6G HRLLC-oriented multi-dimensional intelligent service quality assurance method and system

Abstract

The invention discloses a 6G HRLLC-oriented multi-dimensional intelligent service quality assurance method and system, and relates to the technical field of communication, and the method comprises the steps: obtaining the service quality of multi-dimensional data, and constructing a service quality assurance index of the multi-dimensional data; constructing a feasible region of the multi-dimensional data; performing feature extraction on the service quality of the multi-dimensional data in the feasible region of the multi-dimensional data to obtain a feature vector of the service quality of the multi-dimensional data; processing the feature vector of the service quality of the multi-dimensional data by adopting a trained deep learning network to obtain a service quality collaborative decision of the multi-dimensional data, and converting the service quality collaborative decision of the multi-dimensional data into an executable network instruction; and executing cross-layer scheduling and resource allocation according to the executable network instruction. According to the invention, better multi-dimensional intelligent service quality guarantee can be provided.

Description

Technical Field

[0001] This invention belongs to the field of communication technology, specifically relating to a multi-dimensional intelligent quality of service assurance method and system for 6G HRLLC. Background Technology

[0002] With the gradual development of the sixth-generation mobile communication system (6G), it is widely used in industrial automation, vehicle networking, telemedicine, smart grid, immersive virtual reality and holographic communication and other fields. These fields have the common feature of having extremely stringent requirements for indicators such as reliability, latency and latency jitter.

[0003] However, most existing quality of service (QoS) assurance mechanisms rely on single-dimensional statistical modeling, typically considering only latency default probability or reliability limits. In 6G scenarios, traditional QoS assurance methods often optimize only a single metric (such as latency or bit error rate), failing to simultaneously satisfy multi-dimensional constraints such as latency, jitter, peak information freshness, and reliability. Furthermore, they cannot dynamically extract the coupling relationships between various metrics, thus wasting significant network resources. In addition, traditional static scheduling and resource allocation methods struggle to maintain stable QoS assurance for unpredictable wireless network conditions, and are mostly limited to local optimization at the physical or MAC layers, lacking cross-layer joint modeling and global optimization mechanisms. Moreover, existing solution schemes are generally based on convex optimization or iterative solution mechanisms, making it difficult to proactively adjust resource allocation strategies based on predictions of multi-dimensional QoS assurance constraints. This results in the system's inability to anticipate and mitigate multi-dimensional QoS assurance default risks.

[0004] Therefore, there is an urgent need to provide a multi-dimensional intelligent service quality assurance method and system for 6G HRLLC to improve the shortcomings of existing technologies. Summary of the Invention

[0005] To address the aforementioned problems in the existing technology, this invention provides a multi-dimensional intelligent quality of service assurance method and system for 6G HRLLC. The technical problem to be solved by this invention is achieved through the following technical solution: In a first aspect, the present invention provides a multi-dimensional intelligent quality of service assurance method for 6G HRLLC, comprising: To acquire the service quality of multidimensional data, we construct service quality assurance indicators for multidimensional data; among which, multidimensional data includes latency, peak information freshness, jitter, and reliability. Based on the current network status and business characteristics, and combined with the service quality assurance indicators of multidimensional data, the feasible region of multidimensional data is obtained; within the feasible region of multidimensional data, features of service quality of multidimensional data are extracted to obtain the feature vector of service quality of multidimensional data. A trained deep learning network is used to process the feature vectors of service quality of multidimensional data to obtain collaborative service quality decisions for multidimensional data, and then the collaborative service quality decisions for multidimensional data are transformed into executable network instructions. Based on executable network instructions, perform cross-layer scheduling and resource allocation; at the same time, check whether the service quality status characteristics of the multidimensional data after execution meet the service quality assurance indicators of the multidimensional data, so as to form a closed-loop control cycle.

[0006] Secondly, the present invention also provides a multi-dimensional intelligent quality of service assurance system for 6G HRLLC, used to implement the above-mentioned multi-dimensional intelligent quality of service assurance method for 6G HRLLC, including: The data acquisition module is used to collect service quality data in multiple dimensions, including latency, peak information freshness, jitter, and reliability. The basic module is used to build service quality assurance indicators for multidimensional data; The perception module is used to obtain the feasible region of multidimensional data based on the current network status and business characteristics and behavior, combined with the service quality assurance indicators of multidimensional data; within the feasible region of multidimensional data, the service quality of multidimensional data is extracted to obtain the feature vector of service quality of multidimensional data. The intelligent module is used to process the feature vectors of service quality of multidimensional data using a trained deep learning network, obtain collaborative service quality decisions for multidimensional data, and transform the collaborative service quality decisions for multidimensional data into executable network instructions. The control module is used to perform cross-layer scheduling and resource allocation based on executable network instructions; at the same time, it checks whether the service quality status characteristics of the multidimensional data after execution meet the service quality assurance indicators of the multidimensional data, so as to form a closed-loop control cycle.

[0007] The beneficial effects of this invention are: This invention provides a multi-dimensional intelligent quality of service (QoS) assurance method and system for 6G HRLLC. It proposes a multi-dimensional QoS assurance method based on latency, peak information freshness, jitter, and reliability. By identifying and defining new statistical multi-dimensional QoS assurance indicators, it addresses the challenge that existing single-dimensional QoS assurance methods cannot be applied to diverse application scenarios and multi-dimensional performance indicators. Based on the multi-dimensional QoS assurance constraints, an artificial intelligence algorithm is introduced, enabling the system to perceive changes in the network environment in real time and dynamically adjust resource allocation strategies, thereby avoiding the overly conservative problems caused by static design. Furthermore, the cross-layer collaborative design proposed in this invention utilizes AI-driven QoS assurance to achieve feature sharing and joint optimization among the physical layer, MAC layer, transport layer, and application layer, breaking down inter-layer barriers and achieving end-to-end multi-dimensional QoS assurance. This ensures low computational complexity while meeting the real-time requirements of 6G ultra-reliable low-latency communication.

[0008] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0009] Figure 1 This is a flowchart of a multi-dimensional intelligent service quality assurance method for 6G HRLLC provided in an embodiment of the present invention; Figure 2 This is another flowchart of the multi-dimensional intelligent service quality assurance method for 6G HRLLC provided in the embodiments of the present invention; Figure 3 This is a schematic diagram of a multi-dimensional intelligent service quality assurance system for 6G HRLLC provided in an embodiment of the present invention; Figure 4 This is a schematic diagram comparing the performance of the AI-QoS mechanism provided in this embodiment of the invention with the baseline statistical latency QoS scheme and the URLLC optimized scheduler; Figure 5 This is a schematic diagram of the violation distribution of various service quality indicators of AI-QoS provided in an embodiment of the present invention. Detailed Implementation

[0010] The present invention will be further described in detail below with reference to specific embodiments, but the implementation of the present invention is not limited thereto.

[0011] In existing technologies, the following shortcomings still exist in the multi-dimensional Quality of Service (QoS) assurance for 6G Ultra Reliable Low Latency Communication (HRLLC): First, the limitations of designing a single service quality assurance dimension: Most existing methods focus on statistical modeling and assurance of single performance indicators (such as latency), lacking joint modeling and optimization of multi-dimensional service quality assurance indicators (including latency, jitter, peak information freshness, and reliability). As a result, when facing emerging and complex business scenarios, the system cannot balance multiple business or service requirements, and often can only perform "worst-case design" for the most stringent indicators, thus sacrificing the overall efficiency of the system.

[0012] Second, the correlation between multidimensional service quality assurance indicators is not fully utilized; existing statistical latency service quality assurance methods only handle latency indicators independently, failing to capture the correlation and coupling between latency, jitter, information freshness and reliability. This fragmented design results in limited optimization effects and cannot effectively reduce the overall probability of service quality assurance defaults globally.

[0013] Third, there is a lack of dynamic adaptability and intelligent optimization capabilities. Existing methods rely heavily on fixed thresholds and static optimization methods for resource scheduling and quality of service assurance, which are difficult to adapt to the highly dynamic channel conditions and traffic fluctuations in the 6G environment. Once the network environment changes, the preset scheduling strategy is prone to failure, leading to a decline in service level assurance or even a violation of the constraint probability.

[0014] Fourth, there is a lack of cross-layer collaboration and intelligent agent cooperation mechanisms. Current service quality assurance optimization is mostly limited to a single network layer (such as the MAC layer or transport layer), lacking cross-layer joint optimization and distributed cooperation mechanisms. In 6G heterogeneous networks with multiple access points and multiple nodes, the lack of collaboration mechanisms leads to the dispersion of resource allocation, making it difficult to achieve globally consistent service assurance.

[0015] Fifth, it is difficult to handle non-convex optimization and high-dimensional complex constraints; the multi-service quality assurance problem is essentially a multi-objective, non-convex, time-varying optimization problem, which is difficult to solve with existing analytical methods, and existing heuristic algorithms are also difficult to guarantee convergence and real-time performance, especially in the HRLLC scenario, where the real-time requirements are extremely high, and the computational complexity of existing algorithms is often insufficient to meet the requirements.

[0016] In view of this, this invention provides a multi-dimensional intelligent quality of service (QoS) assurance method and system for 6G HRLLC. It proposes a multi-dimensional QoS assurance method based on latency, peak information freshness, jitter, and reliability. By identifying and defining new statistical multi-dimensional QoS assurance indicators, it addresses the challenge that existing single-dimensional QoS assurance methods cannot be applied to diverse application scenarios and multi-dimensional performance indicators. Based on the multi-dimensional QoS assurance constraints, an artificial intelligence algorithm is introduced, enabling the system to perceive changes in the network environment in real time and dynamically adjust resource allocation strategies, thereby avoiding the overly conservative problems caused by static design. Furthermore, the cross-layer collaborative design proposed in this invention utilizes AI-driven QoS assurance to achieve feature sharing and joint optimization among the physical layer, MAC layer, transport layer, and application layer, breaking down inter-layer barriers and achieving end-to-end multi-dimensional QoS assurance. This ensures low computational complexity while meeting the real-time requirements of 6G ultra-reliable low-latency communication.

[0017] Please see Figure 1 and Figure 2 , Figure 1 This is a flowchart of a multi-dimensional intelligent service quality assurance method for 6G HRLLC provided in an embodiment of the present invention. Figure 2 This is another flowchart of a multi-dimensional intelligent quality of service assurance method for 6G HRLLC provided in an embodiment of the present invention. The multi-dimensional intelligent quality of service assurance method for 6G HRLLC provided by the present invention includes: S101. Obtain the service quality of multidimensional data and construct service quality assurance indicators for multidimensional data; among which, multidimensional data includes latency, peak information freshness, jitter and reliability.

[0018] Specifically, in this embodiment, multimodal data is collected by intelligent devices and preprocessed, including capturing traffic patterns, user mobility, channel quality indicators, service level agreements (SLAs), and device-specific metrics (such as energy levels and processing latency) to obtain multidimensional data.

[0019] Construct service quality assurance indicators for multidimensional data, including: Using the statistical latency service quality assurance model as a theoretical benchmark, the tail behavior of service quality violation probabilities for latency, peak information freshness, jitter, and reliability is modeled. Optionally, tail behavior refers to the decay law or probability of the probability distribution in the extreme region when performance indicators (such as queuing latency and peak information freshness) exceed a certain threshold, which describes the service quality of the system in the worst case.

[0020] To study the multidimensional service quality supply mechanism, this embodiment proposes a unified definition of multidimensional service quality assurance indicators: make This represents a general stochastic transformation process of service quality for multidimensional data, including latency, peak information freshness, jitter, and reliability. Using the large deviation principle, under sufficient conditions, the probability of a service quality violation exceeding a preset threshold for multidimensional data satisfies the following: ; in, This represents the probability of service quality violations in multidimensional data. This represents the stochastic transformation process of service quality for multidimensional data. This represents the decay rate function corresponding to multidimensional data, used to quantify the exponential decay rate of violation probabilities driven by service quality. This represents a preset threshold; as the preset threshold increases, the probability of service quality assurance violations decreases exponentially, thus providing a strict probability limit for extreme service quality assurance events. The decay rate function... The size directly reflects the strength of service quality assurance; The smaller the value, the slower the exponential decay of the violation probability, indicating that the system can only provide a relatively lenient quality of service guarantee. The larger the value, the faster the decay rate, thus enabling the support of more stringent service quality requirements.

[0021] It should be noted that when When the probability of a violation no longer decays, it means the system can actually tolerate any level of violation; when When the probability of a violation approaches zero, it means that no violation is tolerated below the preset threshold. It should be noted that the exponential decay rate function described here can be equally applied to the service quality assurance indicators of latency, peak information freshness, jitter, and reliability introduced below.

[0022] Based on the decay rate function corresponding to multidimensional data, service quality assurance indicators for latency, peak information freshness, jitter, and reliability are constructed respectively.

[0023] 1. Statistical Service Quality Assurance Indicator for Quantitatively Determining Delay. Based on the large deviation principle, under sufficient conditions, the probability of queuing delay violation follows an exponential fading pattern. The rate of this exponential fading is defined as the delay-based service quality assurance indicator. This indicator reflects the stringency level of service quality supply corresponding to the statistical delay boundary as the buffer overflow threshold increases, considering service data volume, buffer status, and delay violation probability. The larger the value of this delay service quality assurance indicator, the more effectively the network can reduce delay risk by making small adjustments to its operating indicators, thereby ensuring a higher level of service quality.

[0024] It should be noted that the network mentioned in this embodiment is a 6G network.

[0025] 2. Statistical Service Quality Assurance Indicators for Quantitatively Defining Peak Information Freshness. The stringency of the service quality assurance measure for peak information freshness constraints is determined by its corresponding peak information freshness service quality assurance index. This index defines the rate at which the probability of a peak information freshness violation decreases exponentially with increasing peak information freshness threshold. For a given peak information freshness threshold, its violation probability can be used to characterize the tail characteristics of peak information freshness. Therefore, a sufficiently small peak information freshness constraint service quality assurance index allows the system to accommodate arbitrarily large peak information freshness, while a sufficiently large index means that the system cannot tolerate any peak information freshness exceeding the threshold.

[0026] 3. Statistical Service Quality Assurance Metrics for Quantitatively Measuring Jitter. Jitter-constrained service quality metrics are used to limit the permissible range of jitter variation in the network, identify and enforce deterministic latency violation thresholds necessary for different service types, ensuring the network can meet the service level requirements of applications such as Voice over IP (VoIP), video streaming, and real-time gaming. These metrics assess and manage packet latency fluctuations within strict limits, ensuring latency does not exceed preset thresholds, which is crucial for maintaining data flow integrity. In time-sensitive applications, maintaining predictable low jitter is particularly critical for ensuring reliable service quality.

[0027] 4. Quantitatively characterizing statistical quality of service (QoS) assurance metrics for reliability. In the domain of finite code length, a metric for QoS assurance mechanisms for reliability is constructed, systematically describing the attenuation of bit error probability as code length increases. Based on the large deviation principle, when the coding rate is lower than the channel capacity, the QoS assurance index based on bit error probability characterizes the exponential decay rate of the reliability QoS violation probability (i.e., bit error probability), measuring the increase in the rigor of statistical reliability QoS assurance with increasing block length.

[0028] S102. Based on the current network status and business characteristics, and combined with the service quality assurance indicators of multidimensional data, obtain the feasible region of multidimensional data; within the feasible region of multidimensional data, extract features of the service quality of multidimensional data to obtain the feature vector of the service quality of multidimensional data.

[0029] Specifically, in this embodiment, based on network status and service characteristic behavior, and combined with the service quality assurance indicators of multidimensional data, the feasible domain of multidimensional data is obtained, including: Based on the current network status, estimate the service quality boundary of multidimensional data, and dynamically adjust the priority of service quality assurance indicators of multidimensional data according to business characteristics and behaviors. Combine the service quality assurance indicators of multidimensional data to set the feasible domain of multidimensional data.

[0030] In this embodiment, the acquired multidimensional data is converted into operable quality indicators, and the network state (including channel changes, service arrival process, and mobility dynamics) is mapped to the optimal operation strategy. For example, the current network's adaptive scheduling, resource allocation, and cooperative retransmission strategy are used as the optimal operation strategy. Based on queuing theory, effective capacity analysis, and finite code length coding (FBC) theory, statistical performance boundaries such as the upper bound of latency, peak AoI distribution, reliability probability, and jitter threshold are estimated under the current network state to further evaluate the statistical assurance capability of key service quality performance indicators. Furthermore, the priority of the service quality assurance indicators of the multidimensional data is dynamically adjusted according to service characteristic behavior. Combining the service quality assurance indicators of the multidimensional data, a feasible domain of the multidimensional data is constructed to guide the feature extraction process.

[0031] It should be noted that the feasible region of multidimensional data represents the inherent trade-offs among various service quality assurance indicators. For example, achieving extremely low latency limits may weaken the ability to control jitter, reflecting the mutual constraint between latency and variance. In situations where spectrum resources are limited, forcing extreme peak information freshness may come at the cost of a reduced reliability index. Through perceptual learning, a multivariate statistical learning method is employed to capture the dependencies between latency, jitter, reliability, and peak information freshness, identifying the joint statistical structure among the multidimensional service quality dimensions.

[0032] Furthermore, in this embodiment, within the feasible domain of the multidimensional data, feature extraction is performed on the service quality of the multidimensional data to obtain a feature vector of the service quality of the multidimensional data, including: Feature extraction is performed on latency-oriented quality of service to obtain the latency distribution function, which characterizes the sensitivity to latency; Feature extraction is performed on the service quality oriented towards peak information freshness to obtain information freshness statistics, which characterize the requirements for peak information freshness; Feature extraction is performed on the quality of service (QoS) oriented towards jitter to obtain jitter variance, which characterizes the periodicity of traffic. Feature extraction is performed on reliability-oriented service quality to obtain reliability curves, which characterize reliability constraints.

[0033] In this embodiment, the feasible domain of multidimensional data defines the acceptable latency, peak information freshness, jitter, and reliability boundaries. The feature extraction process is performed within this feasible domain, ensuring that all learning and adaptation processes are always feasible within this probability region.

[0034] S103. The trained deep learning network is used to process the feature vector of service quality of multidimensional data to obtain the service quality collaborative decision of multidimensional data, and the service quality collaborative decision of multidimensional data is transformed into executable network instructions.

[0035] Specifically, in this embodiment, the training process of the deep learning network includes: Obtain training samples to construct a training dataset; wherein, the training samples include feature vectors of service quality of pre-defined multidimensional data, and the true labels of the training samples include the dependencies between multidimensional data; Input a portion of the samples from the training dataset into the first... The deep learning network to be trained is then trained to obtain the first... The prediction results output during this training process; According to the The prediction results output during the training process are the same as those during the training phase. The true labels of the samples for the next deep learning network to be trained are used to calculate the loss and serve as the first... Loss during each training session; According to the The loss from the training process is backpropagated to update the loss from the training process. The network parameters of the deep learning network to be trained are obtained. The deep learning network to be trained is iterated until the number of training iterations or the degree of convergence meets the preset conditions, and a trained deep learning network is obtained.

[0036] Furthermore, the process of obtaining the dependencies between multidimensional data includes: By employing a multivariate statistical learning method, the dependencies between latency, peak information freshness, jitter, and reliability are constructed through learning and approximating the Pareto optimal strategy.

[0037] In this embodiment, an offline multi-objective policy based on multi-dimensional quality of service (QoS) assurance probability constraints is generated. A deep reinforcement learning algorithm is used to select a suitable policy from the offline policy library and dynamically adjust it in combination with the real-time network status to obtain a QoS collaborative decision based on multi-dimensional data. This collaborative decision determines how to allocate resources such as time-frequency resource blocks, transmit power, and cache space to meet diverse QoS assurance requirements.

[0038] It should be noted that when training deep learning networks, strategies from an offline policy library are used.

[0039] S104. Based on the executable network instructions, perform cross-layer scheduling and resource allocation; at the same time, check whether the service quality status characteristics of the multidimensional data after execution meet the service quality assurance indicators of the multidimensional data, so as to form a closed-loop control cycle.

[0040] Specifically, in this embodiment, collaborative decisions are transformed into executable network instructions. These collaborative decisions are not generated independently at a single network layer, but rather through joint optimization to ensure end-to-end multi-dimensional quality of service performance. For example, when high-priority service flows encounter increased jitter, the system can simultaneously reduce transmission intervals, increase signal power, and adjust the processing queues of edge servers. Furthermore, the execution results are monitored and intelligently optimized. Actual performance metrics after the decision, such as the user's actual latency and packet loss events, are collected and fed back into the learning process, thus forming a closed-loop control cycle.

[0041] In summary, the multi-dimensional intelligent quality of service assurance method for 6G HRLLC provided by this invention has the following beneficial effects: First, this invention fills the gaps in traditional service quality assurance performance indicators. Addressing the current lack of research on service quality assurance indicators such as latency and jitter, it re-identifies and defines a multi-dimensional service quality assurance performance indicator system and constructs an intelligent-driven unified multi-dimensional service quality assurance framework. It is the first to propose a multi-dimensional service quality assurance mechanism based on artificial intelligence. Under the unified framework, it dynamically realizes unified measurement modeling and collaborative optimization of latency, peak information freshness, jitter, and reliability, effectively reflecting diverse and extreme performance indicators in special application scenarios. It supports multiple end-to-end application scenarios in the complete future wireless network and adapts to the dynamic environment of heterogeneous 6G networks.

[0042] Secondly, this invention designs a security constraint mechanism based on statistical multidimensional quality of service assurance boundaries. Under network uncertainty conditions, interactive verification is performed through boundary constraint models to define a unified security margin for multidimensional quality of service assurance statistical constraints, ensuring that probabilistic service guarantees are not violated, and that the generated resource scheduling strategy always remains within the mathematical boundaries that can provide probabilistic service guarantees, thus achieving "constrained intelligence".

[0043] Based on the same inventive concept, please refer to Figure 3 , Figure 3 This is a schematic diagram of a multi-dimensional intelligent quality of service assurance system for 6G HRLLC provided in an embodiment of the present invention. The present invention also provides a multi-dimensional intelligent quality of service assurance system for 6G HRLLC, used to implement the multi-dimensional intelligent quality of service assurance method for 6G HRLLC provided in the above embodiments. Embodiments of the method can be referred to above and will not be repeated here. The system includes: The data acquisition module is used to collect service quality data in multiple dimensions, including latency, peak information freshness, jitter, and reliability. The basic module is used to build service quality assurance indicators for multidimensional data; The perception module is used to obtain the feasible region of multidimensional data based on the current network status and business characteristics and behavior, combined with the service quality assurance indicators of multidimensional data; within the feasible region of multidimensional data, the service quality of multidimensional data is extracted to obtain the feature vector of service quality of multidimensional data. The intelligent module is used to process the feature vectors of service quality of multidimensional data using a trained deep learning network, obtain collaborative service quality decisions for multidimensional data, and transform the collaborative service quality decisions for multidimensional data into executable network instructions. The control module is used to perform cross-layer scheduling and resource allocation based on executable network instructions; at the same time, it checks whether the service quality status characteristics of the multidimensional data after execution meet the service quality assurance indicators of the multidimensional data, so as to form a closed-loop control cycle.

[0044] In an optional embodiment of the present invention, it further includes: a coordination module, used to determine whether the executable network instructions are coordinated with the service quality assurance indicators of the multidimensional data.

[0045] In an optional embodiment of the present invention, the control module includes a cross-layer scheduling module, a resource allocation module, and a service quality status feature guarantee module for multi-dimensional data; wherein, the cross-layer scheduling module is used to implement cross-layer scheduling of the network, the resource allocation module is used to implement resource allocation according to executable network instructions, and the service quality status feature guarantee module for multi-dimensional data is used to obtain the service quality status features of the multi-dimensional data after execution.

[0046] In summary, this invention proposes a hierarchical processing method that integrates intelligent feature extraction, adaptive learning, federated collaboration, and predictive resource management. It provides agile, stable, and customized low-tail QoS guarantees for service quality requirements such as latency, peak information freshness, jitter, and bit error rate. The core of this method consists of the following hierarchical modules: The basic module is responsible for collecting underlying network resources, including channel state information (CSI), queue length, traffic intensity, and historical multi-dimensional quality of service assurance indicators.

[0047] Perception module: It performs feature extraction and modeling on the data collected from the base layer, and generates multi-dimensional service quality assurance feature vectors, such as latency distribution function, jitter variance, information freshness statistics and reliability curves.

[0048] The intelligent module adopts the principle of "offline training, online decision-making." This means training the deep learning network using generated offline strategies, and then learning online strategies based on the trained deep learning network. It integrates multi-objective reinforcement learning and deep reinforcement learning to transform high-dimensional network observation data into dynamic control and resource allocation strategies. Specifically, multi-objective reinforcement learning explores Pareto-optimal strategies that balance mutually constraining service demands to generate offline strategies. Deep reinforcement learning, building upon this, enables low-latency decision-making under highly dynamic conditions. Specifically, deep reinforcement learning trained based on the perceptron layer input can generate refined control operations in real time, including scheduling, resource slicing, and power adaptation.

[0049] The control module translates the decisions of the intelligent layer into specific operations at the network layer. A statistical multidimensional quality of service (QoS) assurance model is used as a filter in the action space, performing constraint verification on the generated decisions and proposing decisions regarding transmit power control, channel access control, and priority scheduling. By enforcing low-tailed bounded guarantees on violation probabilities, the feasible operational region is ensured to remain intact, thus improving performance without compromising statistical reliability during the adaptive process.

[0050] Collaboration Module: Enables federated learning and collaborative optimization among distributed intelligent agents, ensures end-to-end multi-dimensional service quality assurance, and achieves predictive risk control on a global network scale.

[0051] In summary, this invention proposes a hierarchical and collaborative strategy driven by multi-agent and multi-objective optimization. The hierarchical architecture (including a perception layer, intelligence layer, control layer, and collaboration layer) achieves a complete closed loop from feature extraction, learning optimization, policy execution, and multi-agent collaboration. Based on this, a coupled design of functions at different layers is proposed, such as channel-flow representation learning at the feature layer, reinforcement learning and evolutionary search at the intelligence layer, and federated learning and multi-agent cooperation at the collaboration layer, enhancing the practicality and scalability of the solution.

[0052] In an optional embodiment of the present invention, the effectiveness of the multi-dimensional intelligent service quality assurance method for 6GHRLLC provided in the above embodiment is verified by simulation experiments, specifically as follows: Please see Figure 4 , Figure 4This is a schematic diagram comparing the performance of the AI-QoS mechanism provided in this embodiment of the invention with the baseline statistical QoS scheme and the URLLC optimized scheduler. The AI-QoS mechanism (proposed AI-QoS) is the mechanism proposed in the above embodiment of the invention. The baseline statistical QoS scheme and the URLLC optimized scheduler are both existing schemes in the prior art. The baseline statistical QoS scheme only considers the latency boundary QoS constraint, while the URLLC optimized scheduler only considers the latency and error rate constraints. Its performance evaluation uses the Multi-QoS Violation Probability (MQVP) as a unified indicator. This indicator is used to quantify the probability of violating any condition among multiple service quality assurance constraints such as latency, jitter, reliability, and peak information freshness. Figure 4 The graphs show the MQVP (Message Queuing VP) of the three schemes as a function of normalized network load level. Under low load conditions (network load = 1), all schemes perform reasonably well, but the AI-QoS framework shows a significant advantage: its MQVP drops to 2%, while the baseline statistical latency QoS scheme and the URLLC optimized scheduler reach 15% and 10%, respectively. As the load increases, the performance difference becomes increasingly significant: under medium load conditions (network load = 3), the MQVP of the baseline statistical latency QoS scheme rises sharply to 60%, while the AI-QoS framework only maintains 10%, indicating its strong robustness under peak traffic conditions. Under the highest load conditions (network load = 5), the baseline statistical latency QoS scheme reaches full saturation (MQVP = 100%), the URLLC optimized scheduler also rises to 95%, while the AI-QoS framework remains at 30%. Compared to the baseline statistical latency QoS scheme and the URLLC optimized scheduler, it achieved a relative reduction of 70% and 68.4% in violation probability, respectively. These improvements stem from the AI-QoS framework's ability to jointly predict and control multi-dimensional QoS metrics, dynamically reallocate resources based on learned traffic patterns, and adaptively adjust scheduling strategies.

[0053] Please see Figure 5 , Figure 5This diagram illustrates the distribution of violations among various AI-QoS service quality indicators provided in this embodiment of the invention. It demonstrates the relationship between different violations of multi-dimensional service quality indicators based on latency, peak information freshness (AoI), jitter, and reliability, and network load level. As network load increases, the proportion of peak information freshness violations remains relatively constant, while the proportion of latency violations shows a significant downward trend. Simultaneously, the proportions of reliability and latency / jitter violations increase significantly. This indicates that the higher the network load, the more stringent the requirements for jitter and reliability in services.

[0054] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations are intended to cover non-exclusive inclusion, such that an article or device comprising a list of elements includes not only those elements but also other elements not expressly listed. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the article or device comprising said element. Terms such as "connected" or "linked" are not limited to physical or mechanical connections but can include electrical connections, whether direct or indirect. The orientations or positional relationships indicated by terms such as "upper," "lower," "left," and "right" are based on the orientations or positional relationships shown in the accompanying drawings and are used only for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as limiting the invention.

[0055] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Furthermore, those skilled in the art can combine and integrate the different embodiments or examples described in this specification.

[0056] The above description, in conjunction with specific preferred embodiments, provides a further detailed explanation of the present invention. It should not be construed that the specific implementation of the present invention is limited to these descriptions. For those skilled in the art, various simple deductions or substitutions can be made without departing from the concept of the present invention, and all such modifications and substitutions should be considered within the scope of protection of the present invention.

Claims

1. A multi-dimensional intelligent quality of service assurance method for 6G HRLLC, characterized in that, include: The service quality of multidimensional data is acquired, and service quality assurance indicators for multidimensional data are constructed; wherein, the multidimensional data includes latency, peak information freshness, jitter, and reliability; Based on the current network status and business characteristics, and combined with the service quality assurance indicators of the multidimensional data, the feasible region of the multidimensional data is obtained; within the feasible region of the multidimensional data, the service quality of the multidimensional data is feature-extracted to obtain the feature vector of the service quality of the multidimensional data. A trained deep learning network is used to process the feature vectors of the service quality of the multidimensional data to obtain the collaborative decision of the service quality of the multidimensional data, and the collaborative decision of the service quality of the multidimensional data is transformed into executable network instructions. Based on the executable network instructions, cross-layer scheduling and resource allocation are performed; at the same time, the service quality status characteristics of the multidimensional data after execution are checked to see if they meet the service quality assurance indicators of the multidimensional data, so as to form a closed-loop control cycle.

2. The multi-dimensional intelligent quality of service assurance method for 6G HRLLC according to claim 1, characterized in that, The service quality assurance indicators for constructing multidimensional data include: Using the large deviation principle, under sufficient conditions, the probability that the service quality violation probability of multidimensional data is greater than a preset threshold satisfies the following: ； in, This represents the probability of service quality violations in multidimensional data. This represents the stochastic transformation process of service quality for multidimensional data. This represents the decay rate function corresponding to multidimensional data, used to quantify the exponential decay rate of violation probabilities driven by service quality. Indicates a preset threshold; Based on the decay rate function corresponding to the multidimensional data, service quality assurance indicators for latency, peak information freshness, jitter, and reliability are constructed respectively.

3. The multi-dimensional intelligent quality of service assurance method for 6G HRLLC according to claim 1, characterized in that, Based on network status and service characteristics, and combined with the service quality assurance indicators of the multidimensional data, the feasible domain of the multidimensional data is obtained, including: Based on the current network status, the service quality boundary of the multidimensional data is estimated, and the priority of the service quality assurance indicators of the multidimensional data is dynamically adjusted according to the business characteristic behavior. In combination with the service quality assurance indicators of the multidimensional data, the feasible domain of the multidimensional data is set.

4. The multi-dimensional intelligent quality of service assurance method for 6G HRLLC according to claim 1, characterized in that, Within the feasible region of the multidimensional data, feature extraction is performed on the service quality of the multidimensional data to obtain a feature vector of the service quality of the multidimensional data, including: Feature extraction is performed on latency-oriented quality of service to obtain the latency distribution function; Feature extraction is performed on the service quality oriented towards peak information freshness to obtain information freshness statistics; Feature extraction is performed on jitter-oriented quality of service to obtain jitter variance; Feature extraction is performed on reliability-oriented service quality to obtain reliability curves.

5. The multi-dimensional intelligent quality of service assurance method for 6G HRLLC according to claim 1, characterized in that, The training process of the deep learning network includes: Obtain training samples to construct a training dataset; wherein, the training samples include feature vectors of service quality of pre-defined multidimensional data, and the true labels of the training samples include the dependencies between multidimensional data; Input a portion of the samples from the training dataset into the first... The deep learning network to be trained is then trained to obtain the first... The prediction results output during this training process; According to the The prediction results output during the training process are the same as those during the training phase. The true labels of the samples for the next deep learning network to be trained are used to calculate the loss and serve as the first... Loss during each training session; According to the The loss from the training process is backpropagated to update the loss from the training process. The network parameters of the deep learning network to be trained are obtained. The deep learning network to be trained is iterated until the number of training iterations or the degree of convergence meets the preset conditions, and the trained deep learning network is obtained.

6. The multi-dimensional intelligent quality of service assurance method for 6G HRLLC according to claim 5, characterized in that, The process of obtaining the dependencies between the multidimensional data includes: By employing a multivariate statistical learning method, the dependencies between latency, peak information freshness, jitter, and reliability are constructed through learning and approximating the Pareto optimal strategy.

7. A multi-dimensional intelligent quality of service assurance system for 6G HRLLC, used to implement the multi-dimensional intelligent quality of service assurance method for 6G HRLLC as described in any one of claims 1 to 6, characterized in that, include: The data acquisition module is used to collect service quality data of multidimensional data; wherein, the multidimensional data includes latency, peak information freshness, jitter and reliability; The basic module is used to build service quality assurance indicators for multidimensional data; The perception module is used to obtain the feasible region of the multidimensional data based on the current network status and service characteristic behavior, combined with the service quality assurance indicators of the multidimensional data; within the feasible region of the multidimensional data, the service quality of the multidimensional data is extracted to obtain the feature vector of the service quality of the multidimensional data. The intelligent module is used to process the service quality feature vector of the multidimensional data using a trained deep learning network, obtain the service quality collaborative decision of the multidimensional data, and convert the service quality collaborative decision of the multidimensional data into executable network instructions. The control module is used to perform cross-layer scheduling and resource allocation according to the executable network instructions; at the same time, it checks whether the service quality status characteristics of the multidimensional data after execution meet the service quality assurance indicators of the multidimensional data, so as to form a closed-loop control cycle.

8. The multi-dimensional intelligent quality of service assurance system for 6G HRLLC according to claim 7, characterized in that, Also includes: The collaboration module is used to determine whether the executable network instructions are coordinated with the service quality assurance indicators of the multidimensional data.

9. The multi-dimensional intelligent quality of service assurance system for 6G HRLLC according to claim 7, characterized in that, The control module includes a cross-layer scheduling module, a resource allocation module, and a multi-dimensional data quality of service status feature guarantee module; wherein, the cross-layer scheduling module is used to implement cross-layer scheduling of the network, the resource allocation module is used to allocate resources according to the executable network instructions, and the multi-dimensional data quality of service status feature guarantee module is used to obtain the quality of service status features of the multi-dimensional data after execution.

Images