Multi-node remote control method supporting high resolution and related equipment

Abstract

The invention provides a multi-node remote control method supporting high resolution and related equipment, and relates to the field of electric digital data processing. The management control system serves as a core hub to receive and analyze a client instruction and then accurately forwards the client instruction to a target node, a remote control panel of the target node converts the instruction into an action to execute a corresponding operation, and the management control system collects feedback data after the operation is completed. Compared with the prior art, the framework avoids additional network delay caused by centralized resource allocation of similar thin clients. And in the aspect of high-resolution graph transmission, each node independently undertakes a task, so that the data round-trip transmission time is shortened, the operation real-time performance is ensured, and the image transmission is smooth without blockage and lag. Meanwhile, compared with the high cost of a special game server or workstation, the scheme realizes the cost optimization. In an application scene, the quality of a high-resolution picture is guaranteed, and high-quality, low-cost, efficient and stable remote control experience is provided for a user.

Description

Technical Field

[0001] This application relates to the field of digital data processing, and particularly to a multi-node remote control method and related devices supporting high resolution. Background Art

[0002] In the current era of rapid digital development, multi-node remote control systems play a crucial role in many fields, such as film and television special effects production, large-scale data simulation analysis, and remote game competitions. These fields often need to process massive amounts of data and high-resolution graphic image information, posing high requirements for the computing power, data transmission efficiency, and accuracy of remote control of the system.

[0003] Currently, to effectively handle the increasing demand for high-performance computing tasks, some suppliers use dedicated game servers or workstations to build FARM nodes to replace traditional ordinary PCs. Such dedicated devices are usually equipped with powerful CPUs and GPUs, and rely on their strong computing power to significantly improve the overall performance. For example, in the simulation operations of some large 3D game scenes or complex graphic rendering tasks, compared with ordinary PCs, they can process massive graphic data and complex calculation instructions more quickly, thus meeting the task scenarios with high requirements for computing resources to a certain extent. However, this solution is not perfect. Its acquisition cost is extremely high, often requiring a large amount of capital for equipment procurement, which is an insurmountable obstacle for individuals with limited budgets.

[0004] In addition, some manufacturers have introduced the solution of "thin clients". Its core principle is to use a centralized computing resource allocation mechanism to uniformly allocate and distribute the heavy computing tasks that originally needed to be independently borne by a single node, so as to reduce the pressure on a single node and optimize the resource utilization efficiency to a certain extent. However, this solution has caused new problems, namely additional network latency. During data transmission, due to frequent interaction with centralized computing resources, the round-trip transmission time of data has increased significantly. For application scenarios with high real-time requirements, such as online video editing and real-time game control, this will lead to problems such as lagging operation responses and frame stuttering, seriously affecting the user experience.

[0005] In summary, in terms of high-resolution graphic transmission, either the cost is too high or the real-time performance is poor, resulting in a poor user experience. Summary of the Invention

[0006] This application provides a multi-node remote control method and related devices supporting high resolution, which are used to reduce network latency and improve the user experience in the case of low cost in high-resolution graphic transmission.

[0007] In a first aspect, the present application provides a multi-node remote control method supporting high resolution, which is applied to a multi-node remote control system supporting high resolution. The multi-node remote control system supporting high resolution includes: a number of nodes, each node including a remote control board, and a management control system installed on one or more servers at a network center location. The method includes: the management control system receives an operation instruction issued by a client application; after parsing the operation instruction, the management control system forwards it to the corresponding target node; enabling the remote control board located on the target node to convert the operation instruction into an action, and the action is used for the target node to execute the corresponding operation; after the target node executes the corresponding operation, the management control system collects feedback data from the target node.

[0008] By adopting the above technical solution, the management control system, as the core hub, receives and parses the client instructions, and then accurately forwards them to the target node. The remote control board of the target node will convert the instructions into actions to execute the corresponding operations. After the operations are completed, the management control system collects the feedback data. Compared with the related technologies, this architecture avoids the additional network latency caused by the centralized resource allocation of similar "thin clients". In terms of high-resolution graphics transmission, each node independently undertakes tasks, reducing the data round-trip transmission time, ensuring the real-time nature of operations, and the image transmission is smooth without stuttering and lag. At the same time, compared with the high costs of dedicated game servers or workstations, this solution is based on ordinary node devices and realizes cost optimization through effective management and control. Moreover, through the effective monitoring of the nodes by the management control system and the targeted optimization of the remote control board, in application scenarios such as large 3D games, the high-resolution picture quality is guaranteed, and the system reliability and availability are significantly improved, providing users with a high-quality, low-cost and highly efficient and stable remote control experience.

[0009] In some embodiments in combination with some embodiments of the first aspect, after the step that the management control system receives an operation instruction issued by the client application, the method further includes: the management control system extracts the target node identifier, instruction type, and parameter information in the operation instruction; the management control system queries the historical operation instruction data and the corresponding network traffic change data of the target node from the database according to the target node identifier, and obtains the normal traffic and peak traffic of the target node under different instruction types; the management control system inputs the target node identifier, instruction type, and parameter information into the artificial intelligence-assisted decision-making model to obtain the traffic feature vector generated by the target node in response to the instruction type; the management control system compares the traffic feature vector with the normal traffic and peak traffic respectively; if the traffic feature vector is higher than the normal traffic but lower than the peak traffic, a traffic monitoring task is generated. The traffic monitoring task is that after the management control system collects feedback data from the target node, it extracts network traffic data at a preset time interval and compares it with the peak traffic. If the network traffic data is higher than the peak traffic, a traffic optimization strategy is started; the traffic optimization strategy includes finding an alternative transmission link or adjusting the data transmission protocol. If the traffic feature vector is higher than the peak traffic, the management control system issues a request for delaying instruction execution and starts the traffic optimization strategy at the same time.

[0010] By adopting the above technical solution, the management control system extracts the key information of the operation instruction and queries the historical data, generates the traffic feature vector through the artificial intelligence-assisted decision-making model and conducts comparative analysis. Through learning a large amount of historical data, the input layer of the artificial intelligence model can accurately receive various types of information, the hidden layer conducts effective feature extraction and transformation, and the output layer obtains the traffic feature vector. When the traffic feature vector is in different ranges, a traffic monitoring task can be generated or the traffic optimization strategy can be started in a targeted manner. This effectively avoids the impact of sudden network traffic on high-resolution data transmission, ensures stable data transmission, guarantees the continuity of high-resolution graphic transmission, makes the image display clear and coherent during the remote control process, improves the system's response ability to different instructions and traffic conditions, and further improves the overall efficiency and quality of remote control.

[0011] In some embodiments in combination with some embodiments of the first aspect, before the step of inputting the target node identifier, instruction type, and parameter information into the artificial intelligence-assisted decision-making model to obtain the traffic feature vector generated by the target node in response to the instruction type, the method further includes: obtaining historical operation instruction data, historical target node identifiers, historical instruction types, historical parameter information, and historical actual traffic feature vectors from a database; classifying and storing them according to the historical instruction type and historical target node to obtain a structured training sample set; training the artificial intelligence-assisted decision-making model with the training sample set, where the input layer of the artificial intelligence-assisted decision-making model includes a node identifier unit for receiving the historical target node identifier, a type coding unit for receiving the historical instruction type, and a parameter parsing unit for receiving the historical parameter information; the hidden layer includes multiple fully connected layers, and the ReLU activation function is used between the fully connected layers, and the number of neurons in the fully connected layers decreases sequentially; the output layer is a layer containing the same number of neurons as the dimension of the traffic feature vector; where, in each training iteration, sample data is randomly selected from the training sample set, and the historical target node identifier, historical instruction type, and historical parameter information are input into the artificial intelligence-assisted decision-making model. After being parsed and encoded by the input layer, the data sequentially passes through the fully connected layers of the hidden layer for feature extraction and transformation, and finally a predicted traffic feature vector is generated at the output layer.

[0012] By adopting the above technical solution, a structured training sample set is constructed using rich historical data to train the artificial intelligence-assisted decision-making model. During the training process, the sample data is parsed and encoded by each unit of the input layer, and complex feature extraction and transformation are performed in the fully connected layers of the hidden layer with the help of the ReLU activation function. Finally, a predicted traffic feature vector is generated at the output layer. By continuously comparing the error between the predicted vector and the actual vector and adjusting the model parameters, the model can accurately predict the traffic conditions under different instruction types and parameter information. In actual high-resolution multi-node remote control, the traffic changes can be predicted in advance, providing a reliable basis for network resource allocation and traffic control strategy formulation, ensuring the smoothness of the data transmission channel, and avoiding the jamming or interruption of high-resolution graphic transmission caused by inaccurate traffic prediction.

[0013] In some embodiments in combination with some embodiments of the first aspect, after the step of training the artificial intelligence-assisted decision-making model with the training sample set, the method further includes: calculating the error value between the predicted traffic feature vector and the actual traffic feature vector in the training sample set using the mean square error function; adjusting the parameters of each layer in the artificial intelligence-assisted decision-making model according to the error value, where the partial derivatives of the weight and bias parameters of the mean square error function are calculated, and then multiplied by the learning rate and momentum factor to update the weight and bias parameters; after a preset number of training iterations, check whether the mean square error converges to within a preset threshold; if it does not converge to within the preset threshold, continue training.

[0014] By adopting the above technical solution, the mean square error function is used to calculate and adjust the parameters of the artificial intelligence-assisted decision-making model according to the error. The mean square error function accurately quantifies the difference between the predicted traffic feature vector and the actual vector. By taking the partial derivatives of its weight and bias parameters and combining the learning rate and momentum factor to update the parameters, the model can be continuously optimized. During the preset number of training iterations, the convergence of the mean square error is checked to ensure the training effect of the model. After such a rigorous training and optimization process, when the model is applied to high-resolution multi-node remote control, it can more accurately adapt to different network environments and instruction situations, effectively allocate network resources, reduce the risk of high-resolution graphic transmission errors or delays caused by inaccurate models, and improve the overall stability and reliability of the system.

[0015] Combined with some embodiments of the first aspect, in some embodiments, the mean square error function is: In the formula, is the error value, is the th element in the actual traffic feature vector, is the th element in the predicted traffic feature vector; the function for adjusting the parameters of each layer in the artificial intelligence-assisted decision-making model is: In the formula, is the weight at the current moment, is the weight at the previous moment, is the weight at the moment before the previous moment, is the calculus, is the learning rate, is the momentum factor, is the bias parameter at the current moment, is the bias parameter at the previous moment, is the bias parameter at the moment before the previous moment.

[0016] By adopting the above technical solutions, the mean square error function and the adjustment parameter function provide precise quantification and optimization means for the training of the artificial intelligence assisted decision-making model. The mean square error function comprehensively reflects the overall deviation degree between the predicted traffic feature vector and the actual vector by summing and averaging the squared errors of each element, providing an accurate direction for model adjustment. The learning rate and momentum factor in the adjustment parameter function reasonably control the step size and inertia of parameter update, avoiding over-adjustment or slow convergence. In the high-resolution multi-node remote control system, this enables the model to quickly and stably learn the relationships between different instructions and traffic characteristics, so that in actual operation, it can efficiently provide a scientific basis for traffic control and instruction execution strategies, ensuring the efficiency and accuracy of high-resolution graphic transmission.

[0017] In combination with some embodiments of the first aspect, in some embodiments, after the step of the management control system collecting feedback data from the target node, the method further includes: updating the training samples according to the new operation instruction data, the new target node identifier, the new instruction type, the new parameter information, and the new actual traffic feature vector, where the proportion of the new data in the training samples does not exceed a preset proportion threshold; training the artificial intelligence assisted decision-making model with the updated training sample set; in each training iteration, the new data is used as the validation set.

[0018] By adopting the above technical solutions, the training samples are updated by using the new operation instruction data, etc., and the artificial intelligence assisted decision-making model is retrained. During the update process, the proportion of the new data is controlled not to exceed the threshold, and the new data is used as the validation set. This can not only enable the model to continuously learn new situations and adapt to system changes, but also avoid the instability of the model caused by too much new data. In the high-resolution multi-node remote control scenario, as the system continues to run and the business changes, the model can be continuously optimized in this way, always maintaining the accurate prediction ability for different instruction and traffic changes, ensuring the stable quality of high-resolution graphic transmission, and preventing the transmission effect from deteriorating due to the long-term operation of the system or business changes.

[0019] In combination with some embodiments of the first aspect, in some embodiments, if the traffic feature vector is higher than the peak traffic, the management control system marks the operation instruction as low priority; if the traffic feature vector is higher than the normal traffic but lower than the peak traffic, the management control system marks the operation instruction as medium priority; the management control system queues the operation instructions according to the instruction priority determination result.

[0020] By adopting the above technical solution, the management control system determines the priority of operation instructions based on the traffic feature vector and processes them in a queued manner. When the traffic feature vector is in different ranges, the instruction priorities are accurately marked. High-priority instructions are processed first, and medium-priority instructions are executed orderly under monitoring. This avoids the overoccupation of network resources by low-priority instructions, ensures that high-priority instructions related to high-resolution graphic transmission can be processed in a timely manner, and enables fast and efficient data transmission. In multi-node remote control, the real-time performance and smoothness of high-resolution graphic transmission are guaranteed, enabling image data to be transmitted between nodes preferentially and orderly, and improving the response speed and transmission efficiency of the entire system in high-resolution graphic processing scenarios.

[0021] In a second aspect, the present application provides a multi-node remote control system supporting high resolution. The multi-node remote control system supporting high resolution includes: one or more processors and a memory; the memory is coupled to the one or more processors, and the memory is used to store computer program code. The computer program code includes computer instructions, and the one or more processors call the computer instructions to cause the multi-node remote control system supporting high resolution to execute the method described in the first aspect and any implementation manner in the first aspect.

[0022] In a third aspect, the present application provides a computer program product containing instructions. When the computer program product runs on a multi-node remote control system supporting high resolution, it causes the multi-node remote control system supporting high resolution to execute the method described in the first aspect and any implementation manner in the first aspect.

[0023] In a fourth aspect, the present application provides a computer-readable storage medium including instructions. When the instructions run on a multi-node remote control system supporting high resolution, they cause the multi-node remote control system supporting high resolution to execute the method described in the first aspect and any implementation manner in the first aspect.

[0024] One or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages: 1. The management and control system receives and parses the client instructions as the core hub, and then accurately forwards them to the target node. The remote control board of the target node will convert the instructions into actions to perform the corresponding operations. After the operation is completed, the management and control system collects feedback data. Compared with related technologies, this architecture avoids the additional network delay caused by centralized resource allocation similar to "thin clients". In terms of high-resolution graphics transmission, each node independently undertakes tasks to reduce the round-trip data transmission time, ensure the real-time operation, and the image transmission is smooth without jamming and lag. At the same time, compared with the high cost of dedicated game servers or workstations, this solution is based on ordinary node devices and achieves cost optimization through effective management and control. In addition, through the effective monitoring of nodes by the management and control system and the targeted optimization of the remote control board, in application scenarios such as large-scale 3D games, high-resolution picture quality can be guaranteed, and the system reliability and availability are significantly improved, providing users with a high-quality, low-cost, efficient and stable remote control experience.

[0025] 2. The management and control system extracts key information of operation instructions and queries historical data, and uses the artificial intelligence-assisted decision-making model to generate traffic feature vectors and conduct comparative analysis. Through learning a large amount of historical data, the artificial intelligence model can accurately receive various types of information at its input layer, perform effective feature extraction and transformation at its hidden layer, and obtain traffic feature vectors at its output layer. When the traffic feature vector is in different ranges, it can generate traffic monitoring tasks or start traffic optimization strategies in a targeted manner. This effectively avoids the impact of network traffic bursts on high-resolution data transmission, ensures stable data transmission, and guarantees the continuity of high-resolution graphics transmission, making the image display clear and coherent during remote control, improving the system's ability to respond to different instructions and traffic conditions, and thus improving the efficiency and quality of overall remote control.

[0026] 3. Use rich historical data to build a structured training sample set to train the artificial intelligence-assisted decision-making model. During the training process, the sample data is parsed and encoded through each unit of the input layer, and complex feature extraction and transformation are performed in the fully connected layer of the hidden layer with the help of the ReLU activation function, and finally a predicted traffic feature vector is generated in the output layer. By constantly comparing the error between the predicted vector and the actual vector and adjusting the model parameters, the model can accurately predict the traffic conditions under different instruction types and parameter information. In this way, in actual high-resolution multi-node remote control, traffic changes can be predicted in advance, providing a reliable basis for network resource allocation and traffic control strategy formulation, ensuring the smoothness of the data transmission channel, and avoiding high-resolution graphics transmission jams or interruptions due to inaccurate traffic estimates. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 This is a flowchart of a multi-node remote control method supporting high resolution in an embodiment of the present application; Figure 2 It is another flowchart of the multi-node remote control method supporting high resolution in the embodiments of the present application; Figure 3 It is a specific flowchart of step S209 in the embodiments of the present application; Figure 4 It is an exemplary hardware structure diagram of the multi-node remote control system supporting high resolution in the embodiments of the present application. Detailed implementation manners

[0028] The terms used in the following embodiments of the present application are only for the purpose of describing specific embodiments, and are not intended to limit the present application. As used in the specification and appended claims of the present application, the singular forms "a", "an", "the", "above-mentioned", "said", and "this" are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term "and / or" used in the present application refers to and includes any or all combinations of one or more of the listed items.

[0029] Hereinafter, the terms "first" and "second" are only used for descriptive purposes and cannot be construed as implying or suggesting relative importance or implicitly indicating the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more of such features. In the description of the embodiments of the present application, unless otherwise stated, the meaning of "a plurality" is two or more.

[0030] A multi-node remote control method supporting high resolution is applied to a multi-node remote control system supporting high resolution. The multi-node remote control system supporting high resolution includes: a plurality of nodes, each node includes a remote control board, and a management control system installed on one or more servers at the network center location; the method includes: Before describing the multi-node remote control method supporting high resolution, the multi-node remote control system supporting high resolution will be described first.

[0031] Among them, a node represents a single computing unit in the system, and the remote control board is the hardware on the node for receiving and converting control instructions, and is the bridge connecting the management control system and the node. The management control system is the core control part on the network center server and is used to command and dispatch the work of each node.

[0032] Compared with related technologies, related dedicated game servers or workstations rely on high-end CPUs and GPUs to improve performance, resulting in high acquisition costs. This system is based on ordinary node hardware (such as conventional PCs), and through the cooperation of the management control system and the remote control board, it reduces the hardware cost and avoids high procurement expenditures. At the same time, although the centralized resource allocation of "thin clients" optimizes efficiency, there is network latency. Each node of this system is relatively independent, receives instructions from the management control system and executes them, reducing interaction latency and ensuring the real-time transmission of high-resolution graphics.

[0033] Please refer to Figure 1 , Figure 1 which is a schematic flowchart of a multi-node remote control method supporting high resolution in an embodiment of this application; S101. The management control system receives an operation instruction issued by the client application; Among them, the operation instruction refers to specific command information generated by the client application according to the user's operation behavior on the interface, and is used to inform the management control system of the specific task to be executed.

[0034] In some embodiments, when the user performs an operation related to remote control on the client application, the execution of step S101 is triggered. In this step, the client application first collects relevant parameters generated by the user operation, such as the function option corresponding to the click operation, the file path involved (if any), etc. Then, these information are integrated and encoded according to the preset communication protocol format to generate an operation instruction data packet. Next, the data packet is sent to the specified listening port of the server where the management control system is located through the established network connection (such as a network link based on the TCP / IP protocol). The management control system has been keeping a listening state on the corresponding listening port. Once it receives this data packet, it will first perform an integrity check on the data packet, such as through checksum verification and other methods, to ensure that the data has not been lost or damaged. Subsequently, the data packet is decoded, the operation instruction content therein is extracted, and it is converted into a format that can be recognized and processed within the management control system. Thus, step S101 is completed, preparing for subsequent instruction processing.

[0035] S102. After parsing the operation instruction, the management control system forwards it to the corresponding target node; Among them, the target node refers to the specific node individual that the operation instruction ultimately acts on in the entire multi-node remote control system supporting high resolution. Each target node has its unique network identifier (such as an IP address, etc.), and is used to receive the instruction forwarded by the management control system and execute the corresponding operation.

[0036] In some embodiments, when the management control system completes step S101 and successfully receives and extracts the operation instructions sent by the client application, it enters step S102. In this step, the management control system first calls its built-in instruction parsing module, which will perform a detailed parsing of the operation instructions according to the predefined instruction format specifications and syntax rules. It will accurately identify the unique identifier of the target node contained in the operation instructions, which can be the IP address of the node or the node number defined within the system, etc. At the same time, it parses out the specific operation type corresponding to the instruction, such as performing high-resolution graphics rendering, etc., and various parameter information required for the operation, such as the name and version of the software to be started, the specific content of the file to be modified, the resolution parameters for graphics rendering, etc. After the parsing is completed, the network forwarding module of the management control system will, according to the target node identifier information obtained from the parsing, query the network topology table or node information database maintained within the system to obtain information such as the detailed network address of the target node, the current network connection status, and the load situation of the node. Then, the management control system repackages the parsed operation instructions in a format that the target node can recognize and process, such as adding appropriate network protocol header information, setting the source address (the address of the management control system itself) and destination address (the target node address) of the data packet, etc. Finally, through network devices such as network switches or routers, according to the network address of the target node, the packaged instruction data packet is accurately forwarded to the corresponding target node to ensure that the instruction can be accurately delivered.

[0037] S103. Cause the remote control board located on the target node to convert the operation instruction into an action, where the action is used for the target node to execute the corresponding operation; Among them, the action refers to the specific signal or operation behavior directly generated by the remote control board according to the conversion of the operation instruction, which can directly act on the hardware device of the target node to prompt it to execute the corresponding task.

[0038] In some embodiments, when the target node receives the operation instruction data packet forwarded by the management control system, it starts to execute step S103. This step is applicable to various scenarios where a node needs to execute specific tasks through remote control, such as remotely controlling a node to perform high-resolution image rendering, remotely controlling a server node to perform data backup tasks, etc.

[0039] In this step, the remote control board on the target node first unpacks the received instruction data packet to extract the operation instruction content contained therein. Then, the control chip (such as a microprocessor, etc.) inside the remote control board will, according to the specific requirements of the operation instruction, look up the pre-stored instruction-action mapping table or execution rule library inside it. This mapping table or rule library records the correspondence between different types of operation instructions and the specific execution actions of the target node's hardware. For example, if the operation instruction is to start a certain high-resolution graphics rendering software and load a specific image for rendering, the control chip of the remote control board, by looking up the mapping table, will determine that it needs to first send a control signal to start the relevant rendering program to the node's CPU, and at the same time send control signals to the GPU to load the image data and set rendering parameters (such as resolution, color mode, etc.), and send a signal to allocate corresponding cache space to the memory, etc. Then, the control chip, according to the determined action sequence, sends corresponding electrical signals or control instructions to each hardware component of the target node through the corresponding hardware interface circuit (such as the PCIe interface for connecting the GPU, etc.), so that the hardware devices of the target node perform corresponding actions as required, thereby completing the tasks corresponding to the operation instructions, such as realizing specific operations such as high-resolution image rendering and accurate data backup.

[0040] S104. After the target node executes the corresponding operation, the management control system collects feedback data from the target node.

[0041] In some embodiments, when the target node has completed the operation tasks corresponding to the operation instructions in step S103, regardless of whether the task execution result is successful or there are some abnormal situations, it will enter step S104. This step is applicable to any scenario where it is necessary to master the task execution situation of the remote node for subsequent management and optimization. For example, in a remote scientific research computing task, check the accuracy of the node's calculation results, and in remote device control, confirm whether the device has completed the actions as required.

[0042] It can be seen that the management control system, as the core hub, receives and parses client instructions, and then accurately forwards them to the target nodes. The remote control board of the target nodes will convert the instructions into actions to perform corresponding operations. After the operations are completed, the management control system collects feedback data. Compared with related technologies, this architecture avoids the additional network latency caused by centralized resource allocation in similar "thin clients". In terms of high-resolution graphics transmission, each node independently undertakes tasks, reducing the data round-trip transmission time, ensuring the real-time nature of operations, and enabling smooth and lag-free image transmission. At the same time, compared with the high costs of dedicated game servers or workstations, this solution is based on ordinary node devices and realizes cost optimization through effective management and control. Moreover, through the effective monitoring of nodes by the management control system and the targeted optimization of the remote control board, in application scenarios such as large 3D games, high-resolution picture quality is guaranteed, and the system reliability and availability are significantly improved, providing users with a high-quality, low-cost, and efficient and stable remote control experience.

[0043] In the actual usage process, in the face of various operation instructions and complex and changeable network traffic conditions, it is difficult to accurately judge the trend of traffic changes and flexibly adjust strategies according to traffic characteristics, resulting in low efficiency when processing different instructions, being unable to guarantee the high efficiency and high quality of overall remote control, and being difficult to well meet the requirements of application scenarios with high requirements for transmission stability and timeliness such as high-resolution graphics processing.

[0044] Therefore, in some embodiments, after step S101, it further includes: Please refer to Figure 2 , Figure 2 which is another flowchart of the multi-node remote control method supporting high resolution in the embodiments of the present application; S201. The management control system extracts the target node identifier, instruction type, and parameter information in the operation instruction; The target node identifier is a kind of marking information used to uniquely distinguish different node individuals and represents the node on which the operation instruction specifically acts; the instruction type refers to the specific task category represented by the operation instruction and indicates the nature of the operation to be performed by the target node; the parameter information refers to the relevant data content attached to the operation instruction and used to further clarify the operation details, which varies according to different instruction types.

[0045] In some embodiments, corresponding regular expression templates are constructed for different types of operation instructions. For example, for instructions of the program startup type, the regular expression template is set to match the format of "start [software name] [version number (optional)]". After receiving an operation instruction, the instruction text is sequentially matched with each regular expression template. Once a match is successful, the target node identifier (extracted from the part of the instruction corresponding to the node address), the instruction type (determined by the type of the matched template), and the parameter information (obtained from the positions corresponding to the software name, version number, etc. in the capture group) are extracted according to the capture groups defined in the regular expression. Then, the information extracted is respectively stored in the corresponding variables or data structures inside the management control system for subsequent use; no specific limitation is made here.

[0046] S202. The management control system queries the historical operation instructions, data, and corresponding network traffic change data of the target node from the database based on the target node identifier, and obtains the normal traffic and peak traffic of the target node under different instruction types. In some embodiments, the management control system first constructs a query statement for the database by virtue of the target node identifier obtained in step S201. This query statement is written according to the type and structure of the database to find all the historical operation instruction records of the target node and the associated data and network traffic change data records generated during the execution of these instructions. Then, the management control system filters out the data subsets classified by instruction type from the query results, that is, groups the relevant data corresponding to the same type of operation instructions together. For example, all the historical data corresponding to the "file transfer" type instructions are placed together, and those of the "graphical rendering" type are placed in another group, etc. Next, for each data subset of the instruction type, the network traffic change data is analyzed, and through statistical analysis methods, such as calculating the average value, maximum value, etc., the normal traffic and peak traffic of the target node under this instruction type are determined. The normal traffic can be determined by calculating the average level of the network traffic during multiple executions of this instruction type, and the peak traffic is the maximum traffic value that appears during the execution of this instruction type. In this way, the normal traffic and peak traffic conditions of the target node under different instruction types can be obtained, providing a reference basis for subsequent judgment of the impact of the current operation instruction on the traffic.

[0047] In some specific embodiments, each time a file transfer instruction is executed, the network traffic fluctuates within a certain range. After multiple executions of the same type of instruction, the normal traffic of this instruction type can be obtained by calculating the average value. And among these fluctuating data, the maximum value is the peak traffic. Since the database records a sufficient number of historical situations, these historical data can reflect the typical traffic patterns of the target node under different instruction types.

[0048] S203. The management control system inputs the target node identifier, instruction type, and parameter information into the artificial intelligence-assisted decision-making model to obtain the traffic feature vector generated by the target node in response to the instruction type; S204. The management control system compares the traffic feature vector with the normal traffic and peak traffic respectively; S205. If the traffic feature vector is higher than the normal traffic but lower than the peak traffic, a traffic monitoring task is generated. The traffic monitoring task is that after the management control system collects feedback data from the target node, it extracts network traffic data at a preset time interval and compares it with the peak traffic. If the network traffic data is higher than the peak traffic, a traffic optimization strategy is started; the traffic optimization strategy includes finding an alternative transmission link or adjusting the data transmission protocol; Among them, the traffic feature vector is the dimensional data output by the artificial intelligence model, reflecting the traffic status of the target node's instruction response; the traffic optimization strategy is an adjustment means for dealing with abnormal traffic, such as switching links or protocol adjustment; the alternative transmission link is the system's standby network transmission channel, such as a standby network cable or wireless network; adjusting the data transmission protocol is to change the communication protocol used for data transmission, such as changing from TCP to UDP.

[0049] In some embodiments, the management control system first generates a traffic monitoring task and queues it, sends an instruction to the target node, and after receiving its feedback data, extracts the node traffic data at a preset time interval (such as 3 minutes) and compares it with the peak value. If it exceeds the peak value, first check for an alternative link. If there is a suitable one, switch; if not, adjust the data transmission protocol to stabilize the traffic and ensure the operation of the system.

[0050] S206. If the traffic feature vector is higher than the peak traffic, the management control system issues a delay instruction execution request and simultaneously starts the traffic optimization strategy.

[0051] Among them, the "delay instruction execution request" is a signal to let the node pause the instruction execution.

[0052] In some embodiments, the management control system generates a delay instruction execution request and sends it to the target node, and the node pauses the instruction. The management control system simultaneously starts the traffic optimization strategy, checks the system network resources, checks for alternative links or evaluates the feasibility of protocol adjustment, selects a suitable method to optimize the traffic, and then decides whether to let the node continue to execute the instruction.

[0053] In some specific embodiments, the management control system realizes the trigger of the state transition for the traffic exceeding the peak value and enters the delay and optimization state. First, send a delay request to the node. In the optimization state, call the link resource module to check for alternative links. If there is one, switch and monitor the effect. If not, call the protocol adjustment module to select a new protocol to switch and observe. After the traffic is normal, switch the state and notify the node to continue the execution; this is not limited here.

[0054] In some specific embodiments, the management and control system implements the traffic peak triggering rule according to the rule, and the rule engine executes the action. First, a delay request is sent to the node through the communication module, and then the alternative link resources are checked according to the predefined rules. If the conditions are met, the switch is switched. If not, a new protocol is selected according to the rules for adjustment. After the adjustment, the traffic is checked. If it is normal, the node is notified to resume execution; this is not limited here.

[0055] It can be seen that the management and control system extracts key information of operation instructions and queries historical data, and uses the artificial intelligence-assisted decision-making model to generate traffic feature vectors and conduct comparative analysis. Through learning a large amount of historical data, the artificial intelligence model can accurately receive various types of information at its input layer, perform effective feature extraction and transformation at its hidden layer, and obtain traffic feature vectors at its output layer. When the traffic feature vector is in different ranges, it can generate traffic monitoring tasks or start traffic optimization strategies in a targeted manner. This effectively avoids the impact of network traffic bursts on high-resolution data transmission, ensures stable data transmission, and guarantees the continuity of high-resolution graphics transmission, making the image display clear and coherent during remote control, improving the system's ability to respond to different instructions and traffic conditions, and thus improving the efficiency and quality of overall remote control.

[0056] See also Figure 2 , Figure 2 is another flowchart of a multi-node remote control method supporting high resolution in an embodiment of the present application; In some specific embodiments, before step S203, the method further includes: S207, obtaining historical operation instruction data, historical target node identifiers, historical instruction types, historical parameter information, and historical actual traffic feature vectors from a database; S208, classify and store historical instruction types and historical target nodes to obtain a structured training sample set; In some embodiments, this step is usually carried out in the process of preparing training data for an artificial intelligence-assisted decision-making model. In order to enable the model to better learn the traffic characteristics of different nodes when executing different types of instructions, the original acquired data needs to be reasonably structured. This is applicable to any scenario that requires the use of historical data training models to improve the accuracy of traffic predictions.

[0057] In this step, the management control system first creates appropriate data structures to store the upcoming structured training sample set. For example, in programming languages, data structures such as arrays, lists, dictionaries, or their combinations can be used. Then, the management control system traverses all the historical data records obtained in step S207. For each record, key contents such as the historical target node identifier, historical instruction type, historical parameter information, and historical actual traffic feature vector are extracted. Next, according to the two classification dimensions of historical instruction type and historical target node identifier, each data record is classified and placed into the corresponding category group. By classifying and organizing all data records in this way successively, a structured training sample set is finally formed, where data with the same type of instruction and from the same node is grouped together, facilitating the subsequent learning and feature extraction of the model by category.

[0058] S209. Train the artificial intelligence assisted decision-making model with the training sample set. The input layer of the artificial intelligence assisted decision-making model includes a node identifier unit for receiving the historical target node identifier, a type coding unit for receiving the historical instruction type, and a parameter parsing unit for receiving the historical parameter information; the hidden layer includes multiple fully connected layers, and the ReLU activation function is used between the fully connected layers, and the number of neurons in the fully connected layers decreases successively; the output layer is a layer containing the same number of neurons as the dimension of the traffic feature vector. Among them, in each training iteration, sample data is randomly selected from the training sample set, and the historical target node identifier, historical instruction type, and historical parameter information are input into the artificial intelligence assisted decision-making model. After being parsed and encoded by the input layer, the data passes through the fully connected layers of the hidden layer in turn for feature extraction and transformation, and finally a predicted traffic feature vector is generated at the output layer.

[0059] Among them, the input layer is the first layer of the artificial intelligence assisted decision-making model, which is used to receive externally input data. In this model, it includes a node identifier unit, a type coding unit, and a parameter parsing unit. These units are respectively responsible for receiving the corresponding historical target node identifier, historical instruction type, and historical parameter information, and introducing different types of data into the model for subsequent processing.

[0060] The hidden layer is the part of the artificial intelligence assisted decision-making model between the input layer and the output layer, and includes multiple fully connected layers. The ReLU activation function is used between the fully connected layers, and the number of neurons in the fully connected layers decreases successively. Its main function is to perform complex feature extraction and non-linear transformation on the data transmitted from the input layer. Through the connection between neurons and the action of the activation function, the traffic feature laws hidden behind the input data are mined. For example, through the successive calculation of multiple fully connected layers, the simple input data features are gradually transformed into deep features related to traffic features.

[0061] The fully connected layer is a basic building block in the hidden layer. Each neuron is connected to all neurons in the adjacent layer, which means that the output of each neuron in the previous layer serves as the input for each neuron in the next layer. By setting different weights and performing operations such as weighted summation, data transmission and feature transformation are achieved. It is like individual nodes in a huge neural network, working together to process data. For example, after the first fully connected layer receives the data from the input layer, new data features are output through the connection weights between neurons and passed to the next fully connected layer.

[0062] The output layer is the last layer of the artificial intelligence-assisted decision-making model, which is a layer containing the same number of neurons as the dimension of the traffic feature vector. Its function is to output the finally predicted traffic feature vector.

[0063] In some embodiments, data preprocessing operations are first performed on the training sample set. For example, the data in the sample set is normalized so that data in different dimensions is within a suitable numerical range, avoiding certain features dominating or being ignored in model training due to overly large or small numerical values, which may affect the training effect of the model. Then, the preprocessed training sample set is divided into a training set, a validation set, and a test set according to a certain ratio. Common division ratios can be 7:2:1 or 8:1:1, etc. The training set is used for model training, the validation set is used to evaluate the performance of the model and adjust the model's parameters during training, and the test set is used to finally examine the generalization ability of the model after training is completed. Next, the data in the training set is sequentially input into the input layer of the artificial intelligence-assisted decision-making model. The data first enters the node identification unit, type encoding unit, and parameter parsing unit for corresponding parsing and encoding processing. For example, the node identification unit converts the historical target node identification into a suitable numerical encoding, the type encoding unit performs one-hot encoding on the historical instruction type, etc., and the parameter parsing unit normalizes the historical parameter information. The processed information is sequentially passed to multiple fully connected layers in the hidden layer. In the fully connected layer, each layer of neurons performs a weighted summation operation on the output of the previous layer, and then undergoes a non-linear transformation through the ReLU activation function. Through successive calculations in multiple fully connected layers, data features are continuously extracted and transformed, and finally passed to the output layer. The output layer outputs the predicted traffic feature vector according to the settings of its neurons. In each training iteration, that is, in each round of the training process, sample data is randomly selected from the training sample set to repeat the above input and calculation process. As the number of iterations increases, the model continuously adjusts the weight parameters of the connections between layers, so that the error between the predicted traffic feature vector and the corresponding historical actual traffic feature vector in the sample set gradually decreases. For example, loss functions such as mean squared error (MSE) are used to measure the gap between the predicted value and the true value, and the weights of the model are updated by backpropagation based on the loss value through an optimization algorithm (such as the stochastic gradient descent algorithm, etc.), enabling the model to continuously learn the patterns in the data and improve the prediction accuracy. The entire training process will continue for multiple rounds of iteration until the performance metrics (such as accuracy, loss value, etc.) of the model on the validation set reach the preset requirements or there is no obvious improvement. At this time, the model training is completed and can be used to predict the traffic feature vector corresponding to new operation instructions in the future.

[0064] It can be seen that a structured training sample set is constructed using rich historical data to train an artificial intelligence-assisted decision-making model. During the training process, the sample data is parsed and encoded by the units of the input layer, and complex feature extraction and transformation are performed in the fully connected layer of the hidden layer with the help of the ReLU activation function. Finally, a predicted traffic feature vector is generated in the output layer. By continuously comparing the error between the predicted vector and the actual vector and adjusting the model parameters, the model can accurately predict the traffic conditions under different instruction types and parameter information. In this way, in actual high-resolution multi-node remote control, the traffic changes can be predicted in advance, providing a reliable basis for network resource allocation and traffic control strategy formulation, ensuring the smoothness of the data transmission channel, and avoiding the jamming or interruption of high-resolution graphic transmission caused by inaccurate traffic prediction.

[0065] Please refer to Figure 3 , Figure 3 which is a schematic diagram of the specific process of step S209 in the embodiment of the present application; In some embodiments, step S209 includes: S2091. Calculate the error value between the predicted traffic feature vector and the actual traffic feature vector in the training sample set using the mean square error function; The mean square error function is: In the formula, is the error value, is the th element in the actual traffic feature vector, is the th element in the predicted traffic feature vector; It should be noted that the mean square error function provides a standard and comprehensive way to measure the difference between the predicted value and the actual value. It squares the error in each dimension, which has two advantages. One is to amplify the error, so that even a small error can be reflected in the calculation result, making the model more concerned about these differences; the other is to avoid the situation where positive and negative errors cancel each other out. For example, the error that the predicted traffic is larger than the actual traffic and the error that the predicted traffic is smaller than the actual traffic will not cancel each other out when summing, so as to comprehensively and objectively reflect the inaccuracy of the prediction.

[0066] The calculated error value provides a clear direction for subsequent parameter adjustment. If the error value is large, it means that the current prediction effect of the model is not good, and the model parameters need to be adjusted significantly; if the error value is small, it means that the model is already relatively close to the actual situation and only needs fine-tuning.

[0067] S2092. Adjust the parameters of each layer in the artificial intelligence-assisted decision-making model according to the error value, where the partial derivatives of the weight and bias parameters of the mean square error function are calculated, and then multiplied by the learning rate and momentum factor to update the weight and bias parameters; The function for adjusting the parameters of each layer in the artificial intelligence-assisted decision-making model is: In the formula, is the weight at the current moment, is the weight at the previous moment, is the weight at the moment before the previous moment, is the calculus, is the learning rate, is the momentum factor, is the bias parameter at the current moment, is the bias parameter at the previous moment, is the bias parameter at the moment before the previous moment.

[0068] It should be noted that calculating the partial derivatives of the mean square error function with respect to the weight and bias parameters is the core operation based on the gradient descent algorithm. The partial derivative represents the rate of change of the mean square error function in the direction of each parameter, that is, it tells us how the mean square error changes when each parameter is changed. For example, if the partial derivative of a certain weight parameter is positive and has a large value, it means that increasing this weight will cause the mean square error to increase rapidly. Then, when updating the parameters, this weight should be decreased. In this way, the model can adjust the parameters along the direction where the mean square error decreases fastest to achieve the purpose of quickly optimizing the model.

[0069] The learning rate controls the step size of parameter adjustment. If the learning rate is too large, the model will skip the optimal parameter value during training, resulting in non-convergence or oscillating around the optimal value; if the learning rate is too small, the training speed of the model will be very slow, and a large number of iterations are required to achieve good results. The momentum factor, to a certain extent, takes into account the historical trend of parameter updates, making the parameter updates smoother. When the model encounters situations such as local minima or saddle points during training, the momentum factor can help the model utilize the previous update trend, jump out of these points that cause training stagnation, and continue to explore the more optimal parameter space, so as to more effectively find the global optimum or a parameter combination close to the global optimum, improving the convergence speed and stability of the model.

[0070] S2093. After a preset number of training iterations, check whether the mean square error converges within the preset threshold; S2094. If it does not converge within the preset threshold, continue training.

[0071] It can be seen that the mean squared error function is used for calculation and the parameters of the artificial intelligence-assisted decision-making model are adjusted based on the error. The mean squared error function accurately quantifies the difference between the predicted traffic feature vector and the actual vector. By taking the partial derivatives of its weight and bias parameters and combining the learning rate and momentum factor to update the parameters, the model can be continuously optimized. During the preset number of training iterations, the convergence of the mean squared error is checked to ensure the training effect of the model. Through such a rigorous training and optimization process, when the model is applied to high-resolution multi-node remote control, it can more accurately adapt to different network environments and instruction situations, effectively allocate network resources, reduce the risk of high-resolution graphic transmission errors or delays caused by inaccurate models, and improve the overall stability and reliability of the system.

[0072] It can be seen that the mean squared error function and the parameter adjustment function provide precise quantification and optimization means for the training of the artificial intelligence-assisted decision-making model. The mean squared error function comprehensively reflects the overall deviation degree between the predicted traffic feature vector and the actual vector by summing and averaging the squared errors of each element, providing an accurate direction for model adjustment. The learning rate and momentum factor in the parameter adjustment function reasonably control the step size and inertia of parameter update, avoiding over-adjustment or slow convergence. In the high-resolution multi-node remote control system, this enables the model to quickly and stably learn the relationship between different instructions and traffic features, so that in actual operation, it can efficiently provide a scientific basis for traffic control and instruction execution strategies, ensuring the efficiency and accuracy of high-resolution graphic transmission.

[0073] In some embodiments, after step S104, the method further includes: S301. Update the training sample according to the new operation instruction data, new target node identifier, new instruction type, new parameter information, and new actual traffic feature vector, where the proportion of the new data in the training sample does not exceed a preset proportion threshold; Wherein, the new operation instruction data, new target node identifier, new instruction type, new parameter information, and new actual traffic feature vector refer to the specific content records of the feedback data received by the newly generated target node after step S104 is completed; the preset proportion threshold is a preset upper limit value of a ratio for restricting the maximum proportion of new data in the entire training sample, ensuring that the updated sample will not be damaged due to excessive new data and the statistical laws of the original data and the features learned by the model.

[0074] S302. Train the artificial intelligence-assisted decision-making model with the updated training sample set; in each training iteration, the new data is used as the validation set.

[0075] It should be noted that the principle and process of this step are similar to those of step S209. The relevant principle and process can be referred to step S209 and will not be elaborated here.

[0076] Among them, the new data (specifically referring to the newly generated operation and traffic-related data updated to the training samples in step S301 here) is a data set reflecting the latest operation status of the system, including new operation instruction data, new target node identifiers, new instruction types, new parameter information, and new actual traffic feature vectors, etc. These are new situation information that has not been fully learned by the model. Using them as the validation set can test the model's ability to predict traffic characteristics in new scenarios.

[0077] In some embodiments, this step is mainly applied to scenarios where it is necessary to use newly generated data to accurately evaluate the performance of the model when facing new situations, and then reasonably adjust the model training strategy. For example, in a remote big data analysis system, with the emergence of new data analysis tasks, new data features appear. Using these new data as the validation set can timely discover whether the model can adapt to the traffic changes under the new tasks, thereby optimizing the model training. In this step, in each training iteration, first, the training set data divided from the updated training sample set is input into the artificial intelligence-assisted decision-making model, and the model performs normal training operations such as forward propagation, error calculation, and parameter adjustment. Then, taking the new data as the validation set, the new operation instruction data, new target node identifiers, new instruction types, new parameter information, etc. in the new data are sequentially input into the model being trained currently, and the model will output the corresponding predicted traffic feature vector. Next, a suitable evaluation metric (such as the mean square error function, etc.) is used to calculate the error between the predicted traffic feature vector and the new actual traffic feature vector included in the new data, so as to evaluate the performance of the model on these new data. For example, if the mean square error is small, it indicates that the model's traffic prediction for the new situation is relatively accurate, and the current training direction and parameter adjustment of the model are reasonable; if the mean square error is large, it indicates that the model has not well learned the rules in the new data, and operations such as adjusting the training parameters (such as reducing the learning rate to adjust the parameters more finely, etc.) or increasing the number of training rounds are required. By continuously validating with new data in each training iteration in this way, it is ensured that while the model learns the rules of historical data, it can also well adapt to the newly generated data situation, improving the generalization ability of the model and the accuracy of traffic prediction in new scenarios.

[0078] It can be seen that the training samples are updated using new operation instruction data, etc., and the artificial intelligence-assisted decision-making model is retrained. During the update process, the proportion of new data is controlled not to exceed the threshold, and the new data is used as the validation set. This can not only enable the model to continuously learn new situations and adapt to system changes, but also avoid the instability of the model caused by too much new data. In the high-resolution multi-node remote control scenario, as the system continues to run and the business changes, the model can be continuously optimized in this way, always maintaining the accurate prediction ability for different instruction and traffic changes, ensuring the stable quality of high-resolution graphic transmission, and preventing the transmission effect from degrading due to long-term system operation or business changes.

[0079] In some embodiments, it further includes: S401. If the traffic feature vector is higher than the peak traffic, the management control system marks the operation instruction as low priority; In some embodiments, when the management control system completes the analysis of the traffic feature vector related to the operation instruction and compares it with the peak traffic, and finds that the traffic feature vector is higher than the peak traffic, it will execute step S401. This is to avoid network congestion or even collapse in the remote control scenario caused by the execution of high-traffic instructions.

[0080] S402. If the traffic feature vector is higher than the normal traffic but lower than the peak traffic, the management control system marks the operation instruction as medium priority; In some embodiments, when the management control system compares the traffic feature vector with the normal traffic and the peak traffic and determines that the traffic feature vector is higher than the normal traffic but lower than the peak traffic, it will enter step S402. This step is commonly used in remote control scenarios that require a detailed priority differentiation of operation instructions to more reasonably allocate network resources, so that the system can operate smoothly under different load conditions. Through such priority marking, it can ensure the orderly execution of important and resource-consuming operations without affecting normal operations.

[0081] S403. The management control system queues the operation instructions according to the result of the instruction priority determination.

[0082] In some embodiments, when the management control system completes steps S401 and S402 and marks the corresponding priorities (low priority, medium priority, etc.) for each operation instruction, it will enter step S403. This step is applicable to any remote control scenario where there are multiple operation instructions to be executed and the execution order needs to be reasonably arranged according to factors such as the importance of the instructions and the demand for network resources to ensure the overall performance and stability of the system.

[0083] In this step, the management control system first creates or obtains a queue structure for storing operation instruction queuing information (which can be a data structure in memory, such as a queue implemented by a linked list, or by leveraging an external message queue system, etc.). Then, the management control system traverses all the operation instruction records with marked priorities (these records are stored in different places such as a database table, an instruction list in memory, etc.), and adds the operation instructions to the queue structure in order from highest to lowest priority (for example, high-priority first, medium-priority second, and low-priority last). For example, if it is a linked list queue in memory, for a high-priority operation instruction, it will be inserted at the head of the linked list, while low-priority instructions will be inserted at the tail of the linked list in sequence. After all the operation instructions are queued in priority order, the management control system takes out the operation instructions in the order of the queue, allocates corresponding system resources (such as network bandwidth, computing resources, etc.) in sequence, sends the operation instructions to the target node for execution, and monitors the instruction execution situation at the same time. After one instruction is executed, the next instruction is taken out from the queue for execution, and so on, to ensure that the operation instructions in the system can be executed orderly and efficiently.

[0084] It can be seen that the management control system determines the priorities of operation instructions based on the traffic feature vectors and performs queuing processing. When the traffic feature vectors are in different ranges, the instruction priorities are accurately marked. High-priority instructions are processed first, and medium-priority instructions are executed orderly under monitoring. This avoids the excessive occupation of network resources by low-priority instructions, ensures that high-priority instructions related to high-resolution graphic transmission can be processed in a timely manner, and enables fast and efficient data transmission. In multi-node remote control, it guarantees the real-time performance and smoothness of high-resolution graphic transmission, enables image data to be transmitted preferentially and orderly among nodes, and improves the response speed and transmission efficiency of the entire system in high-resolution graphic processing scenarios.

[0085] Next, an exemplary multi-node remote control system 400 supporting high resolution provided by the embodiments of the present application will be introduced. Figure 4 It is an exemplary hardware structure diagram of the multi-node remote control system 400 supporting high resolution provided by the embodiments of the present application.

[0086] In some embodiments, the multi-node remote control system 400 that supports high resolution is a computer device or the multi-node remote control system 400 that supports high resolution includes a computer device. The computer device includes a processor, a memory, and a network interface connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store data. The network interface of the computer device is used to communicate with other external terminals or servers through a network connection. In some embodiments, the network interface can be a wired network interface, and in some embodiments, the network interface can also be a wireless network interface. When the computer program is executed by the processor, it implements the method in the embodiments of the present application.

[0087] Those skilled in the art can understand that Figure 4 the structure shown in is only a block diagram of some structures related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0088] As mentioned above, the above embodiments are only used to illustrate the technical solutions of the present application, and are not intended to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application.

[0089] As used in the above embodiments, depending on the context, the term "when..." can be interpreted to mean "if..." or "after..." or "in response to determining..." or "in response to detecting...". Similarly, depending on the context, the phrase "when determining..." or "if detecting (the stated condition or event)" can be interpreted to mean "if determining..." or "in response to determining..." or "when detecting (the stated condition or event)" or "in response to detecting (the stated condition or event)".

[0090] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the processes or functions described in the embodiments of the present application are generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions may be transmitted from one website, computer, server, or data center to another website, computer, server, or data center by wire (such as coaxial cable, optical fiber, digital subscriber line) or wirelessly (such as infrared, wireless, microwave, etc.). The computer-readable storage medium may be any available medium that can be accessed by a computer or a data storage device such as a server or a data center that includes one or more integrated available media. The available medium may be a magnetic medium (such as a floppy disk, hard disk, magnetic tape), an optical medium (such as a DVD), or a semiconductor medium (such as a solid-state drive), etc.

[0091] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware with a computer program. The program can be stored in a computer-readable storage medium. When the program is executed, it may include the processes of the above method embodiments. The foregoing storage medium includes: various media that can store program codes such as ROM or random access memory RAM, magnetic disks, or optical discs.

Claims

1. A multi-node remote control method supporting high resolution, characterized in that: The invention is applied to a multi-node remote control system supporting high resolution, the multi-node remote control system supporting high resolution comprises: a plurality of nodes, each of which comprises a remote control panel, and a management control system installed on one or more servers at a network center; the method comprises: The management control system receives an operation instruction from a client application; The management and control system parses the operation instruction and forwards it to the corresponding target node; enabling a remote control panel located on the target node to convert the operation instruction into an action, wherein the action is used for the target node to perform a corresponding operation; After the target node performs a corresponding operation, the management control system collects feedback data from the target node.

2. The method according to claim 1, characterized in that: After the management control system receives the operation instruction issued by the client application, the method further includes: The management and control system extracts the target node identifier, instruction type and parameter information in the operation instruction; The management control system queries the historical operation instruction data and corresponding network traffic change data of the target node from the database according to the target node identifier, and obtains the normal traffic and peak traffic of the target node under different instruction types; The management and control system inputs the target node identifier, the instruction type and the parameter information into an artificial intelligence-assisted decision-making model to obtain a traffic feature vector generated by the target node in response to the instruction type; The management and control system compares the flow characteristic vector with the normal flow and the peak flow respectively; If the traffic characteristic vector is higher than the regular traffic but lower than the peak traffic, a traffic monitoring task is generated. The traffic monitoring task is that after the management and control system collects feedback data from the target node, it extracts network traffic data at a preset time interval and compares it with the peak traffic. If the network traffic data is higher than the peak traffic, a traffic optimization strategy is initiated; the traffic optimization strategy includes finding an alternative transmission link or adjusting the data transmission protocol; If the traffic characteristic vector is higher than the peak traffic, the management control system issues a delay instruction execution request and simultaneously starts the traffic optimization strategy.

3. The method according to claim 2, characterized in that Before the step of inputting the target node identifier, the instruction type and the parameter information into an artificial intelligence-assisted decision-making model by the management and control system to obtain a traffic feature vector generated by the target node in response to the instruction type, the method further includes: Acquire historical operation instruction data, historical target node identifiers, historical instruction types, historical parameter information, and historical actual traffic feature vectors from the database; Classify and store the historical instruction types and historical target nodes to obtain a structured training sample set; The training sample set is used to train the artificial intelligence-assisted decision-making model, wherein the input layer of the artificial intelligence-assisted decision-making model includes a node identification unit for receiving the historical target node identification, a type encoding unit for receiving the historical instruction type, and a parameter parsing unit for receiving the historical parameter information; the hidden layer includes a plurality of fully connected layers, and the ReLU activation function is used between the fully connected layers, and the number of neurons in the fully connected layers decreases in sequence; the output layer is a layer including the same number of neurons as the dimension of the traffic feature vector; In each training iteration, sample data is randomly extracted from the training sample set, and the historical target node identifier, the historical instruction type and the historical parameter information are input into the artificial intelligence assisted decision-making model. After parsing and encoding at the input layer, the data is sequentially extracted and transformed through the fully connected layer of the hidden layer, and finally the predicted traffic feature vector is generated at the output layer.

4. The method according to claim 3, characterized in that After the step of training the artificial intelligence-assisted decision-making model with the training sample set, the method further includes: Using a mean square error function to calculate the error value between the predicted flow feature vector and the actual flow feature vector in the training sample set; According to the error value, adjusting the parameters of each layer in the artificial intelligence-assisted decision-making model, wherein the partial derivatives of the weight and bias parameters of the mean square error function are calculated, and then multiplied by the learning rate and momentum factor to update the weight and bias parameters; After a preset number of training iterations, check whether the mean square error converges to within a preset threshold; If it does not converge to within the preset threshold, continue training.

5. The method according to claim 4, characterized in that The mean square error function is: In the formula, is the error value, is the first elements, is the predicted traffic feature vector elements; The function for adjusting the parameters of each layer in the artificial intelligence assisted decision-making model is: In the formula, is the weight at the current moment, is the weight at the previous moment, is the weight of the previous moment, For calculus, is the learning rate, is the momentum factor, is the bias parameter at the current moment, is the bias parameter of the previous moment, is the bias parameter at the previous moment.

6. The method according to claim 3, characterized in that After the step of the management control system collecting feedback data from the target node, the method further comprises: Update the training sample according to the new operation instruction data, the new target node identifier, the new instruction type, the new parameter information and the new actual traffic feature vector, wherein the proportion of the new data in the training sample does not exceed a preset proportion threshold; Training the artificial intelligence-assisted decision-making model with the updated training sample set; In each training iteration, the new data is used as a validation set.

7. The method according to claim 2, characterized in that If the traffic characteristic vector is higher than the peak traffic, the management control system marks the operation instruction as a low priority; If the traffic characteristic vector is higher than the normal traffic but lower than the peak traffic, the management control system marks the operation instruction as medium priority; The management control system queues the operation instructions according to the instruction priority determination result.

8. A multi-node remote control system supporting high resolution, characterized in that: The multi-node remote control system supporting high resolution includes: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code includes computer instructions, and the one or more processors call the computer instructions to enable the multi-node remote control system supporting high resolution to execute the method described in any one of claims 1-7.

9. A computer program product comprising instructions, characterized in that When the computer program product runs on a multi-node remote control system supporting high resolution, the multi-node remote control system supporting high resolution executes the method according to any one of claims 1 to 7.

10. A computer-readable storage medium comprising instructions, characterized in that: When the instructions are executed on a multi-node remote control system supporting high resolution, the multi-node remote control system supporting high resolution executes the method according to any one of claims 1 to 7.

Images