A cloud architecture distributed data transmission control method and system

Abstract

The application relates to the technical field of video transmission, and discloses a cloud architecture distributed data transmission control method and system. The method comprises the following steps: collecting a real-time state vector of a target video stream and attaching a collection timestamp to obtain a terminal state observation sequence, and delivering the terminal state observation sequence to a cloud control node; determining an actual uplink time delay, combining historical downlink time delay data to obtain a predicted uplink time delay and a predicted downlink time delay; performing state extrapolation on the real-time state vector, the predicted uplink time delay and the predicted downlink time delay to obtain a predicted state estimate; mapping the predicted state estimate into a coding control quantity, and constructing an effective tolerance time window; comparing a current execution time with the effective tolerance time window; calculating a prediction error of an actual running state and the predicted state estimate to correct parameters of the state extrapolation online, and feeding back the corrected parameters to the state extrapolation operation; and the application can improve the accuracy of distributed data transmission control.

Description

Technical Field

[0001] This invention relates to the field of video transmission technology, and in particular to a distributed data transmission control method and system based on a cloud architecture. Background Technology

[0002] Currently, in distributed video data transmission scenarios within a cloud architecture, control decisions are typically generated in the cloud and distributed to edge terminals via network links for execution. Existing technologies primarily focus on optimizing aspects such as adaptive bitrate, congestion control, and network state prediction, attempting to improve the stability and smoothness of video transmission by enhancing network state awareness or introducing trend prediction models. However, these solutions generally rely on an implicit premise: that control commands can be executed immediately and remain effective upon reaching the terminal.

[0003] For example, existing technologies often rely on single-dimensional network bandwidth estimation or buffer occupancy rates to adjust encoding parameters, failing to comprehensively collect the encoding parameters, network status indicators, and historically executed control variables of the target video stream. This makes it difficult to form a terminal state observation sequence with temporal semantics, resulting in an incomplete characterization of the actual transmission state. Simultaneously, existing technologies mostly employ fixed estimations or simple trend predictions to handle link latency, without combining actual uplink latency with historical downlink latency for dynamic extrapolation, leading to discrepancies between predictions and actual link changes. Furthermore, existing technologies neglect a fundamental issue at the control execution level: due to network jitter, queuing latency, and link fluctuations, control commands generated in the cloud experience time drift during transmission to the terminal. By the time the control command arrives at the terminal, its designed effective time may have deviated from the current actual transmission state, causing the control command to be no longer suitable for the current link environment. In this situation, even if the control strategy itself is reasonable, inappropriate execution timing can lead to control failure.

[0004] Therefore, existing technologies cannot fully meet the demand for precise control of distributed video data transmission under dynamically changing network latency conditions. There is an urgent need for a cloud-based technology that can combine latency prediction results with state extrapolation mechanisms to determine whether control commands are still valid at the time of execution, constrain their execution timing through time tolerance, and introduce an online correction mechanism for prediction errors to improve the accuracy and stability of distributed video data transmission control. Summary of the Invention

[0005] The main purpose of this application is to provide an office automation collaboration system based on multimodal artificial intelligence, which aims to solve the technical problem that the control method in the prior art, which allows the control command to be executed immediately upon arrival at the terminal and is always effective, especially under conditions of severe network jitter and dynamic changes in link latency, cannot achieve timing matching between the control command and the actual link state.

[0006] To achieve the above objectives, the present invention provides a distributed data transmission control method for a cloud architecture, comprising: S1. Collect the real-time state vector of the target video stream and attach a collection timestamp to obtain the terminal state observation sequence of the target video stream, and send the terminal state observation sequence to the cloud control node of the target video stream; S2. Based on the time information in the terminal status observation sequence, determine the actual uplink latency of the target video stream, and combine it with the historical downlink latency data of the target video stream to obtain the predicted uplink latency and predicted downlink latency of the target video stream; S3. Extrapolate the real-time state vector, the predicted uplink delay, and the predicted downlink delay to obtain the predicted state estimate of the target video stream; S4. Map the predicted state estimate to the coding control quantity of the target video stream, and construct an effective tolerance time window for the target video stream based on the predicted downlink delay; S5. When the cloud control node receives the coded control quantity, it compares the current execution time with the effective tolerance time window; S6. Obtain the actual operating state of the next sampling period based on the comparison results, calculate the prediction error between the actual operating state and the predicted state estimate, correct the parameters of the state extrapolation online, and feed back the corrected parameters to S3.

[0007] Optionally, the step of acquiring the real-time state vector of the target video stream and appending an acquisition timestamp to obtain the terminal state observation sequence of the target video stream, and transmitting the terminal state observation sequence to the cloud control node of the target video stream, includes: Obtain the current encoding parameters and network status indicators of the target video stream; Based on the current encoding parameters and the network status indicators, and combined with the encoding control quantities executed in the previous control cycle of the target video stream, the real-time status vector of the target video stream is obtained. A corresponding acquisition timestamp is appended to the real-time state vector to generate a terminal state observation sequence of the target video stream, and the terminal state observation sequence is sent to the cloud control node of the target video stream.

[0008] Optionally, the step of determining the actual uplink latency of the target video stream based on the time information in the terminal state observation sequence, and combining it with the historical downlink latency data of the target video stream to obtain the predicted uplink latency and predicted downlink latency of the target video stream, includes: The acquisition timestamp of the new state vector in the terminal state observation sequence and the corresponding receiving time of the cloud control node are obtained to obtain the actual uplink delay of the current period; Extract the historical downlink latency data of the target video stream to construct the historical downlink latency sequence of the target video stream; Based on the historical downlink latency sequence, the predicted downlink latency of the target video stream is calculated; Based on the actual uplink latency, the predicted uplink latency of the target video stream is calculated.

[0009] Optionally, the formulas for calculating the predicted downlink latency and the predicted uplink latency are as follows: ; In the formula, For the first Predicted uplink latency for each control cycle For the first One predictive coefficient, For actual uplink latency, To control the cycle number, The preset quantity for prediction; ; In the formula, For the first Predicted downlink latency for each control cycle, The preset quantity for prediction, For the first One predictive coefficient, For historical downlink latency, This is for controlling the cycle number.

[0010] Optionally, the step of extrapolating the real-time state vector, the predicted uplink delay, and the predicted downlink delay to obtain the predicted state estimate of the target video stream includes: Based on the predicted uplink latency and the predicted downlink latency, the end-to-end latency of the target video stream is determined; Obtain the state transition relationships that characterize the state evolution law in the target video stream; The state transition relationship is adjusted based on the full-link time delay to obtain the adjusted state transition relationship of the target video stream; Using the real-time state vector as input, the adjusted state transition relationship is forward extrapolated to obtain the predicted state estimate of the target video stream.

[0011] Optionally, mapping the predicted state estimate to the coding control quantity of the target video stream, and constructing an effective tolerance time window for the target video stream based on the predicted downlink delay, includes: The predicted state estimate is input into a preset coding strategy mapping table for matching and querying to obtain the coding parameter set of the target video stream; Based on the set of encoding parameters, the encoding control quantity of the target video stream is generated, and the encoding control quantity is sent to the encoding instruction queue of the cloud control node; Based on the predicted downlink latency, the estimated arrival time of the target video stream from the cloud control node to the terminal node is determined; Based on the expected arrival time period, the first buffer duration is extended forward and the second buffer duration is extended backward to obtain the effective tolerance time window of the target video stream.

[0012] Optionally, when the cloud control node receives the coded control quantity, comparing the current execution time with the effective tolerance time window includes: Extract the encoding control quantity and the effective tolerance time window from the encoding instruction queue, and obtain the current execution time of the target video stream; Identify the permissible execution start boundary and permissible execution end boundary of the effective tolerance time window; The current execution time is logically determined in relation to the allowed execution start boundary and the allowed execution end boundary.

[0013] Optionally, the step of performing logical relationship determination between the current execution time and the allowed execution start boundary and the allowed execution end boundary includes: Logically determine the relationship between the current execution time and the allowed execution start boundary and the allowed execution end boundary; When the logical relationship judgment result indicates that the current execution time is earlier than the allowable execution start boundary, a waiting execution instruction for the target video stream is generated; When the logical relationship judgment result indicates that the current execution time is between the allowed execution start boundary and the allowed execution end boundary, an immediate execution instruction for the target video stream is generated; When the logical relationship judgment result indicates that the current execution time exceeds the allowable execution deadline boundary, a new control quantity instruction for the target video stream is generated.

[0014] Optionally, the step of obtaining the actual operating state of the next sampling period based on the comparison result, calculating the prediction error between the actual operating state and the predicted state estimate, correcting the parameters of the state extrapolation online, and feeding back the corrected parameters to S3 includes: Based on the comparison results, the actual operating status corresponding to the coded control quantity is obtained from the execution terminal corresponding to the cloud control node; The actual operating state is matched and compared with the predicted state estimate to obtain the state difference between the actual operating state and the predicted state estimate; Based on the direction and magnitude of the state difference, a parameter correction instruction for the target video stream is generated; The parameter correction instructions are parsed to obtain the target parameters and adjustment trends of the target video stream; The adjustment trend is applied to the target parameter to update the parameters used for state extrapolation; The updated parameters are used as the corrected parameters of the target video stream, and the corrected parameters are fed back to S3.

[0015] To address the aforementioned problems, the present invention also provides a distributed data transmission control system based on a cloud architecture, the system comprising: The acquisition and transmission module is used to acquire the real-time state vector of the target video stream and attach an acquisition timestamp to obtain the terminal state observation sequence of the target video stream, and transmit the terminal state observation sequence to the cloud control node of the target video stream. The latency prediction module is used to determine the actual uplink latency of the target video stream based on the time information in the terminal state observation sequence, and to obtain the predicted uplink latency and predicted downlink latency of the target video stream by combining the historical downlink latency data of the target video stream. The state prediction module is used to extrapolate the real-time state vector, the predicted uplink delay, and the predicted downlink delay to obtain a predicted state estimate of the target video stream. The control generation module is used to map the predicted state estimate to the coding control quantity of the target video stream, and to construct an effective tolerance time window for the target video stream based on the predicted downlink delay. The time window comparison module is used to compare the current execution time with the effective tolerance time window when the cloud control node receives the coded control quantity; The error correction module is used to obtain the actual operating state of the next sampling period based on the comparison results, calculate the prediction error between the actual operating state and the predicted state estimate, correct the parameters of the state extrapolation online, and feed back the corrected parameters to the step state prediction module.

[0016] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention provides an efficient control foundation for edge data transmission through precise data acquisition and latency prediction. It collects the encoding parameters, network status indicators, and historical control execution variables of the target video stream, generates a timestamped terminal status observation sequence, and transmits it to the cloud control node. Combining actual uplink latency and historical downlink latency data, it accurately calculates and predicts uplink and downlink latency, comprehensively capturing the spatiotemporal characteristics of edge data transmission and providing precise data support for subsequent control decisions.

[0017] 2. This invention significantly improves the control accuracy and stability of edge data transmission by leveraging intelligent extrapolation and closed-loop optimization. It obtains predicted state estimates through state extrapolation and maps them to coded control quantities. An effective tolerance time window is constructed based on the predicted downlink delay to adapt the execution timing. Combined with online error correction parameters between the actual operating state and the predicted value, a dynamic optimization closed loop is formed, continuously optimizing the control logic for edge data transmission. This ensures the smoothness and reliability of distributed data transmission under a cloud architecture, meeting the requirements for high-quality video stream transmission. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating a distributed data transmission control method for a cloud architecture, as provided in an embodiment of the present invention.

[0019] Figure 2 This is a functional block diagram of a cloud-based distributed data transmission control system according to an embodiment of the present invention.

[0020] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0021] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0022] This application provides a distributed data transmission control method for a cloud architecture. The executing entity of this distributed data transmission control method for a cloud architecture includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this application: a server, a terminal, etc. In other words, this distributed data transmission control method for a cloud architecture can be executed by software or hardware installed on a terminal device or a server device. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster. The server can be an independent server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0023] Reference Figure 1 The diagram shown is a flowchart illustrating a distributed data transmission control method for a cloud architecture according to an embodiment of the present invention. In this embodiment, the distributed data transmission control method for a cloud architecture includes: S1. Collect the real-time state vector of the target video stream and attach a collection timestamp to obtain the terminal state observation sequence of the target video stream, and send the terminal state observation sequence to the cloud control node of the target video stream; In this embodiment of the invention, the step of acquiring the real-time state vector of the target video stream and appending an acquisition timestamp to obtain the terminal state observation sequence of the target video stream, and transmitting the terminal state observation sequence to the cloud control node of the target video stream, includes: Obtain the current encoding parameters and network status indicators of the target video stream; Based on the current encoding parameters and the network status indicators, and combined with the encoding control quantities executed in the previous control cycle of the target video stream, the real-time status vector of the target video stream is obtained. A corresponding acquisition timestamp is appended to the real-time state vector to generate a terminal state observation sequence of the target video stream, and the terminal state observation sequence is sent to the cloud control node of the target video stream.

[0024] The built-in monitoring module of the video streaming terminal collects the current encoding parameters of the target video stream in real time. These parameters cover core information that directly affects the video encoding effect, such as video compression format, image resolution, and frame rate. At the same time, network status indicators are collected simultaneously, including key content reflecting network transmission quality such as data transmission rate, latency, and packet loss rate, ensuring that the acquired parameters and indicators are comprehensive and accurate.

[0025] The system retrieves the encoding control quantities executed in the previous control cycle of the target video stream. These control quantities represent the specific operations performed in the previous cycle to adjust the encoding parameters. The system then integrates the current encoding parameters, network status indicators, and the encoding control quantities from the previous control cycle. By analyzing the interrelationships among these three elements, the system identifies a comprehensive set of core information reflecting the current operational status of the target video stream, thereby obtaining the real-time status vector of the target video stream.

[0026] By leveraging the terminal's time synchronization function, standard time information at the moment of real-time state vector acquisition is obtained. This time information is then appended as an acquisition timestamp to the corresponding real-time state vector, giving each real-time state vector a clear time stamp. Multiple timestamped real-time state vectors are arranged in chronological order of acquisition time to generate a terminal state observation sequence for the target video stream. Subsequently, this terminal state observation sequence is transmitted completely to the cloud control node of the target video stream via a stable network transmission channel, ensuring the timeliness and integrity of data transmission.

[0027] The beneficial effects are that it can comprehensively obtain the current encoding parameters and network status indicators of the target video stream, covering core information such as encoding format, resolution, transmission rate, and latency, ensuring that the basic operating conditions of the video stream are fully described, and providing complete data support for the subsequent construction of state vectors.

[0028] By combining the coded control quantities executed in the previous control cycle, the current parameters, network indicators, and historical execution feedback are linked and integrated to eliminate the limitations of single-dimensional data. The constructed real-time state vector can comprehensively reflect the dynamic operating status of the video stream and improve the accuracy of state representation.

[0029] Adding a collection timestamp to the real-time state vector ensures that each state data has a clear time mark, guaranteeing the temporality and traceability of the terminal state observation sequence, and providing a reliable time reference for subsequent delay calculation and state evolution analysis.

[0030] The terminal status observation sequence is generated in the order of collection time and transmitted to the cloud control node, realizing efficient transmission and centralized management of edge status data to the cloud. This allows the cloud to monitor the dynamic operation of the video stream in real time, providing timely and comprehensive data for subsequent control decisions.

[0031] The entire process forms a complete link from data acquisition and status integration to sequence transmission, ensuring that the status data obtained by the cloud control node is comprehensive, accurate, and has a clear time sequence, providing a solid foundation for subsequent delay prediction, status extrapolation, and coded control, and improving the scientificity and reliability of distributed data transmission control.

[0032] S2. Based on the time information in the terminal status observation sequence, determine the actual uplink latency of the target video stream, and combine it with the historical downlink latency data of the target video stream to obtain the predicted uplink latency and predicted downlink latency of the target video stream; In this embodiment of the invention, the step of determining the actual uplink latency of the target video stream based on the time information in the terminal state observation sequence, and obtaining the predicted uplink latency and predicted downlink latency of the target video stream by combining the historical downlink latency data of the target video stream, includes: The acquisition timestamp of the new state vector in the terminal state observation sequence and the corresponding receiving time of the cloud control node are obtained to obtain the actual uplink delay of the current period; Extract the historical downlink latency data of the target video stream to construct the historical downlink latency sequence of the target video stream; Based on the historical downlink latency sequence, the predicted downlink latency of the target video stream is calculated; Based on the actual uplink latency, the predicted uplink latency of the target video stream is calculated.

[0033] The formulas for calculating the predicted downlink latency and the predicted uplink latency are as follows: ; In the formula, For the first Predicted uplink latency for each control cycle For the first One predictive coefficient, For actual uplink latency, To control the cycle number, The preset quantity for prediction; ; In the formula, For the first Predicted downlink latency for each control cycle, The preset quantity for prediction, For the first One predictive coefficient, For historical downlink latency, This is for controlling the cycle number.

[0034] The acquisition timestamp attached to the new state vector is extracted from the terminal state observation sequence. This timestamp is the standard time recorded when the new state vector is acquired by the video stream transmission terminal. At the same time, the corresponding reception time when the cloud control node successfully receives the new state vector is recorded. By calculating the time difference between the reception time and the acquisition timestamp, the actual uplink latency of the current period of the target video stream is directly obtained, ensuring that the latency data can truly reflect the time consumption of the current uplink transmission.

[0035] By using the historical data storage module of the cloud control node, all downlink latency data recorded during the past transmission of the target video stream are retrieved. These data are latency records generated during the transmission of the video stream from the cloud to the terminal in each control cycle. All extracted historical downlink latency data are arranged in chronological order to form a complete and orderly historical downlink latency sequence of the target video stream, providing comprehensive historical data support for subsequent predictions.

[0036] We systematically sort out all the data in the historical downlink latency sequence, analyze the data's changing trends, fluctuation ranges and stable intervals, identify the key factors affecting downlink latency and the patterns of data changes, judge the possible changes in subsequent downlink latency based on these patterns, and then calculate the predicted downlink latency of the target video stream that can reflect the future downlink transmission time consumption, ensuring that the prediction results are consistent with the actual transmission scenario.

[0037] By combining the actual uplink latency of the current cycle with reference to the changes in actual uplink latency of multiple recent cycles, the influence patterns of factors such as network status and encoding parameters on latency during uplink transmission are analyzed. The possible effects of these factors in subsequent cycles are determined, and based on this determination, the actual uplink latency is reasonably extrapolated. The predicted uplink latency of the target video stream, which can predict the uplink transmission time consumption in the future, is calculated, providing an accurate latency reference for subsequent transmission control.

[0038] The actual execution latency is calculated from the time difference between the acquisition timestamp of the new state vector in the terminal state observation sequence and the corresponding receiving time of the cloud control node. This data is generated and stored for archiving in each control cycle.

[0039] Historical downlink latency originates from the downlink latency records of each control cycle during the past transmission of the target video stream stored in the cloud control node. These records are the actual time consumption data generated when the video stream is transmitted from the cloud to the terminal node, and are archived in chronological order.

[0040] The prediction coefficient is determined by analyzing the influence of different historical period data on future prediction results based on the actual uplink latency data of the target video stream over multiple control cycles. The degree of influence is positively correlated with the coefficient weight.

[0041] The control cycle number is a sequential identifier for each consecutive cycle in the video stream transmission control process. It is used to clarify the temporal relationship between different cycles and ensure that historical data and prediction cycles correspond accurately.

[0042] The preset quantity for prediction is a value pre-set based on the stability requirements of video stream transmission and the validity range of historical data, used to limit the total amount of historical period data participating in the prediction calculation.

[0043] The significance of calculating the predicted uplink delay formula lies in selecting a preset number of recent historical control cycle actual uplink delay data, assigning a corresponding prediction coefficient to each historical data, and then accumulating these weighted historical data to obtain the predicted uplink delay for the next control cycle.

[0044] The significance of calculating the predicted downlink delay formula lies in selecting a preset number of historical downlink delay data from past control cycles, assigning a corresponding prediction coefficient to each historical data, and integrating the characteristics of these historical data through weighted summation to calculate the predicted downlink delay for the next control cycle.

[0045] This calculation method fully integrates the latency characteristics of multiple historical periods, and reflects the influence weight of different historical data through coefficient weighting. It effectively avoids the random interference of single data, so that the prediction results can accurately reflect the dynamic change trend of uplink latency, and provide a reliable uplink latency reference for subsequent full-link latency determination and transmission control decisions.

[0046] This calculation method uses coefficient weighting to reflect the influence of different historical downlink latency data, making full use of the changing patterns of multi-period historical data, effectively reducing the prediction bias caused by the randomness of single-period data, and enabling the prediction results to accurately reflect the dynamic change trend of downlink latency. This provides a reliable downlink latency reference for subsequent steps such as determining the end-to-end latency and constructing an effective tolerance time window, ensuring the scientific nature of distributed data transmission control.

[0047] The beneficial effect is that, based on the acquisition timestamp of the new state vector in the terminal state observation sequence and the corresponding reception time of the cloud control node, the actual uplink latency of the current period can be obtained by accurately calculating the time difference, ensuring that the latency data can truly reflect the time consumption of the current uplink transmission, and providing accurate real-time data support for subsequent predictions.

[0048] Historical downlink latency data of the target video stream is extracted and constructed into a historical downlink latency sequence in chronological order. This comprehensively integrates the latency information of past transmissions, avoids the limitations of single historical data, and provides rich and time-series-based historical reference for predicting downlink latency.

[0049] Based on the analysis of historical downlink delay sequence data, the trend of change, fluctuation range and influencing factors are analyzed to accurately calculate and predict downlink delay, so that the prediction results are consistent with the actual change law of downlink transmission, and provide a reliable downlink delay reference for the determination of end-link delay and subsequent control decisions.

[0050] By combining the actual uplink latency of the current cycle with the latency changes of several recent cycles, we analyze the influencing factors of uplink transmission and reasonably extrapolate and predict uplink latency to ensure that uplink latency prediction can adapt to the real-time transmission status and improve the accuracy of end-to-end latency prediction.

[0051] By collecting actual latency data, integrating historical data, and making accurate predictions, both predicted uplink latency and predicted downlink latency are obtained simultaneously. This comprehensively captures the spatiotemporal characteristics of video stream transmission, providing complete and accurate latency data for state extrapolation. This ensures the scientific nature of subsequent encoding control quantity generation and the construction of effective tolerance time windows, thereby improving the accuracy of distributed data transmission control.

[0052] The formula for predicting uplink latency integrates a preset number of historical period uplink latency data and assigns a corresponding prediction coefficient to each data point. It fully considers the impact weight of different historical period latency on future uplink transmission, so that the prediction results can fit the dynamic change pattern of uplink latency and improve the accuracy of uplink latency prediction.

[0053] The formula for predicting downlink latency is based on a preset amount of historical downlink latency data. It quantifies the impact of historical data in different periods through prediction coefficients, comprehensively integrates the latency characteristics of past downlink transmissions, and ensures that the predicted downlink latency can accurately reflect the time consumption trend of future downlink transmissions.

[0054] Both formulas employ a weighted summation calculation logic based on multiple historical data, avoiding the limitations of single data or simple trend predictions, effectively reducing the interference of accidental factors on time delay predictions, and making the prediction results more stable and reliable.

[0055] The prediction coefficients in the formula provide flexible adjustment space for latency prediction. The coefficient configuration can be optimized according to the transmission characteristics of the target video stream, changes in the network environment, and other actual conditions, so as to further adapt to the latency prediction needs in different scenarios.

[0056] By using standardized formulas to calculate the predicted uplink and downlink latency, unified and accurate latency data is provided to support subsequent determination of end-to-end latency, state extrapolation, and generation of coded control quantities. This ensures the scientific nature and consistency of the distributed data transmission control process and improves the overall control accuracy.

[0057] S3. Extrapolate the real-time state vector, the predicted uplink delay, and the predicted downlink delay to obtain the predicted state estimate of the target video stream; In this embodiment of the invention, the step of extrapolating the real-time state vector, the predicted uplink delay, and the predicted downlink delay to obtain the predicted state estimate of the target video stream includes: Based on the predicted uplink latency and the predicted downlink latency, the end-to-end latency of the target video stream is determined; Obtain the state transition relationships that characterize the state evolution law in the target video stream; The state transition relationship is adjusted based on the full-link time delay to obtain the adjusted state transition relationship of the target video stream; Using the real-time state vector as input, the adjusted state transition relationship is forward extrapolated to obtain the predicted state estimate of the target video stream.

[0058] By combining the predicted uplink latency and predicted downlink latency, the two are integrated and calculated to comprehensively cover the time consumed in the complete transmission process of the video stream from the terminal to the cloud and from the cloud to the terminal, and finally the full-link latency of the target video stream that can reflect the time delay of the entire transmission link is determined.

[0059] Retrieve the pre-constructed state transition relationship, which is established based on the transmission characteristics, encoding mechanism and network transmission rules of the target video stream. It can accurately characterize the inherent rules of the state evolution of the target video stream over time in different control cycles and clarify the correlation logic between the current state and the subsequent state.

[0060] Based on end-to-end time delay as the core adjustment basis, the impact of time delay on the state transition process is analyzed. Adaptive adjustments are made to the time dimension parameters involved in the state transition relationship to correct the state evolution deviation caused by time delay. This ensures that the adjusted state transition relationship can accurately match the state change scenario under end-to-end time delay, thus obtaining the adjusted state transition relationship of the target video stream.

[0061] Using the real-time state vector as the core input data, it is substituted into the adjusted state transition relationship. According to the state evolution logic defined by this relationship, it is deduced step by step from the current real-time state, and the possible states of each subsequent time node are derived in turn. Finally, a predictive state estimate that can predict the future running state of the target video stream is obtained, ensuring that the estimate can accurately reflect the state trend of the video stream in subsequent periods.

[0062] The beneficial effects are that by combining the predicted uplink latency and the predicted downlink latency for integrated calculation, the complete transmission link of the video stream uplink and downlink is fully covered, the full link latency is accurately determined, and a time benchmark that fits the actual transmission scenario is provided for the adjustment of state transition relationship, avoiding the deviation of state inference caused by incomplete consideration of latency.

[0063] Obtaining pre-constructed state transition relationships, which are established based on video stream transmission characteristics, encoding mechanisms, and network rules, can accurately characterize the internal logic of state evolution, provide a scientific basis for state extrapolation, and ensure that the extrapolation process conforms to the essential laws of video stream operation.

[0064] Based on the end-to-end time delay, the state transition relationship is adjusted accordingly to correct the evolution deviation in the time dimension, so that the adjusted state transition relationship is fully adapted to the actual transmission time delay, thereby improving the timeliness and accuracy of state inference.

[0065] Using the real-time state vector as input, the adjusted state transition relationship is substituted for forward deduction. The possible states at subsequent time nodes are gradually deduced from the current state. The generated predicted state estimate can accurately predict the future operation trend of the video stream and provide a reliable state reference for the generation of coded control quantities.

[0066] The entire process achieves deep integration of real-time status, link delay, and state evolution laws, ensuring that the predicted state estimates are comprehensive, accurate, and in line with reality. This lays a solid foundation for subsequent coding control, time window construction, and other stages, significantly improving the scientific nature and foresight of distributed data transmission control.

[0067] S4. Map the predicted state estimate to the coding control quantity of the target video stream, and construct an effective tolerance time window for the target video stream based on the predicted downlink delay; In this embodiment of the invention, mapping the predicted state estimate to the coding control quantity of the target video stream, and constructing an effective tolerance time window for the target video stream based on the predicted downlink delay, includes: The predicted state estimate is input into a preset coding strategy mapping table for matching and querying to obtain the coding parameter set of the target video stream; Based on the set of encoding parameters, the encoding control quantity of the target video stream is generated, and the encoding control quantity is sent to the encoding instruction queue of the cloud control node; Based on the predicted downlink latency, the estimated arrival time of the target video stream from the cloud control node to the terminal node is determined; Based on the expected arrival time period, the first buffer duration is extended forward and the second buffer duration is extended backward to obtain the effective tolerance time window of the target video stream.

[0068] A pre-defined encoding strategy mapping table is retrieved. This table is pre-constructed based on factors such as video stream transmission quality requirements, network adaptability, and encoding efficiency. It stores fixed relationships between various prediction state estimates and their corresponding encoding parameter sets. Each prediction state estimate uniquely corresponds to a set of encoding parameter combinations that can adapt to its state. The obtained prediction state estimates are input into this mapping table, and a query is performed according to the exact matching principle to extract the encoding parameter set that perfectly corresponds to the current prediction state estimate, ensuring that the encoding parameter set can adapt to the prediction operation state of the target video stream.

[0069] Based on the matched set of encoding parameters, which includes core encoding parameters such as video compression format, resolution, and frame rate, these parameters are converted into control instructions recognizable by the encoding device. This clarifies the specific execution standards for each parameter during the encoding process and generates encoding control quantities for the target video stream that directly guide the encoding operation. Subsequently, according to the storage rules of the cloud control node's encoding instruction queue, these encoding control quantities are stored in the queue in an orderly manner, ensuring that the execution order of the encoding instructions is not disordered and providing a guarantee for the orderly conduct of subsequent encoding operations.

[0070] Based on the transmission path of the target video stream, network transmission characteristics, and predicted downlink latency, the time range from when the encoded control quantity is issued from the cloud control node to when it reaches the terminal node via network transmission is determined. Combined with the planned transmission time of the current encoded instructions, the earliest and latest estimated arrival times of the target video stream from the cloud control node to the terminal node are calculated. The range defined by these two time points constitutes the expected arrival time period of the target video stream, clearly predicting the transmission time interval.

[0071] Based on the determined estimated arrival time period, and considering the possibility of slight delays or early arrivals during network transmission, a buffer space is set for this time period. Before the earliest estimated arrival time of the estimated arrival time period, the first buffer duration is extended forward to accommodate scenarios of early transmission; after the latest estimated arrival time, the second buffer duration is extended backward to accommodate transmission delays. By extending the buffer duration forward and backward, a time range encompassing the estimated arrival time period and the buffer intervals is formed. This time range constitutes the effective tolerance time window for the target video stream, ensuring that the terminal node can accurately receive the target video stream within a reasonable time range.

[0072] The beneficial effect is that the preset encoding strategy mapping table stores the fixed correlation between the predicted state estimate and the encoding parameter set. The encoding parameter set obtained by precise matching query can perfectly adapt to the predicted running state of the target video stream, avoiding transmission quality problems caused by the mismatch between the encoding parameters and the actual state.

[0073] Based on the set of encoding parameters, encoding control quantities are generated to clarify the specific execution standards of encoding operations, ensuring that the encoding device can accurately execute instructions. At the same time, the encoding control quantities are stored in an orderly manner in the encoding instruction queue of the cloud control node to ensure that encoding operations are carried out in sequence and to avoid chaotic instruction execution.

[0074] Based on the predicted downlink latency combined with factors such as transmission path and network characteristics, the expected arrival time of the target video stream from the cloud to the terminal is accurately determined, and the transmission time interval is clearly predicted, providing an accurate time reference for the subsequent construction of an effective tolerance time window.

[0075] By extending the buffer duration before and after the expected arrival time, an effective tolerance time window is formed, which can flexibly cope with the possible early or late situations in network transmission, avoid the terminal node being unable to receive the video stream normally due to transmission timing fluctuations, and improve the stability of transmission.

[0076] The entire process achieves seamless integration from predicted state to encoding control and transmission timing adaptation. The combination of the precision of encoding control quantities and the flexibility of the effective tolerance time window ensures that distributed data transmission not only meets the requirements of video stream state but also adapts to changes in transmission timing, significantly improving the accuracy and reliability of transmission control.

[0077] S5. When the cloud control node receives the coded control quantity, it compares the current execution time with the effective tolerance time window; In this embodiment of the invention, the step of comparing the current execution time with the effective tolerance time window when the cloud control node receives the coded control quantity includes: Extract the encoding control quantity and the effective tolerance time window from the encoding instruction queue, and obtain the current execution time of the target video stream; Identify the permissible execution start boundary and permissible execution end boundary of the effective tolerance time window; The current execution time is logically determined in relation to the allowed execution start boundary and the allowed execution end boundary.

[0078] The step of performing logical relationship determination between the current execution time and the allowed execution start boundary and the allowed execution end boundary includes: Logically determine the relationship between the current execution time and the allowed execution start boundary and the allowed execution end boundary; When the logical relationship judgment result indicates that the current execution time is earlier than the allowable execution start boundary, a waiting execution instruction for the target video stream is generated; When the logical relationship judgment result indicates that the current execution time is between the allowed execution start boundary and the allowed execution end boundary, an immediate execution instruction for the target video stream is generated; When the logical relationship judgment result indicates that the current execution time exceeds the allowable execution deadline boundary, a new control quantity instruction for the target video stream is generated.

[0079] Following the order stored in the encoding instruction queue, the encoding control variables corresponding to the target video stream are accurately extracted. Simultaneously, the effective tolerance time window associated with this encoding control variable is extracted to ensure a one-to-one correspondence without misalignment. The time synchronization module of the cloud control node obtains the current system's standard time as the current execution time of the target video stream, ensuring the accuracy and consistency of the time information.

[0080] The extracted effective tolerance time window is structured and analyzed to clarify the time range of the time window. The starting point of the time window is the allowable execution start boundary, which represents the earliest time when the coded control quantity can start execution; the ending point of the time window is the allowable execution end boundary, which represents the latest time when the coded control quantity must be completed, thus clearly dividing the legal execution time interval.

[0081] The current execution time is compared one by one with the allowable execution start boundary and the allowable execution end boundary to determine whether the current execution time is after the allowable execution start boundary and before the allowable execution end boundary. Through this clear logical relationship, it is determined whether the current execution time is within the effective tolerance time window, providing a direct basis for the execution decision of subsequent coded control quantities.

[0082] Using the permissible start boundary and permissible end boundary of the effective tolerance time window as the judgment benchmark, the current execution time of the target video stream is compared with these two boundaries one by one to clarify the specific positional relationship of the current execution time relative to the effective tolerance time window and obtain the corresponding logical relationship judgment result.

[0083] When the logical relationship judgment result shows that the current execution time is earlier than the allowable execution start boundary, it indicates that the execution of the encoding control quantity does not meet the preset time requirements. Execution too early may cause the terminal node to fail to adapt in time or cause data synchronization problems. Based on this judgment result, a waiting execution instruction for the target video stream is generated. The instruction clearly states that the execution of the encoding control quantity needs to be paused until the current execution time enters the effective tolerance time window before resuming the operation.

[0084] When the logical relationship judgment result indicates that the current execution time is between the permissible execution start boundary and the permissible execution end boundary, it means that the timing conditions for executing the encoding control quantity are fully met, ensuring that the target video stream can be successfully received and processed after being transmitted to the terminal node. Based on this, an immediate execution instruction for the target video stream is generated, requiring the encoding device to immediately start the encoding operation according to the parameter configuration of the encoding control quantity, ensuring the smooth progress of the transmission process.

[0085] When the logical relationship judgment result shows that the current execution time has exceeded the allowable execution deadline, it means that the encoded control quantity has missed the optimal execution time. Execution at this time may lead to problems such as video stream transmission delay, stuttering, or data loss. Based on this judgment result, a new control quantity request instruction for the target video stream is generated. The instruction triggers the cloud control node to recalculate the predicted state estimate, match the encoding parameter set, and generate a new encoded control quantity to adapt to the current transmission state and ensure the quality of video stream transmission.

[0086] The beneficial effects are that the coded control quantities and associated effective tolerance time windows are extracted sequentially from the coded instruction queue to ensure that the two correspond one-to-one without misalignment. At the same time, the current execution time of the standard is obtained through the time synchronization module of the cloud control node, providing accurate and matching basic data for subsequent comparison and avoiding comparison errors caused by data disorder.

[0087] The effective tolerance time window is analyzed in a structured manner, clearly defining the allowable execution start boundary and allowable execution end boundary, and clearly defining the legal execution time interval, so that the comparison of execution time has a clear basis and avoids judgment bias caused by ambiguity of boundaries.

[0088] By logically determining the relationship between the current execution time and the two boundaries, the system comprehensively covers three scenarios: execution time earlier than, within, and exceeding the time window. This ensures that the comparison results are comprehensive and accurate, providing a reliable basis for generating targeted execution instructions in the future.

[0089] The entire comparison process is standardized and rigorous, forming a complete link from data extraction and boundary identification to logical judgment. This ensures that the comparison results between the execution time and the effective tolerance time window are objective and accurate, avoids errors caused by subjective judgment, and guarantees the rationality of the timing of the execution of the coded control quantity.

[0090] By accurately comparing and clarifying whether the execution timing of the encoded control quantity is compliant, it provides direct support for subsequent decisions on waiting for execution, immediate execution, or requesting new control quantities, ensuring that the encoding operation is adapted to the timing of video stream transmission and improving the stability and controllability of distributed data transmission.

[0091] Using the permissible execution start boundary and permissible execution end boundary as clear judgment criteria, a comprehensive logical verification is performed on the current execution moment to ensure coverage of all possible time-series scenarios and avoid execution decision deviations caused by incomplete judgment dimensions.

[0092] When the current execution time is earlier than the allowed execution start boundary, a waiting execution instruction is generated to avoid terminal adaptation anomalies or data synchronization problems caused by premature execution of coded control quantities, thus ensuring the rationality of transmission timing.

[0093] When the current execution time is between two boundaries, an immediate execution instruction is generated to ensure that the encoded control quantity is executed within the optimal timing window, so that the video stream transmission and the terminal reception requirements are accurately matched, thereby improving transmission efficiency and quality.

[0094] When the current execution time exceeds the allowable execution deadline, a new control quantity request instruction is generated to promptly discard outdated coded control quantities and trigger the calculation and generation of new control parameters, thereby avoiding problems such as transmission delays and data loss caused by timing delays.

[0095] Based on different logical judgment results, targeted instructions are generated to achieve dynamic adaptation of code execution, enabling real-time matching of control decisions and transmission timing. At the same time, it provides clear execution status basis for subsequent error correction, helps closed-loop optimization, and improves the flexibility and reliability of distributed data transmission control.

[0096] S6. Obtain the actual operating state of the next sampling period based on the comparison results, calculate the prediction error between the actual operating state and the predicted state estimate, correct the parameters of the state extrapolation online, and feed back the corrected parameters to S3.

[0097] In this embodiment of the invention, the step of obtaining the actual operating state of the next sampling period based on the comparison result, calculating the prediction error between the actual operating state and the predicted state estimate, correcting the parameters of the state extrapolation online, and feeding back the corrected parameters to S3 includes: Based on the comparison results, the actual operating status corresponding to the coded control quantity is obtained from the execution terminal corresponding to the cloud control node; The actual operating state is matched and compared with the predicted state estimate to obtain the state difference between the actual operating state and the predicted state estimate; Based on the direction and magnitude of the state difference, a parameter correction instruction for the target video stream is generated; The parameter correction instructions are parsed to obtain the target parameters and adjustment trends of the target video stream; The adjustment trend is applied to the target parameter to update the parameters used for state extrapolation; The updated parameters are used as the corrected parameters of the target video stream, and the corrected parameters are fed back to S3.

[0098] Based on the comparison results between the current execution time and the effective tolerance time window obtained previously, the execution status of the encoded control variables is clarified, including the execution status after waiting for execution, immediate execution, or requesting a new control variable. Through the real-time data transmission channel between the cloud control node and the execution terminal, the actual operating status of the target video stream after the encoding control variables are applied is obtained from the execution terminal. This status includes core information that can truly reflect the current operating status of the video stream, such as encoding effect and network transmission feedback.

[0099] The actual operating state is comprehensively matched and compared with the previously obtained predicted state estimate. The specific performance of the two in each state dimension is matched one by one. The system sorts out the differences between the actual operating state and the predicted state estimate in each dimension, and integrates these differences to form a state difference that can fully reflect the differences between the two.

[0100] For the obtained state differences, we conduct in-depth analysis of the direction of deviation to clarify whether the actual operating state is higher or lower than the standard corresponding to the predicted state estimate. At the same time, we accurately judge the magnitude of the difference and determine the severity of the deviation. Based on the direction and magnitude of the difference, we formulate targeted adjustment plans and generate parameter correction instructions for the target video stream. The instructions include the types of parameters to be adjusted and the core logic of the adjustment.

[0101] The generated parameter correction instructions are structured and parsed, and the format information and auxiliary instructions in the instructions are stripped away to extract the target parameters that need to be adjusted for state extrapolation. At the same time, the adjustment trend of these target parameters is clarified, that is, whether the parameters need to be increased, decreased or maintained at a specific ratio, to ensure that the core requirements of parameter adjustment are accurately grasped.

[0102] Based on the adjustment trend obtained from the analysis, it is specifically applied to the corresponding target parameters. According to the requirements of the adjustment trend, the target parameters are actually adjusted numerically or logically optimized. The original parameter settings that do not conform to the current actual operating state are discarded, and the updated parameters are used for state extrapolation to adapt to the current video stream operating conditions.

[0103] The adjusted and updated parameters are determined as the corrected parameters for the target video stream. These corrected parameters are then accurately fed back to the state extrapolation operation through the data feedback channel, replacing the original state extrapolation parameters. This allows subsequent state extrapolation processes to be based on the corrected parameters, improving the accuracy of the predicted state estimate and ensuring dynamic optimization of video stream transmission control.

[0104] The beneficial effects are that by combining the comparison results, the execution scenario of the coded control quantity can be accurately located, and the corresponding actual operating status can be obtained from the execution terminal. This ensures that the actual data collected is directly related to the effect of the coded control quantity, providing real and realistic basic data for error calculation.

[0105] By performing a full-dimensional matching and comparison between the actual operating state and the predicted state estimate, the system sorts out the differences in each state dimension. The obtained state differences can comprehensively and accurately reflect the deviation between the predicted value and the actual situation, providing a clear basis for parameter correction.

[0106] Based on the direction and magnitude of the state differences, targeted parameter correction instructions are formulated to ensure that the correction instructions can accurately target the root cause of the deviation, avoid blind adjustments, and make parameter correction more scientific and effective.

[0107] The parsing parameter correction command clearly defines the target parameter and adjustment trend, and applies the adjustment trend precisely to the target parameter to achieve targeted updates of the state extrapolation parameters. This ensures that the updated parameters can adapt to the current video stream transmission status and improves the accuracy of subsequent state extrapolation.

[0108] The corrected parameters are fed back to the state extrapolation step to build a dynamic optimization closed loop of "prediction-execution-comparison-correction-feedback". The parameter settings of the state extrapolation are continuously optimized so that the control logic can dynamically adapt to the complex fluctuations in the transmission process and avoid control rigidity.

[0109] The entire process continuously reduces the deviation between predicted and actual values ​​through real-time error correction and parameter iteration, thereby continuously improving the accuracy and stability of distributed data transmission control and ensuring that the demand for high-quality video stream transmission is met.

[0110] like Figure 2 The diagram shown is a functional block diagram of a cloud-based distributed data transmission control system provided in an embodiment of the present invention.

[0111] The distributed data transmission control system 100 based on a cloud architecture described in this invention can be installed in an electronic device. Depending on the functions implemented, the distributed data transmission control system 100 may include a data acquisition and transmission module 101, a delay prediction module 102, a state prediction module 103, a control generation module 104, a time window comparison module 105, and an error correction module 106. The module described in this invention can also be referred to as a unit, which refers to a series of computer program segments that can be executed by the processor of an electronic device and can perform a fixed function, and are stored in the memory of the electronic device.

[0112] In this embodiment, the functions of each module / unit are as follows: The acquisition and transmission module 101 is used to acquire the real-time state vector of the target video stream and attach an acquisition timestamp to obtain the terminal state observation sequence of the target video stream, and transmit the terminal state observation sequence to the cloud control node of the target video stream. The latency prediction module 102 is used to determine the actual uplink latency of the target video stream based on the time information in the terminal state observation sequence, and to obtain the predicted uplink latency and predicted downlink latency of the target video stream by combining the historical downlink latency data of the target video stream. The state prediction module 103 is used to extrapolate the real-time state vector, the predicted uplink delay, and the predicted downlink delay to obtain a predicted state estimate of the target video stream. The control generation module 104 is used to map the predicted state estimate to the coding control quantity of the target video stream, and to construct an effective tolerance time window for the target video stream based on the predicted downlink delay. The time window comparison module 105 is used to compare the current execution time with the effective tolerance time window when the cloud control node receives the coded control quantity; The error correction module 106 is used to obtain the actual operating state of the next sampling period based on the comparison result, calculate the prediction error between the actual operating state and the predicted state estimate, correct the parameters of the state extrapolation online, and feed back the corrected parameters to the state prediction module.

[0113] In the several embodiments provided by this invention, it should be understood that the disclosed methods and systems can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.

[0114] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0115] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0116] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0117] This application embodiment can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0118] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A distributed data transmission control method for a cloud architecture, characterized in that, The method includes: S1. Collect the real-time state vector of the target video stream and attach a collection timestamp to obtain the terminal state observation sequence of the target video stream, and send the terminal state observation sequence to the cloud control node of the target video stream; S2. Based on the time information in the terminal status observation sequence, determine the actual uplink latency of the target video stream, and combine it with the historical downlink latency data of the target video stream to obtain the predicted uplink latency and predicted downlink latency of the target video stream; S3. Extrapolate the real-time state vector, the predicted uplink delay, and the predicted downlink delay to obtain the predicted state estimate of the target video stream; S4. Map the predicted state estimate to the coding control quantity of the target video stream, and construct an effective tolerance time window for the target video stream based on the predicted downlink delay; S5. When the cloud control node receives the coded control quantity, it compares the current execution time with the effective tolerance time window; S6. Obtain the actual operating state of the next sampling period based on the comparison results, calculate the prediction error between the actual operating state and the predicted state estimate, correct the parameters of the state extrapolation online, and feed back the corrected parameters to S3.

2. The distributed data transmission control method for a cloud architecture as described in claim 1, characterized in that, The process of acquiring the real-time state vector of the target video stream and appending an acquisition timestamp to obtain the terminal state observation sequence of the target video stream, and then transmitting the terminal state observation sequence to the cloud control node of the target video stream, includes: Obtain the current encoding parameters and network status indicators of the target video stream; Based on the current encoding parameters and the network status indicators, and combined with the encoding control quantities executed in the previous control cycle of the target video stream, the real-time status vector of the target video stream is obtained. A corresponding acquisition timestamp is appended to the real-time state vector to generate a terminal state observation sequence of the target video stream, and the terminal state observation sequence is sent to the cloud control node of the target video stream.

3. The distributed data transmission control method for a cloud architecture as described in claim 1, characterized in that, The step of determining the actual uplink latency of the target video stream based on the time information in the terminal state observation sequence, and obtaining the predicted uplink latency and predicted downlink latency of the target video stream by combining the historical downlink latency data of the target video stream, includes: The acquisition timestamp of the new state vector in the terminal state observation sequence and the corresponding receiving time of the cloud control node are obtained to obtain the actual uplink delay of the current period; Extract the historical downlink latency data of the target video stream to construct the historical downlink latency sequence of the target video stream; Based on the historical downlink latency sequence, the predicted downlink latency of the target video stream is calculated; Based on the actual uplink latency, the predicted uplink latency of the target video stream is calculated.

4. The distributed data transmission control method for a cloud architecture as described in claim 3, characterized in that, The formulas for calculating the predicted downlink latency and the predicted uplink latency are as follows: ； In the formula, For the first Predicted uplink latency for each control cycle For the first One predictive coefficient, For actual uplink latency, To control the cycle number, The preset quantity for prediction; ； In the formula, For the first Predicted downlink latency for each control cycle, The preset quantity for prediction, For the first One predictive coefficient, For historical downlink latency, This is for controlling the cycle number.

5. The distributed data transmission control method for a cloud architecture as described in claim 2, characterized in that, The step of extrapolating the real-time state vector, the predicted uplink delay, and the predicted downlink delay to obtain the predicted state estimate of the target video stream includes: Based on the predicted uplink latency and the predicted downlink latency, the end-to-end latency of the target video stream is determined; Obtain the state transition relationships that characterize the state evolution law in the target video stream; The state transition relationship is adjusted based on the full-link time delay to obtain the adjusted state transition relationship of the target video stream; Using the real-time state vector as input, the adjusted state transition relationship is forward extrapolated to obtain the predicted state estimate of the target video stream.

6. The distributed data transmission control method for a cloud architecture as described in claim 1, characterized in that, The step of mapping the predicted state estimate to the coding control quantity of the target video stream, and constructing an effective tolerance time window for the target video stream based on the predicted downlink delay, includes: The predicted state estimate is input into a preset coding strategy mapping table for matching and querying to obtain the coding parameter set of the target video stream; Based on the set of encoding parameters, the encoding control quantity of the target video stream is generated, and the encoding control quantity is sent to the encoding instruction queue of the cloud control node; Based on the predicted downlink latency, the estimated arrival time of the target video stream from the cloud control node to the terminal node is determined; Based on the expected arrival time period, the first buffer duration is extended forward and the second buffer duration is extended backward to obtain the effective tolerance time window of the target video stream.

7. The distributed data transmission control method for a cloud architecture as described in claim 6, characterized in that, When the cloud control node receives the coded control quantity, it compares the current execution time with the effective tolerance time window, including: Extract the encoding control quantity and the effective tolerance time window from the encoding instruction queue, and obtain the current execution time of the target video stream; Identify the permissible execution start boundary and permissible execution end boundary of the effective tolerance time window; The current execution time is logically determined in relation to the allowed execution start boundary and the allowed execution end boundary.

8. The distributed data transmission control method for a cloud architecture as described in claim 7, characterized in that, The step of performing logical relationship determination between the current execution time and the allowed execution start boundary and the allowed execution end boundary includes: Logically determine the relationship between the current execution time and the allowed execution start boundary and the allowed execution end boundary; When the logical relationship judgment result indicates that the current execution time is earlier than the allowable execution start boundary, a waiting execution instruction for the target video stream is generated; When the logical relationship judgment result indicates that the current execution time is between the allowed execution start boundary and the allowed execution end boundary, an immediate execution instruction for the target video stream is generated; When the logical relationship judgment result indicates that the current execution time exceeds the allowable execution deadline boundary, a new control quantity instruction for the target video stream is generated.

9. The distributed data transmission control method for a cloud architecture as described in claim 1, characterized in that, The step of obtaining the actual operating state of the next sampling period based on the comparison result, calculating the prediction error between the actual operating state and the predicted state estimate, correcting the parameters of the state extrapolation online, and feeding back the corrected parameters to S3 includes: Based on the comparison results, the actual operating status corresponding to the coded control quantity is obtained from the execution terminal corresponding to the cloud control node; The actual operating state is matched and compared with the predicted state estimate to obtain the state difference between the actual operating state and the predicted state estimate; Based on the direction and magnitude of the state difference, a parameter correction instruction for the target video stream is generated; The parameter correction instructions are parsed to obtain the target parameters and adjustment trends of the target video stream; The adjustment trend is applied to the target parameter to update the parameters used for state extrapolation; The updated parameters are used as the corrected parameters of the target video stream, and the corrected parameters are fed back to S3.

10. A distributed data transmission control system based on a cloud architecture, characterized in that, The system is used to implement the distributed data transmission control method for a cloud architecture as described in claim 1, the system comprising: The acquisition and transmission module is used to acquire the real-time state vector of the target video stream and attach an acquisition timestamp to obtain the terminal state observation sequence of the target video stream, and transmit the terminal state observation sequence to the cloud control node of the target video stream. The latency prediction module is used to determine the actual uplink latency of the target video stream based on the time information in the terminal state observation sequence, and to obtain the predicted uplink latency and predicted downlink latency of the target video stream by combining the historical downlink latency data of the target video stream. The state prediction module is used to extrapolate the real-time state vector, the predicted uplink delay, and the predicted downlink delay to obtain a predicted state estimate of the target video stream. The control generation module is used to map the predicted state estimate to the coding control quantity of the target video stream, and to construct an effective tolerance time window for the target video stream based on the predicted downlink delay. The time window comparison module is used to compare the current execution time with the effective tolerance time window when the cloud control node receives the coded control quantity; The error correction module is used to obtain the actual operating state of the next sampling period based on the comparison results, calculate the prediction error between the actual operating state and the predicted state estimate, correct the parameters of the state extrapolation online, and feed back the corrected parameters to the state prediction module.

Images