effect immediately as the interaction baseline, and completing the asynchronous fusion of difference data in the background; the second fusion mode aims to quickly resolve key conflicts, by presenting a merged view focusing on core differences to the user and obtaining confirmation, and performing selective fusion; the third fusion mode aims to restore efficiency, by fusion of difference data in batches and progressively according to preset business logic priorities; the fourth fusion mode aims to ensure data integrity and consistency, by uniformly updating the local state after ensuring that all difference data has been synchronized and verified. Asynchronous fusion includes completing data synchronization and merging in a background thread without interrupting the user's current interface operation; selective fusion includes merging only the core differences confirmed by the user, while the rest are processed according to preset rules; progressive fusion includes merging the differences into the local state in batches according to the importance or dependency of the data, so that users can work in advance based on the partially merged context. The edge nodes perform a fusion operation on the differences between the local state and the cloud state based on the target fusion strategy.

[0023] Specifically, the urgency level of the transport coordination is high, and the degree of influence in R_diff is high; Based on the cache strategy type Strategy_Summary identified in the preheating strategy tag associated with this recovery, the dynamic threshold corresponding to this strategy type, i.e., the third value, is found from the currently maintained threshold table and assigned to Th_diff1; in this embodiment, this threshold is obtained through historical learning and its current value is 0.7. The impact level of R_diff is high, which means that although Similarity=0.85 is higher than Th_diff1 (0.7), the independent event of adding a new issue has triggered the high impact flag. When the impact level in the status difference impact report is judged to be high, regardless of the comparison result between the comprehensive difference measure D and the threshold Th_diff1, it will be preferentially judged as having a high impact flag. According to the rules, the second fusion mode should be selected, namely focus-guided fusion; The system immediately uses the locally cached state as the interactive baseline to restore the interface, allowing users to immediately see the core content before the interruption and resume operations. At the same time, a difference merge view pops up in the non-blocking area on the side of the interface, clearly highlighting: "New topic risk discussion detected during the interruption, should it be included in the current discussion? Accept or reject." The user clicks "Accept"; the system performs selective fusion: merging the new topic into the local topic list, and based on the user's selection, pulling and merging the changed parts of the cloud document; for semantically minor changes, since the differences are not significant, the system automatically performs silent updates based on the cloud version according to preset rules; throughout the process, the user's main working interface is not blocked, achieving rapid recovery and resolution of key conflicts.

[0024] This invention provides another technical solution: a real-time collaboration system for smart meeting screens based on cloud-edge collaboration. The system includes: an interruption prediction module, a warm-up strategy module, a cloud scheduling module, a report generation module, and a fusion execution module. The interruption prediction module is deployed on edge nodes. When a collaboration interruption is detected, it generates an expected duration level and a collaboration urgency level based on the interruption event trigger type, associated running status parameters, user interaction status, and collaboration task status, and constructs an interruption prediction matrix. The preheating strategy module is deployed at the edge node and connected to the interruption prediction module. It is used to receive the interruption prediction matrix and select a target caching strategy according to a specific combination of duration and urgency in the interruption prediction matrix; execute the corresponding data preheating action according to the duration level, and encode the identifier of the target caching strategy and the data preheating action into a preheating strategy label. The cloud scheduling module, deployed on a cloud server, communicates with each edge node. It receives and analyzes the interruption prediction matrix and preheating strategy labels of each node. The analysis specifically includes identifying high-urgency collaborative links as high-pressure areas based on the prediction matrix, adjusting the pressure level of the high-pressure areas according to the strategy identifier in the preheating strategy label, constructing a global collaborative pressure heat map based on this, generating and issuing resource coordination instructions and behavior optimization instructions to relevant nodes. The parameters of the behavior optimization instructions are dynamically adjusted according to the preheating strategy labels of the associated nodes. The report generation module is deployed on the edge node and connected to the preheating strategy module. It is used to determine the comparison granularity and confidence level based on the caching strategy and data integrity indicated by the preheating strategy label during interruption recovery, compare the local cache status with the global status in the cloud, and generate a status difference impact report. The fusion execution module is deployed on an edge node and is connected to the interruption prediction module and the report generation module, respectively. It is used to extract the collaborative urgency level in the interruption prediction matrix and analyze the degree of difference in the state difference impact report. The first difference threshold used for decision-making is dynamically set according to the target caching strategy type selected by the preheating strategy label generation unit, and the corresponding fusion mode is selected and executed based on the urgency level, the degree of difference and the dynamic threshold.

[0025] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A real-time collaboration method for smart meeting screens based on cloud-edge collaboration, characterized in that: The method includes: S1. When an edge node detects a collaboration interruption, it performs parallel predictions on the expected duration of the interruption and the collaboration urgency level after the interruption is restored, and generates an interruption prediction matrix. S2. Based on the interruption prediction matrix, the edge node performs dynamic allocation of cache resources and data preheating for rapid recovery, and generates a preheating strategy label containing the identifier of the executed strategy. S3. The cloud receives the interruption prediction matrix and the preheating strategy label from each edge node, constructs a global collaborative pressure heat map, and issues resource coordination and behavior optimization instructions to relevant nodes. S4. When the interruption recovery condition is triggered, the edge node locates the analysis status of the local cache based on the preheating strategy label, quickly compares the analysis status of its local cache with the latest global status in the cloud, and generates a status difference impact report. S5. The edge node executes the fusion strategy based on the coordination urgency level in the interruption prediction matrix and the state difference impact report.

2. The real-time collaboration method for a smart conference screen based on cloud-edge collaboration according to claim 1, characterized in that: Step S1 includes: The edge node determines the expected duration of the interruption as short-term, medium-term, or long-term based on the triggering type of the interruption event and the associated running status parameters. Based on the user interaction status and collaborative task status before the interruption, the edge node predicts the urgency of collaboration after the interruption is restored as high urgency level, medium urgency level, or low urgency level. The interruption prediction matrix is ​​composed of the expected duration level and the coordination urgency level.

3. The real-time collaboration method for a smart conference screen based on cloud-edge collaboration according to claim 2, characterized in that: The triggering types of the interruption events include network connection abnormality, device state switching, or application abnormality; the running status parameters include network signal attenuation rate, historical average device switching time, or application process abnormal stack depth. The user interaction status includes whether the edge node was executing a collaborative task issued by the cloud before the interruption; The collaborative task status includes the priority label of the task processed by the edge node before the interruption, and the priority label is assigned by the cloud during the collaboration process.

4. The real-time collaboration method for a smart conference screen based on cloud-edge collaboration according to claim 1, characterized in that: Step S2 includes: The edge node selects and executes a target caching strategy from the caching strategies based on the interruption prediction matrix, and the identifier of the target caching strategy constitutes the first part of the warm-up strategy label; The edge node performs a data preheating action based on the interruption prediction matrix, and the type identifier of the data preheating action constitutes the second part of the preheating strategy label.

5. A real-time collaboration method for a smart conference screen based on cloud-edge collaboration according to claim 4, characterized in that: When the interrupt prediction matrix corresponds to a combination of short-term interruption and high urgency, the edge node selects a limit retention strategy; when the interrupt prediction matrix corresponds to a combination of long-term interruption and high urgency, the edge node selects a critical anchor point caching strategy. For other combinations, the edge nodes select an intelligent summary caching strategy; When the expected duration level is predicted to be a long duration level, the data preheating action includes sending a request to the cloud, pre-generating a state difference data packet, and attaching the preheating strategy label to the request.

6. The real-time collaboration method for a smart conference screen based on cloud-edge collaboration according to claim 1, characterized in that: Step S3 includes: The cloud receives interruption status information from each edge node, including the expected duration of interruption, the coordination urgency level, and the coordination relationship between nodes. The cloud analyzes the interruption prediction matrix of each edge node. If multiple nodes with the same high urgency level and on the same collaborative link are detected, the link is marked as a collaborative pressure high-pressure area. The cloud analyzes the preheating strategy label of each edge node. If a node's strategy is detected as an extreme retention strategy, the pressure level of the collaborative pressure high-pressure area corresponding to that node is increased. Collaborative pressure high-pressure areas and high-pressure recovery points are identified globally, and a global collaborative pressure heat map is constructed based on the identification results. Based on the global collaborative pressure heatmap, the cloud sends resource coordination instructions to the interrupted nodes. These instructions include instructions for caching resource allocation adaptation and instructions for reserving the necessary computing power and bandwidth resources for recovery. The cloud also sends behavior optimization instructions to edge nodes that are collaboratively associated with the interrupted nodes. These instructions include instructions for suspending modifications to non-core collaborative data and instructions for controlling the frequency of collaborative data updates. The update frequency is dynamically adjusted based on the preheating strategy tags of the associated nodes.

7. The real-time collaboration method for a smart conference screen based on cloud-edge collaboration according to claim 1, characterized in that: Step S4 includes: When the interruption recovery condition is triggered, the edge node determines the storage structure and integrity of the local cache analysis status based on the preheating strategy label; The edge node sends a request to the cloud to obtain the latest global status of the cloud, which includes the global collaborative analysis status and the current collaborative status of each associated edge node. Based on the storage structure, the edge node determines the comparison granularity and focus range for comparing with the latest global state in the cloud; based on the integrity, it determines the confidence assessment of the difference range when generating the state difference impact report.

8. The real-time collaboration method for a smart conference screen based on cloud-edge collaboration according to claim 1, characterized in that: Step S5 includes: The edge node obtains the collaboration urgency level and extracts the difference range and impact degree from the state difference impact report. Based on the collaboration urgency level, difference range and impact degree, a target fusion strategy is selected from the fusion strategies. The edge nodes perform a fusion operation on the differences between the local state and the cloud state based on the target fusion strategy.

9. A real-time collaboration method for a smart conference screen based on cloud-edge collaboration as described in claim 8, characterized in that: The selection of a target fusion strategy from the fusion strategies includes: When the collaboration urgency level is high urgency level, if the degree of difference is lower than a first difference threshold, a first fusion mode is selected and executed; if the degree of difference is higher than or equal to the first difference threshold, a second fusion mode is selected and executed; when the collaboration urgency level is medium urgency level, the granularity and order of fusion are determined based on the amount of difference data or conflict complexity associated with the degree of difference, and a third fusion mode is selected and executed; when the collaboration urgency level is low urgency level, a fourth fusion mode is selected and executed; the first difference threshold includes: if the target caching strategy is a limit retention strategy, the first difference threshold is set to a first value; if the target caching strategy is a key anchor point caching strategy, the first difference threshold is set to a second value; if the target caching strategy is an intelligent summarization caching strategy, the first difference threshold is set to a third value; wherein, the first value is greater than the second value, and the second value is greater than the third value; The first fusion mode aims for real-time interactive continuity, allowing the local state to take effect immediately as the interaction baseline, and completing the asynchronous fusion of difference data in the background; the second fusion mode aims to quickly resolve key conflicts, by presenting a merged view focusing on core differences to the user and obtaining confirmation, and performing selective fusion; the third fusion mode aims to restore efficiency, by fusion of difference data in batches and progressively according to preset business logic priorities; the fourth fusion mode aims to ensure data integrity and consistency, by uniformly updating the local state after ensuring that all difference data has been synchronized and verified.

10. A real-time collaboration system for smart meeting screens based on cloud-edge collaboration, characterized in that: The system includes: an interruption prediction module, a preheating strategy module, a cloud scheduling module, a report generation module, and a fusion execution module; The interruption prediction module is deployed on edge nodes. When a collaboration interruption is detected, it generates an expected duration level and a collaboration urgency level based on the interruption event trigger type, associated running status parameters, user interaction status, and collaboration task status, and constructs an interruption prediction matrix. The preheating strategy module is deployed at the edge node and connected to the interruption prediction module. It is used to receive the interruption prediction matrix and select a target caching strategy according to a specific combination of duration and urgency in the interruption prediction matrix; execute the corresponding data preheating action according to the duration level, and encode the identifier of the target caching strategy and the data preheating action into a preheating strategy label. The cloud scheduling module, deployed on a cloud server, communicates with each edge node. It receives and analyzes the interruption prediction matrix and preheating strategy labels of each node. The analysis specifically includes identifying high-urgency collaborative links as high-pressure areas based on the prediction matrix, adjusting the pressure level of the high-pressure areas according to the strategy identifier in the preheating strategy label, constructing a global collaborative pressure heat map based on this, generating and issuing resource coordination instructions and behavior optimization instructions to relevant nodes. The parameters of the behavior optimization instructions are dynamically adjusted according to the preheating strategy labels of the associated nodes. The report generation module is deployed on the edge node and connected to the preheating strategy module. It is used to determine the comparison granularity and confidence level based on the caching strategy and data integrity indicated by the preheating strategy label during interruption recovery, compare the local cache status with the global status in the cloud, and generate a status difference impact report. The fusion execution module is deployed on an edge node and is connected to the interruption prediction module and the report generation module, respectively. It is used to extract the collaborative urgency level in the interruption prediction matrix and analyze the degree of difference in the state difference impact report. The first difference threshold used for decision-making is dynamically set according to the target caching strategy type selected by the preheating strategy label generation unit, and the corresponding fusion mode is selected and executed based on the urgency level, the degree of difference and the dynamic threshold.

Images