Intelligent information display system based on end-to-end cooperation

Abstract

The invention provides an intelligent information display system based on end-to-end cooperation. The system improves the real-time performance, the accuracy and the display efficiency of intelligent information processing. Firstly, a dynamic data classifier is adopted to intelligently classify information; thirdly, common information and important information are processed separately through a dual-channel processing module, computing resources are dynamically allocated according to the priority of the common information and the important information, it is ensured that high-priority real-time information can respond in time, the computing resources are dynamically allocated, and redundancy and delay of resource allocation are avoided; and finally, the difference visualization module displays the aggregation result of the common information and the streaming calculation result of the important information on the terminal equipment, and generates a connection path, so that the interactivity and the visualization effect of information display are enhanced. Through intelligent classification, double-channel parallel processing, dynamic resource scheduling and intelligent visual display, the processing efficiency, response speed and user experience of intelligent information are remarkably improved, and the defects in the prior art are overcome.

Description

Technical Field

[0001] The present invention relates to the field of intelligent information display, and in particular to an intelligent information display system based on end-to-end collaboration. Background Art

[0002] With the development of intelligent technology, terminal devices and sensors are playing an increasingly important role in real-time data collection. In fields such as smart buildings, smart cities, and industrial automation, a large number of sensors and devices can monitor intelligent information such as the environment, energy, and equipment status in real time, and generate massive amounts of data. However, when faced with such a huge amount of data, traditional intelligent information display systems are not smart and timely enough in processing information. Moreover, a large amount of data means high load, and traditional technologies are often unable to intelligently allocate computing resources, resulting in system resource waste or task processing delays.

[0003] Existing technologies use a technical solution that collaborates with edge devices and terminal devices. By migrating data processing from the cloud to edge nodes, computing tasks are shared, reducing data transmission delays and improving response speeds. However, existing solutions still have some shortcomings. First, they fail to intelligently classify intelligent information based on its importance, resulting in an even distribution of system resources and computing power, which may lead to missed timely responses to emergencies. Furthermore, they rely on a single processing channel, resulting in an inability to respond in a timely manner when processing large-scale real-time information, affecting the real-time nature and decision-making efficiency of the system. Summary of the Invention

[0004] In order to solve the technical problems mentioned in the current background technology, the present invention proposes an intelligent information display system based on end-to-end collaboration.

[0005] To this end, the technical solution adopted in the present invention is as follows: An intelligent information display system based on end-to-end collaboration includes several terminal devices and edge nodes. The terminal devices communicate with the edge nodes and collect intelligent information. The edge nodes have built-in dynamic data classifiers that classify the intelligent information into general information and important information. The system includes: M1, a dual-channel processing module, is deployed on edge nodes and includes a first channel, a second channel, and a channel resource scheduling unit. The second channel has a higher priority than the first channel. The first channel aggregates common information in batches, while the second channel performs real-time streaming calculations on important information. The channel resource scheduling unit dynamically allocates computing resources between the first and second channels based on the current load status. M2, the difference visualization module, is deployed on the terminal device. The aggregation results of the first channel are displayed in the edge area; the real-time streaming calculation results of the second channel are displayed in the core area; when the ordinary information and important information are associated, a connection path is automatically generated; when the number of times the user visits the connection path exceeds the set number, the automatic merging mechanism is triggered.

[0006] Furthermore, the specific division steps of the dynamic data classifier are: 1) Calculate the information structure complexity, information mutation coefficient, and time sensitivity of the intelligent information. The information structure complexity is expressed as: in, Indicates the complexity of information structure; Indicates the number of information fields; Indicates the field nesting level; Indicates information volume; The information mutation coefficient is The mutation degree of intelligent information is calculated in a time window, which is expressed as: in, represents the information mutation coefficient; Represents intelligent information In the time window The mean within Represents a time window Standard deviation within; is a numerical stability term; The time sensitivity is expressed as: in, Indicates time sensitivity; Represents intelligent information The permissible delay from generation to the latest acceptable display time; 3) Calculate the importance of the intelligent information using the formula: in, Represents intelligent information the importance of 、 and is the weight factor; set the importance threshold ,when , indicating intelligent information If it is not satisfied, it is ordinary information.

[0007] Furthermore, the first channel extracts the overall main trend sequence of the general information through a weighted principal component analysis method, specifically, Get The general information set within a time window is expressed as: Perform covariance calculation on the common information set to obtain the covariance matrix , The covariance matrix is ​​used to extract the principal components to obtain the principal component results and the corresponding unit eigenvectors, which are expressed as: The weight coefficient of the principal component is defined as follows: in, It is The weights of the principal components; Map the general information set to the principal component space and calculate the overall main trend sequence. The formula is: in, Indicates general information The main trend response value of Indicates general information In the The projection in the direction of the principal component; the overall main trend sequence constitutes the visualization display result of general information.

[0008] Furthermore, the specific steps of the real-time streaming calculation of the second channel are: Get the important information set of time step t, expressed as: Using the self-attention mechanism in the Transformer model, the important information set is mapped to the query vector Q, key vector K and value vector V, which can be expressed as: in, and Represents the weight matrix of training; Calculate the attention score matrix , expressed as: in, is the dimension of the vector; The value vector is weighted averaged by the attention score matrix, expressed as: in, It is a temporal feature; The time series features are input into the LSTM layer and expressed as: in, is the hidden state at the previous moment; Based on the output of the isolation forest algorithm and the LSTM layer, an anomaly score of the important information is calculated, and a visualization result of the important information is constructed according to the anomaly score.

[0009] Furthermore, the channel resource scheduling unit adjusts computing resources through deep reinforcement learning, sets the state space, action space and reward function, State Space Including current load , First channel resource requirements , Second channel resource requirements ; Action Space Defined as the allocation strategy for computing resources, the allocation strategy is the ratio of resources allocated to the second channel and the first channel and ; Reward Function Based on throughput , priority violation Perform optimization, defined as: in, and is the weight parameter; The throughput is defined as: The priority violation is defined as the computational resources of the second channel not being satisfied, expressed as: According to the current status Choose the best action , the formula is: in, It is the action at the present moment; is the discount factor; is the Q value of the current state and action; Take action After the state; Is in state The next action to be taken.

[0010] Furthermore, the connection path is generated according to the mutual information value of the common information and the important information, and the formula is: in, is the joint probability distribution of common information and important information; and is the probability distribution of the individual variables; set the mutual information threshold ,when , then there is a correlation between ordinary information and important information, and a connection path is automatically generated.

[0011] Compared with the prior art, the advantages of the present invention are: 1. By introducing a dynamic data classifier, the present invention can dynamically classify intelligent information, ensure that important information is given priority when allocating resources, and improve the accuracy and efficiency of information processing.

[0012] 2. The dual-channel processing mechanism of the present invention is deployed at the edge node, processing common information and important information separately, ensuring that high-priority important information can be streamed in real time, significantly improving the real-time performance and reliability of information processing.

[0013] 3. The present invention adopts a channel resource scheduling unit, which can monitor the system load status in real time and dynamically allocate computing resources according to load changes. When the system load is high, the resource allocation of the second channel is prioritized, thereby effectively avoiding resource waste. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0015] Figure 1 This is a schematic diagram of the architecture of the end-to-end collaboration of the present invention; Figure 2 This is a flow chart of the dual-channel processing module of the present invention; Figure 3 This is a flow chart of the difference visualization module of the present invention. DETAILED DESCRIPTION

[0016] To achieve the above objectives, the present invention is implemented through the following technical solutions: the present invention provides an intelligent information display system based on end-to-end collaboration, please refer to Figures 1 to 3 , specifically, The intelligent information display system adopts a distributed architecture, including several terminal devices and edge nodes. The terminal devices communicate with the edge nodes and collect intelligent information. The edge nodes have built-in dynamic data classifiers to divide the intelligent information into general information and important information. In the embodiment, the terminal device is a smart large screen, mobile terminal, in-vehicle interactive screen, etc., which collects intelligent information through a standardized data interface (such as MQTT, gRPC or OPC UA) and establishes a stable communication connection with the edge node. The communication protocol used is 5G network, Wi-Fi6 or industrial Ethernet to ensure low latency and high availability of intelligent information transmission; all intelligent information is recorded as a collection , where each Represents an information instance.

[0017] After receiving the intelligent information uploaded by the terminal device, the edge node executes the dynamic data classification processing mechanism through the built-in dynamic data classifier to classify each intelligent information Divide into general information or important information, For each piece of intelligent information, three indicators are calculated, which are divided into information structure complexity , information mutation coefficient and time sensitivity , Information structure complexity is an important metric for measuring the internal field structure, nesting level, and semantic density of intelligent information. It is used to determine the resource consumption and semantic richness required during the processing and analysis of intelligent information. The calculation formula is: in, Indicates the number of information fields, such as the number of key-value pairs in JSON; Indicates the field nesting level; Indicates information volume; The larger the value, the more complex the intelligent information structure is; The information mutation coefficient is used to measure the magnitude of the value change or the degree of mutation of the behavior pattern of the intelligent information item within a recent time window to determine whether the intelligent information may correspond to an emergency, a state switch, or an out-of-control trend. In the embodiment, mutation behavior means that the terminal device has an abnormal operating condition, the environment has entered a critical state, or the user has performed a critical operation. Therefore, intelligent information with high mutation is classified as important information and is given priority in the streaming processing channel and the core area for display. In the edge node, the dynamic data classifier is based on The mutation degree of intelligent information is quantified in a time window, which is expressed as: in, Represents intelligent information In the time window The mean within Represents a time window Standard deviation within; is a numerical stability term; Time sensitivity is used to measure whether a piece of intelligent information is time-sensitive. That is, whether the display must be completed within a specified time limit. Failure to do so will result in information failure, delayed policy response, or inability for users to make effective decisions. By introducing a time sensitivity indicator, it is used to express the tolerance of intelligent information to processing delays. The formula is: in, Represents intelligent information The acceptable delay from generation to the latest display time. The larger the value of the time sensitivity index, the more sensitive the intelligent information is and the sooner it needs to be displayed.

[0018] After normalizing the three indicators, the dynamic data classifier calculates the importance of each piece of intelligent information, expressed as: in, Represents intelligent information the importance of 、 and is a weight factor that satisfies the sum equal to 1; sets the importance threshold ,when , indicating intelligent information Important information, if not met, it is ordinary information.

[0019] M1, a dual-channel processing module, is deployed at the edge node and includes a first channel, a second channel, and a channel resource scheduling unit. The first channel aggregates common information in batches, while the second channel performs real-time streaming computation on important information. The channel resource scheduling unit dynamically allocates computing resources between the first and second channels based on the current load status. The first channel is deployed in the edge node and is used to batch aggregate intelligent information that has been determined as ordinary information by the dynamic data classifier, quickly extract the overall main trend sequence, and generate visual input. In the embodiment, general information generally refers to data with strong stability, low rate of change, and low decision-making priority from scenarios such as environmental monitoring, urban management, and equipment operation, such as room temperature, regional voltage, and traffic flow. This type of data is characterized by a relatively stable update frequency, no sudden characteristics, and low timeliness requirements, making it suitable for periodic summary analysis. Structured aggregation processing is performed on this type of general information, and a weighted principal component analysis method is introduced in the first channel of the edge node. The specific steps are as follows: First, get The general information set within a time window is expressed as: To ensure the mathematical validity of principal component analysis, common information is pre-clustered according to the field structure. For common information with inconsistent field structures, field alignment is provided, and missing filling, field mapping or universal coding are supported to ensure that trend extraction is carried out in a unified field structure. Then, the covariance of the field and the subsequent common information set is calculated to obtain the covariance matrix , further extract the principal components of the covariance matrix to obtain the principal component results , and the corresponding unit eigenvector , indicating the direction of the principal axis, the number of eigenvalues ​​is equal to the number of fields of ordinary information, The principal component results are fused and the weighted coefficient of each principal component is defined as: in, It is The weight of each principal component indicates the contribution ratio of the principal component to the overall change, and the sum of all weights is 1; Mapping the general information set to the principal component space Calculate the weighted response of each common information in the direction of the principal component to form the overall main trend sequence. The formula is: in, Indicates general information The main trend response value has been reduced to one dimension; express General Information In the The projection in the direction of the principal component; the overall main trend sequence constitutes the final visual display information.

[0020] In the embodiment, the overall main trend sequence is displayed through three types of visualization methods, namely, The line chart shows the change curve of common information in a continuous period, with time as the horizontal axis and the main trend response value as the vertical axis. Regional comparison bar chart: if there are multiple common information sources (such as different regions or buildings), calculate the main trend mean separately and display the order of change intensity. Multi-period trend arrow chart, with The average main trend of a time window is used as a benchmark to draw rising, falling or stable symbol diagrams for status reminders.

[0021] The core goal of the second channel is to use deep neural networks combined with the time series modeling capabilities of the Transformer model to perform real-time streaming calculations on important information and drive dynamic visualization based on the calculation results. The second channel takes precedence over the first channel and processes important information with high priority to ensure the real-time and accuracy of event response. Get the important information set of time step t, expressed as: Using the self-attention mechanism in the Transformer model, the important information set is mapped to the query vector Q, key vector K and value vector V, which can be expressed as: in, and Represents the trained weight matrix, which is responsible for mapping the important information set to the corresponding query, key, and value space; calculates the similarity between the query vector and the key vector, and obtains the attention score matrix through the dot product , expressed as: in, Is the dimension of the vector, used to normalize the dot product result; the value vector is weighted averaged by the attention score matrix, expressed as: in, It is a temporal feature; Although the Transformer model already has powerful temporal modeling capabilities, combining it with LSTM (Long Short-Term Memory) or GRU (Gated Recurrent Unit) models to handle long-term dependencies can further improve real-time processing accuracy. Input the time series features into the LSTM layer for time series modeling. The output of LSTM It is a high-level representation of the important information at the current moment, expressed as: in, It is the hidden state of the previous moment; through the LSTM layer, the relationship between the current moment and the past period can be captured, which can better identify emergencies with long-term dependencies; Based on the isolation forest, the anomaly score is calculated and compared with the anomaly threshold to determine whether the important information at time step t is an anomaly. When the anomaly score is greater than the anomaly threshold, the important information at time step t is an anomaly.

[0022] In the embodiment, when it is determined that important information is abnormal, a corresponding display instruction is generated, which is used to update the display content in real time, specifically, Heatmaps are used to display the intensity distribution of important information (such as regional temperature and energy consumption). The color depth of different areas indicates the degree of information change. Dynamic particle flow is used to display the dynamic changes of important information. The speed and color of the particles reflect the severity and time changes of the events corresponding to the important information. Trend charts display long-term trends and abnormal sudden changes in the timeline, and are used to assist in determining the duration and future trends of events corresponding to important information.

[0023] The core objectives of the channel resource scheduling unit include dynamically allocating computing resources between the first and second channels based on the current load status; computing resources for the second channel (higher priority) must always be guaranteed, especially when the load is high; Dynamically adjust the channel resource allocation strategy through deep reinforcement learning, set the state space, action space and reward function, State Space Including current load , resource requirements 、 ; Specifically represents the upper limit of the total computing resources available to the current edge node, regardless of the channel, and is a unified resource constraint for edge nodes. Indicates the resource requirements of the first channel, that is, the computing resources currently required by the channel. Indicates the resource requirements of the second channel, that is, the computing resources of this channel have a higher priority; Action Space Defined as the allocation strategy for computing resources, that is, the ratio of resources allocated to the second channel and the first channel and ; Reward Function The optimization is based on two objectives: maximizing throughput and minimize priority violations , the function is defined as: in, and is the weight parameter; Throughput is defined as: Priority violation means that the computing resources of the second channel are not satisfied, which is expressed as: Combined with the reward function, the deep Q-learning algorithm is used to calculate the current state Choose the best action , that is, the computing resource allocation strategy. The core of Q-learning is to guide the strategy through Q value update. The formula is: in, It is the action at the present moment; is the discount factor; It is the Q value of the current state and action, which represents the expected long-term benefit of the strategy. Through the deep Q-learning algorithm, it can automatically learn how to allocate resources under different loads and optimize the scheduling strategy.

[0024] In an embodiment, the current state is 40, is 70, is 100, the current action is 0.4, is 0.6, the throughput is calculated as , which is 100, the priority violation is calculated as , that is, 0, weight parameter and If both are set to 0.5, the reward function is calculated as 50. According to the reward function and the current action, the system enters the next state and then executes the next action. The above process is repeated. With multiple iterations, the system learns the action that can best guarantee high-priority channel resources while maximizing throughput and avoiding violations of high-priority channel resources.

[0025] Through deep reinforcement learning algorithms, the allocation of computing resources is dynamically adjusted to achieve efficient and intelligent computing resource scheduling to meet the priority requirements of different channels.

[0026] M2, the difference visualization module, is deployed on the terminal device. The aggregation results of the first channel are displayed in the edge area, and the real-time streaming calculation results of the second channel are displayed in the core area. When there is a correlation between ordinary information and important information, a connection path is automatically generated. When the number of users accessing the connection path exceeds the set number, the automatic merging mechanism is triggered. The difference visualization module is responsible for displaying the aggregated results of the first channel and the real-time streaming calculation results of the second channel, and interacting between the two. This module is deployed on the terminal device to display the system calculation results in real time and provide user interaction. Specifically, The display area of ​​the difference visualization module is divided into two parts, the edge area and the core area. In the embodiment, the difference visualization module is a display screen, the edge area is the corners of the display screen and the bottom of the screen, and the core area is the center of the display screen and the top of the screen. The aggregation results of the first channel are displayed in the corners of the display screen and the bottom of the screen, mainly showing the statistical data and trends of general information (such as temperature, humidity, traffic flow, etc.); the real-time streaming calculation results of the second channel are displayed in the center of the display screen and the top of the screen, which are specifically used to show the dynamic changes of important information, such as emergencies, abnormal monitoring, emergency response, etc.

[0027] The purpose of generating connection paths is to achieve intelligent association between common information and important information, and to help users more intuitively understand the inherent connection between these two types of information. In order to determine whether there is a correlation between common information and important information, the mutual information algorithm is used to calculate the correlation between the two. Mutual information (MI) measures the dependency between two types of information. The formula is: in, is the joint probability distribution of common information and important information; and is the probability distribution of individual variables, setting the mutual information threshold ,when , then it is considered that there is a correlation between the two and a connection path is automatically generated; In an embodiment, when the mutual information value between common information and important information exceeds a set threshold, a connection path is automatically generated on the display screen interface. This path is displayed as a connection line, which can reflect the association between common information and important information.

[0028] By monitoring the frequency of user access to connection paths, it is determined which information is most important to the user. When the number of times a user accesses a connection path exceeds a set number, it is determined that the general information and important information in the connection are highly correlated and the user wishes to further merge and display them. In this embodiment, the number of clicks on a certain connection path by the user is recorded in real time. If the number of clicks exceeds 10, the merging mechanism is triggered. The automatic merging mechanism is designed to optimize the user's interactive experience and ensure that the displayed content can be efficiently merged in the case of large amounts of data and information interaction, reducing interface complexity and making the intelligent information display more logical and hierarchical. The merging mechanism combines the display areas of general information and important information into a unified view. The display ratio of the merged display area is automatically adjusted according to the suddenness of important information and the stability of general information. After the merger, the display ratio of the merged area is continuously optimized according to the user's historical interaction behavior (such as visit frequency, click pattern, etc.), and the display strategy is dynamically adjusted through reinforcement learning or online learning algorithms.

[0029] In the embodiment, intelligent information in intelligent building management is collected. Ordinary information such as regional temperature and humidity are aggregated and displayed through the first channel, while important information such as abnormal energy consumption and equipment failure are calculated and displayed in real time through the second channel. If there is a correlation between temperature and energy consumption, a connection path will be automatically generated to connect the change trends of the two. When the user clicks on the connection path between energy consumption and temperature multiple times, the automatic merging mechanism is triggered to display a unified interface that shows both the temperature change trend and the abnormal energy consumption. The user can view the correlation between the two in real time.

[0030] The intelligent information display system based on end-to-end collaboration proposed in the present invention adopts a dynamic data classifier to classify intelligent information to ensure that important information is processed first. By deploying a dual-channel processing module at the edge node, ordinary information and important information are processed separately, and computing resources are dynamically allocated. At the same time, combined with the difference visualization module, the aggregation results of ordinary information and important information are displayed on the terminal device, and cross-channel connection paths are generated to enhance the user's interactive experience.

[0031] In summary, the present invention significantly improves the system's real-time response capability and information processing efficiency by introducing a dual-channel processing mechanism, intelligent information classification, dynamic resource scheduling, and differential visualization display, solving the shortcomings of the existing technology in effectively processing high-priority information and intelligent display, and has significant technical advantages.

[0032] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. An intelligent information display system based on end-to-end collaboration, comprising several terminal devices and edge nodes. The terminal devices communicate with the edge nodes and collect intelligent information. The edge nodes have built-in dynamic data classifiers that classify the intelligent information into general information and important information. The system is characterized by: The system includes: M1, a dual-channel processing module, is deployed on edge nodes and includes a first channel, a second channel, and a channel resource scheduling unit. The second channel has a higher priority than the first channel. The first channel aggregates common information in batches, while the second channel performs real-time streaming calculations on important information. The channel resource scheduling unit dynamically allocates computing resources between the first and second channels based on the current load status. M2, the difference visualization module, is deployed on the terminal device. The aggregation results of the first channel are displayed in the edge area; the real-time streaming calculation results of the second channel are displayed in the core area; when the ordinary information and important information are associated, a connection path is automatically generated; when the number of times the user visits the connection path exceeds the set number, the automatic merging mechanism is triggered.

2. The intelligent information display system based on end-to-end collaboration according to claim 1 is characterized in that: The specific division steps of the dynamic data classifier are: 1) Calculate the information structure complexity, information mutation coefficient, and time sensitivity of the intelligent information. The information structure complexity is expressed as: in, Indicates the complexity of information structure; Indicates the number of information fields; Indicates the field nesting level; Indicates information volume; The information mutation coefficient is The mutation degree of intelligent information is calculated in a time window, which is expressed as: in, represents the information mutation coefficient; Represents intelligent information In the time window The mean within Represents a time window Standard deviation within; is a numerical stability term; The time sensitivity is expressed as: in, Indicates time sensitivity; Represents intelligent information The permissible delay from generation to the latest acceptable display time; 3) Calculate the importance of the intelligent information using the formula: in, Represents intelligent information the importance of 、 and is the weight factor; set the importance threshold ,when , indicating intelligent information If it is not satisfied, it is ordinary information.

3. The intelligent information display system based on end-to-end collaboration according to claim 1 is characterized in that: The first channel extracts the overall main trend sequence of the general information through the weighted principal component analysis method, specifically, Get The general information set within a time window is expressed as: Perform covariance calculation on the common information set to obtain the covariance matrix , The covariance matrix is ​​used to extract the principal components to obtain the principal component results and the corresponding unit eigenvectors, which are expressed as: The weight coefficient of the principal component is defined as follows: in, It is The weights of the principal components; Map the general information set to the principal component space and calculate the overall main trend sequence. The formula is: in, Indicates general information The main trend response value of Indicates general information In the The projection in the direction of the principal component; the overall main trend sequence constitutes the visualization display result of general information.

4. The intelligent information display system based on end-to-end collaboration according to claim 3 is characterized in that: The specific steps of the real-time streaming calculation of the second channel are: Get the important information set of time step t, expressed as: Using the self-attention mechanism in the Transformer model, the important information set is mapped to the query vector Q, key vector K and value vector V, which can be expressed as: in, and Represents the weight matrix of training; Calculate the attention score matrix , expressed as: in, is the dimension of the vector; The value vector is weighted averaged by the attention score matrix, expressed as: in, It is a temporal feature; The time series features are input into the LSTM layer and expressed as: in, is the hidden state at the previous moment; Based on the output of the isolation forest algorithm and the LSTM layer, an anomaly score of the important information is calculated, and a visualization result of the important information is constructed according to the anomaly score.

5. The intelligent information display system based on end-to-end collaboration according to claim 4 is characterized in that: The channel resource scheduling unit adjusts computing resources through deep reinforcement learning, sets the state space, action space and reward function, State Space Including current load , First channel resource requirements , Second channel resource requirements ; Action Space Defined as the allocation strategy for computing resources, the allocation strategy is the ratio of resources allocated to the second channel and the first channel and ; Reward Function Based on throughput , priority violation Perform optimization, defined as: in, and is the weight parameter; The throughput is defined as: The priority violation is defined as the computational resources of the second channel not being satisfied, expressed as: According to the current status Choose the best action , the formula is: in, It is the action at the present moment; is the discount factor; is the Q value of the current state and action; Take action After the state; Is in state The next action to be taken.

6. The intelligent information display system based on end-to-end collaboration according to claim 5 is characterized in that: The connection path is generated based on the mutual information value of common information and important information, and the formula is: in, is the joint probability distribution of common information and important information; and is the probability distribution of the individual variables; set the mutual information threshold ,when , then there is a correlation between ordinary information and important information, and a connection path is automatically generated.

Images