Method for adjusting downlink outer ring AMC in 5G wireless system and base station

Abstract

The invention discloses a method for adjusting a downlink outer ring AMC in a 5G wireless system and a base station. The method comprises the following steps: acquiring a feature vector in a current decision period; inputting the feature vector in the current decision-making period into an intelligent decision-making model, and determining a target offset of a downlink outer ring AMC in the current decision-making period from a preset MCS offset value range; wherein the feature vector in the current decision period comprises a value of a physical layer parameter and a value of a throughput parameter, and the value of the physical layer parameter is determined according to CQI fed back by the UE in at least two recent TTI periods at present; and determining the value of the throughput parameter according to the size of the TB successfully transmitted by the UE in the preset throughput window at present, thereby realizing more intelligent, stable and accurate adjustment of the downlink outer ring MCS offset.

Description

Technical Field

[0001] This article relates to mobile communication technology, and more particularly to a method for adjusting the downlink outer loop AMC and base station in a 5G wireless system. Background Technology

[0002] In fifth-generation (5G) New Radio (NR) mobile communication systems, downlink adaptive modulation and coding (AMC) technology is one of the key mechanisms for improving downlink spectral efficiency and transmission reliability. Its basic principle is that the base station dynamically adjusts the modulation scheme and coding rate of downlink data based on the radio channel state information fed back by the user equipment (UE).

[0003] Downlink AMC typically comprises two control loops: Inner Loop AMC and Outer Loop AMC. Inner Loop AMC, also known as Fast Link Adaptive, has a faster response time and directly selects a basic Modulation and Coding Scheme (MCS) index value based on the Channel Quality Indicator (CQI) fed back by the UE in real time. Outer Loop AMC, also known as Slow Link Adaptive, supplements and corrects the selection result of Inner Loop AMC. Traditional Outer Loop AMC methods typically rely on statistical analysis of Hybrid Automatic Repeat Request (HARQ) acknowledgment / non-acknowledgment (ACK / NACK) feedback. Specifically, the base station will count the number of ACKs and NACKs of newly transmitted data within a set statistical time window. When the statistical value reaches the preset adjustment threshold, a fixed offset (ΔMCS) will be applied to the MCS index to compensate for the possible estimation deviation of the inner loop AMC. The final MCS value is the sum of the inner loop base MCS value and the outer loop offset.

[0004] However, the aforementioned outer-loop AMC method may cause frequent step changes in the MCS index, which not only fails to make full use of favorable channel conditions to improve average throughput, but may also exacerbate the fluctuation of transmission performance in unstable channel environments, thereby affecting user experience and the overall transmission efficiency of the system. Summary of the Invention

[0005] This application provides a method for adjusting the downlink outer loop AMC and a base station in a 5G wireless system.

[0006] A method for adjusting the downlink outer loop AMC in a 5G wireless system includes: Obtain the feature vector within the current decision-making cycle; The feature vector within the current decision period is input into the intelligent decision model, and the target offset of the downlink outer ring AMC within the current decision period is determined from the preset MCS offset value range. The feature vector within the current decision period includes the values ​​of physical layer parameters and throughput parameters. The physical layer parameters are determined based on the CQI feedback from the UE in the most recent two TTI periods, and the throughput parameters are determined based on the TB successfully transmitted by the UE within a preset throughput window.

[0007] A storage medium storing a computer program, wherein the computer program is configured to execute the method described above when run.

[0008] An electronic device includes a memory and a processor, the memory storing a computer program and the processor being configured to run the computer program to perform the methods described above.

[0009] This application embodiment constructs a multi-dimensional feature vector that integrates the physical layer and the throughput layer, and uses an intelligent decision-making model to determine the target offset. This effectively solves the technical problems of insufficient control accuracy and large system performance fluctuations caused by the single decision information in traditional outer-loop AMC, and achieves more intelligent, stable and accurate adjustment of the downlink outer-loop MCS offset. This solution makes full use of the existing software processing capabilities and interface data of the base station, and achieves performance improvement through algorithmic innovation without adding any additional hardware resources.

[0010] Other features and advantages of this application will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the application. Other advantages of this application can be realized and obtained by means of the solutions described in the description and the accompanying drawings. Attached Figure Description

[0011] The accompanying drawings are used to provide an understanding of the technical solutions of this application and constitute a part of the specification. They are used together with the embodiments of this application to explain the technical solutions of this application and do not constitute a limitation on the technical solutions of this application.

[0012] Figure 1 A flowchart illustrating the adjustment method of the downlink outer loop AMC in a 5G wireless system provided in this application embodiment; Figure 2 This is another flowchart illustrating the adjustment method of the downlink outer loop AMC in a 5G wireless system provided in this application embodiment. Detailed Implementation

[0013] This application describes several embodiments, but these descriptions are exemplary and not restrictive, and it will be apparent to those skilled in the art that many more embodiments and implementations are possible within the scope of the embodiments described herein. Although many possible combinations of features are shown in the drawings and discussed in the detailed description, many other combinations of the disclosed features are also possible. Unless specifically limited, any feature or element of any embodiment may be used in combination with, or may replace, any feature or element of any other embodiment.

[0014] This application includes and contemplates combinations of features and elements known to those skilled in the art. The embodiments, features, and elements disclosed in this application can also be combined with any conventional features or elements to form unique inventive solutions. Any feature or element of any embodiment can also be combined with features or elements from other inventive solutions to form another unique inventive solution. Therefore, it should be understood that any feature shown and / or discussed in this application can be implemented individually or in any suitable combination. Therefore, the embodiments are not limited except by the limitations imposed by the appended claims and their equivalents. Furthermore, various modifications and changes can be made within the scope of the appended claims.

[0015] Furthermore, in describing representative embodiments, the specification may have presented methods and / or processes as a specific sequence of steps. However, the method or process should not be limited to the specific order of steps described herein, to the extent that it does not depend on such a specific order. As will be understood by those skilled in the art, other sequences of steps are also possible. Therefore, the specific order of steps set forth in the specification should not be construed as a limitation of the claims. Moreover, the claims concerning the method and / or process should not be limited to the steps performed in the written order, and those skilled in the art will readily understand that these orders can be varied and still remain within the spirit and scope of the embodiments of this application.

[0016] Traditional methods rely on a single-dimensional statistical approach to ACK / NACK counts for core control. This simple control strategy, based on a single threshold decision, lacks a more refined understanding of channel conditions and transmission performance.

[0017] In view of this, this embodiment provides a method for adjusting the downlink outer loop AMC in a 5G wireless system, enabling 5G base stations to achieve more intelligent, stable and precise adjustment of the outer loop MCS offset in downlink adaptive control.

[0018] The core of this method lies in proposing a downlink outer-loop AMC adjustment mechanism based on multi-dimensional feature vectors and an intelligent decision-making model. This mechanism constructs a feature vector that integrates physical layer and throughput parameters to achieve multi-dimensional, multi-time-scale joint perception of wireless channel state and data transmission performance. This vector is then input into the intelligent decision-making model, which determines the optimal MCS offset. This approach realizes data-driven and intelligent outer-loop AMC adjustment strategy, thereby effectively improving the downlink average throughput and transmission stability.

[0019] See Figure 1 The method includes steps S11 and S12.

[0020] Step S11: Obtain the feature vector within the current decision-making cycle.

[0021] This step is performed within each preset decision cycle, and its purpose is to construct a feature vector that can describe the current channel and transmission state.

[0022] The purpose of this step is to integrate the low-level, high-frequency physical layer measurement (CQI) with the high-level, user experience-reflecting performance metrics (throughput) across layers, providing comprehensive and quantitative contextual information for intelligent decision-making.

[0023] Specifically, the feature vector includes the values ​​of physical layer parameters and throughput parameters.

[0024] Specifically, the values ​​of physical layer parameters are determined based on a series of historical CQI data fed back by the UE within at least the last two transmission time intervals (TTI).

[0025] Furthermore, in order to indirectly but effectively assess the downlink rate status of user equipment, this application defines a set of throughput parameters. The value of this parameter is not a direct measurement of the physical layer rate, but is calculated based on the MAC layer's acknowledgment information of the transmission result.

[0026] Specifically, it is determined based on the size of all successfully transmitted transport blocks (TBs) within a preset throughput window by the UE. The calculation rule is as follows: when the base station receives an ACK from the UE, it indicates that the corresponding TB has been successfully received, and the size of the TB is included in the amount of data successfully transmitted; if a NACK is received, the transmission fails, and the size of the corresponding TB is not included.

[0027] Based on this, this method maintains two rolling time windows with different durations: short-term and long-term. The throughput parameter is the ratio of the sum of all successfully received TB values ​​within these two windows to the duration of each window. In this way, the feature vector can simultaneously reflect the instantaneous state of channel quality, historical fluctuations, and the actual effect of data transmission.

[0028] Compared to existing technologies that rely solely on the number of ACK / NACKs as a decision-making basis, this step introduces and integrates multi-dimensional, cross-layer feature parameters to construct a more informative state description vector, which can provide more accurate and comprehensive input for subsequent decisions.

[0029] The above steps provide the intelligent decision-making model with input information that can accurately characterize the current integrated wireless environment, thus laying a reliable data foundation for generating better MCS offset decisions.

[0030] Step S12: Input the feature vector within the current decision period into the intelligent decision model, and determine the target offset of the downlink outer ring AMC within the current decision period from the preset MCS offset value range.

[0031] This step inputs the feature vector obtained in the previous step into an intelligent decision-making model, which then outputs the optimal target offset within the current decision-making cycle from the preset MCS offset value range.

[0032] The purpose of this step is to map multidimensional features to an optimal control action (i.e., ΔMCS). Specifically, the intelligent decision model, acting as a function approximator or decision engine, operates internally through learning or predefined rules. It can evaluate the expected performance gains of different MCS offset values ​​for a given feature vector representing the network state, and ultimately select the one with the highest expected gain as the target offset.

[0033] Compared with the simple addition and subtraction rules based on fixed thresholds in existing technologies, this step processes high-dimensional features through an intelligent decision model, realizing a fundamental shift from "rule-based hard decision" to "learning-based soft optimization", enabling the MCS adjustment strategy to adapt to more complex channel scenarios.

[0034] Through the above steps, the intelligent and optimized adjustment of the outer ring AMC is realized, making the adjustment of MCS smoother and more accurate, effectively avoiding drastic fluctuations in throughput caused by a simple decision-making mechanism, thereby improving the overall performance and stability of the system.

[0035] Through the aforementioned downlink outer ring AMC adjustment method, this embodiment constructs a multi-dimensional feature vector that integrates the physical layer and the throughput layer, and uses an intelligent decision-making model to determine the target offset. This effectively solves the technical problems of insufficient control precision and large system performance fluctuations caused by the single decision information in traditional outer ring AMC, achieving more intelligent, stable, and accurate adjustment of the downlink outer ring MCS offset. This solution fully utilizes the existing software processing capabilities and interface data of the base station, and achieves performance improvement through algorithmic innovation without adding any additional hardware resources.

[0036] In one specific embodiment, the methods for obtaining the values ​​of the physical layer parameters and the throughput parameters in step S11 are explained separately: For obtaining the values ​​of physical layer parameters: The physical layer parameters include at least one of the following: CQI average value, CQI standard deviation, and CQI fitting slope. The CQI fitting slope is obtained by linearly fitting the CQI sequences of the UE over at least the most recent two TTI periods.

[0037] The core of this approach lies in using multi-dimensional mathematical representation of historical CQI data to break through the limitations of a single mean indicator, thereby providing a more comprehensive and profound characterization of the channel quality status and dynamic trends.

[0038] In the above embodiments: The average CQI value is obtained by taking the arithmetic mean of the CQI values ​​over at least the most recent two TTIs. It reflects the overall level of channel quality over the statistical period and is a basic indicator for judging the quality of the channel.

[0039] CQI Standard Deviation: Calculated by taking the standard deviation of CQI values ​​over at least the most recent two TTIs. This parameter quantifies the dispersion of CQI values, reflecting the volatility and stability of channel quality. A larger standard deviation indicates more severe channel jitter.

[0040] CQI Fit Slope: Obtained by linear fitting of the UE's CQI sequence over at least the most recent two TTI periods. This parameter mathematically quantifies the direction and rate of change in channel quality: a positive slope indicates that the channel quality is improving, a negative slope indicates that it is deteriorating, and a slope close to zero indicates that the channel is stabilizing.

[0041] By processing historical CQI data in multiple dimensions, a joint diagnosis of channel status can be performed from three orthogonal dimensions: time-domain level (average value), stability (standard deviation), and trend (slope). Compared with existing technologies that only focus on instantaneous or average CQI values, this provides richer contextual information, enabling intelligent decision-making models not only to perceive the channel's "current state" but also to determine its "stability" and "trend," providing a crucial basis for making more forward-looking and robust MCS adjustment decisions.

[0042] Optionally, the physical layer parameters may also include BLER instantaneous values.

[0043] Instantaneous BLER value: The instantaneous block error rate calculated based on HARQ ACK / NACK feedback within a preset short-term BLER statistical window. This parameter directly reflects the reliability of recent data transmission.

[0044] The calculation expression is: BLER instantaneous value = NACK count / (NACK count + ACK count). The NACK count and ACK count are both obtained from the HARQ information fed back by the UE within a preset BLER statistical window. This BLER window is a short-term statistical window, for example, it can cover the most recent K TTIs (K is an integer ≥ 2).

[0045] By limiting a shorter statistical window, the instantaneous BLER value can break free from the inertia of long-term average values ​​and react sensitively to events such as instantaneous deep fading and sudden interference in the channel, providing timely negative feedback for outer-loop control.

[0046] In 5G eMBB scenarios with high bandwidth and high throughput, a window that is too short can cause drastic fluctuations in the instantaneous BLER value, potentially leading to frequent oscillations in the MCS; conversely, a window that is too long will negate the significance of instantaneous monitoring. Therefore, it is necessary to configure an appropriate window length based on the specific scenario to achieve a balance between rapid response and system stability, ensuring that outer-loop control can correct problems promptly while avoiding unnecessary fluctuations caused by overreaction.

[0047] Introducing instantaneous BLER values ​​into the physical layer parameter settings adds a direct measure of transmission reliability. It effectively complements and verifies channel quality prediction based on CQI. Since CQI is the UE's prediction of channel quality, while BLER is the actual result of transmission success or failure, combining the two allows for a more accurate assessment of whether the current MCS strategy matches the real channel.

[0048] In summary, by combining "channel prediction (CQI series features)" and "transmission verification (BLER instantaneous value)," this method constructs a composite feature set that can comprehensively and faithfully reflect the physical layer transmission state, greatly improving the accuracy of state awareness. Regarding the value of the throughput parameter: Regarding the acquisition of the throughput parameter value: The throughput parameter includes at least one of the following values: short-term throughput, long-term throughput, throughput change rate, and throughput fitting slope; the corresponding throughput statistics window includes a short-term window and a long-term window, and the length of the short-term window is shorter than that of the long-term window.

[0049] The core of this approach lies in indirectly and effectively perceiving the downlink data throughput dynamics of the UE by introducing multi-timescale throughput observations and their derived indicators, thereby directly linking the control objectives of the outer-loop AMC with the final user experience (i.e., speed). Specifically: Short-term throughput: Within a short window, the sum of all successfully received terabytes (TB) is divided by the window length. It reflects the system's recent, instantaneous data transmission capacity.

[0050] Long-term throughput: The sum of all successfully received terabytes (TB) within a long window, divided by the window length. It reflects the system's average data transmission performance over a longer period and serves as an evaluation benchmark.

[0051] Throughput Change Rate: Defined as the ratio of short-term throughput to long-term throughput, used to quantify the deviation of recent throughput from the historical average. A ratio significantly greater than 1 indicates performance improvement, significantly less than 1 indicates performance degradation, and a ratio close to 1 indicates stable performance.

[0052] Throughput fit slope: The slope obtained by linearly fitting the throughput values ​​over multiple consecutive short-term windows. It characterizes the momentum and persistence of throughput changes.

[0053] By comparing short-term and long-term throughput performance and analyzing its trends, the system can intelligently identify whether throughput is improving, deteriorating, or stable. This allows the intelligent decision-making model to take more proactive actions: for example, it can take a conservative approach to prevent further deterioration when throughput first shows a downward trend; or it can be more willing to try to improve MCS to tap its potential when throughput continues to improve. This overcomes the limitation of traditional methods that only focus on instantaneous BLER and ignore changes in overall throughput performance.

[0054] The rules and relationships for setting the above-mentioned time windows are as follows: TTI: This is the most basic unit of time, and all data (such as CQI reporting, ACK / NACK feedback, and TB transmission) is generated using this unit.

[0055] Decision cycle: This is the working rhythm of the intelligent outer loop controller, which is much longer than one TTI (e.g., making a decision once every M TTIs, where M is an integer ≥ 2), and defines the frequency of feature vector updates and ΔMCS output.

[0056] Long-term window, short-term window, and BLER window: These are all statistical windows used to calculate eigenvalues. Their length is measured by the number of time intervals (TTIs) and does not exceed the length of the decision period. At the beginning of each decision period, historical data within these windows is retrieved to complete the calculation.

[0057] In summary, the structure of time information is as follows: TTI is the basic time unit; multiple TTIs constitute a feature statistical window; at each decision cycle moment, feature vectors are calculated based on the data within each statistical window, thereby triggering an intelligent decision.

[0058] In step S11, the method for obtaining the feature vector within the current decision cycle involves cross-layer fusion of low-level, high-frequency physical layer measurements (CQI) and high-level performance indicators (throughput) reflecting user experience. It also comprehensively utilizes multi-timescale statistical and trend prediction algorithms to ultimately form a low-dimensional but highly information-rich contextual feature vector. This approach enables the intelligent decision model to simultaneously perceive "channel state," "transmission results," and "performance trends." Based on this, it can make adjustment decisions far exceeding those of traditional single-threshold decision methods: more accurate, smoother, and more forward-looking, thereby effectively improving the average throughput and transmission stability of the downlink.

[0059] In one specific embodiment, in step S12, the value range of the MCS offset is a discrete set of integer values, which is represented as {-3, -2, -1, 0, +1, +2, +3}.

[0060] The core purpose of this constraint is to limit the adjustment range of the MCS offset to a finite, discrete, and symmetric integer range, which conforms to the communication protocol specifications and helps to achieve a balance between algorithm efficiency and system stability.

[0061] In the 3GPP standard, the MCS index itself is a discrete integer value (e.g., 0 to 28). Therefore, the offset adjustment (ΔMCS) applied to it naturally also uses a discrete integer value.

[0062] A range of ±3 is sufficient to cover the outer-loop compensation amplitude required under most channel conditions. For example, when the channel deteriorates sharply, the MCS needs to be significantly reduced to ensure reliability (corresponding to -3); when the channel conditions are excellent, the MCS can be aggressively increased to pursue throughput (corresponding to +3). Conversely, an excessively large adjustment range (such as ±5 and above), while flexible, can lead to an expansion of the decision space, increase the learning difficulty, and even cause MCS oscillations, which is not conducive to the rapid convergence of the system. Therefore, the range of ±3 provides the necessary flexibility while constraining the amplitude of a single adjustment, which can avoid the excessive impact of a single decision error and effectively ensure the stability of the system.

[0063] In step S12, the above-mentioned value range is used as the optional value of the MCS offset, which achieves a good balance between algorithm complexity and system performance.

[0064] In one specific embodiment, the intelligent decision-making model in step S12 is further defined.

[0065] Specifically, the intelligent decision-making model is based on the LinUCB (Linear Upper Confidence Bound) algorithm, and its action space is the range of MCS offset values. The model calculates a UCB score for each candidate value within the action space and determines the target offset based on this score.

[0066] The LinUCB algorithm was chosen because its characteristics are highly compatible with the technical problem addressed in this application. Specifically: Compatible with discrete action space: The LinUCB algorithm is naturally suited for selecting the best from a finite discrete action space, which perfectly matches the requirement of the outer loop AMC to output discrete MCS offsets.

[0067] It excels at handling contextual information: the multidimensional feature vectors constructed in this application are precisely the "context" required by the LinUCB algorithm. The algorithm can learn the linear relationship between these contextual features and the expected returns of different ΔMCS, thereby enabling personalized decision-making based on specific network states.

[0068] Balancing Exploration and Exploitation: The core advantage of the UCB mechanism lies in its intelligent balance between "exploration" and "exploration." It not only selects the action with the highest reward given current knowledge but also provides opportunities for actions that are less frequently attempted and have higher uncertainty. This is crucial for time-varying wireless channels, enabling the algorithm to continuously explore better strategies and avoid getting trapped in local optima.

[0069] The intelligent decision-making model operates as follows in each decision cycle: For each candidate ΔMCS value within the action space, the model calculates the UCB score by combining the current feature vector (context) and the historical experience of that action. This score consists of two parts: the historical average reward and compensation for the uncertainty of the valuation. Finally, the algorithm selects the candidate value with the highest UCB score as the target offset.

[0070] In the above embodiments, the intelligent decision-making model based on LinUCB can learn the complex mapping relationship between multi-dimensional contextual features and the optimal ΔMCS, thereby making more accurate adjustment decisions, effectively improving system throughput and reducing the error rate.

[0071] In one specific embodiment, the method further includes step S13 after step S12: determining the final applied MCS index value based on the inner ring base MCS value and the target offset.

[0072] The core of this embodiment lies in integrating the dynamic offset learned by the intelligent outer loop with the fast response result of the inner loop AMC, so as to achieve the collaborative work of the inner and outer loop AMC mechanisms.

[0073] This step serves to apply the output of the intelligent outer-loop controller to the coding and modulation process of downlink data transmission, completing the closed loop from intelligent decision-making to physical layer execution. Specifically, the inner-loop AMC quickly determines a basic MCS value (MCS_inner) based on the CQI fed back by the UE in real time. This value mainly reflects the instantaneous state of the channel. The target offset (ΔMCS) learned by the intelligent outer loop in this application is a fine calibration of the basic MCS value based on contextual features with a longer time scale and more dimensions. Finally, through the algebraic operation MCS_final = MCS_inner + ΔMCS, an MCS index value that responds to rapid channel changes while taking into account long-term performance and stability is generated and immediately applied to subsequent downlink data transmission within the current decision cycle.

[0074] Compared to existing technologies, this step transforms the outer-loop AMC from an independent compensation module based on fixed rules into a learning-based intelligent optimizer deeply integrated with the inner loop, thereby achieving more refined, smooth, and adaptive MCS control. In traditional methods, the adjustment of the outer-loop ΔMCS is relatively coarse and has a weak connection with the inner-loop decision logic. This step, however, combines the "agility" of the inner loop with the "intelligence" of the outer loop, enabling the final MCS decision to possess both rapid response capabilities and a long-term optimization perspective.

[0075] By performing this step, we ensure that the optimized results output by the intelligent decision model can be seamlessly and efficiently applied to the physical layer transmission. This allows the optimal ΔMCS obtained through multidimensional feature analysis and online learning to be directly translated into actions that improve system performance, making the entire adaptive modulation and coding system a responsive and intelligently decision-making organic whole.

[0076] In one specific embodiment, the method further includes step S14 after step S13: after transmitting downlink data using the final value of MCS, the intelligent decision model is updated according to the HARQ ACK / NACK information fed back by the UE.

[0077] The core of this embodiment lies in establishing an online learning mechanism with ACK / NACK as the reward signal, enabling the intelligent decision-making model to continuously optimize its decisions based on actual transmission results, thus achieving long-term system self-adaptation. This step aims to complete the learning loop of the intelligent outer-loop AMC system, allowing the model to self-adjust and optimize based on actual transmission performance.

[0078] Specifically, after the base station uses the final value of the MCS for downlink data transmission, it receives HARQ ACK / NACK information from the UE and generates a reward signal accordingly: if the feedback is ACK, a positive reward is generated, indicating that the selected ΔMCS action was successful in the current channel environment; if it is NACK, a negative reward is generated, indicating that the decision-making effect was poor. Subsequently, the system uses this reward signal and the feature vector used in the current decision cycle to update only the local parameters in the intelligent decision model corresponding to the selected ΔMCS.

[0079] Compared to the fixed or slowly adjusted outer loop AMC parameters in existing technologies, this step introduces an online learning framework, which transforms each transmission attempt into a learning opportunity, enabling real-time, automatic, and data-driven fine-tuning of model parameters.

[0080] Thus, the model continuously accumulates experience during operation, gradually and accurately mastering which ΔMCS action to choose under specific channel and throughput conditions to obtain the best long-term benefits. This capability enables the system to proactively track and adapt to long-term channel changes and traffic load fluctuations, thereby continuously approaching optimal scheduling performance and significantly enhancing robustness in complex real-world environments. This online learning mechanism, combined with the aforementioned decision-making steps, constitutes a complete system that combines intelligent decision-making and continuous optimization capabilities.

[0081] Figure 2 This is another flowchart illustrating the method for adjusting the downlink outer loop AMC in a 5G wireless system provided in this application embodiment. Figure 2As shown, this scheme comprises three core components: feature extraction, intelligent decision-making, and online learning, forming a continuously optimized closed-loop control system. Specifically, this method includes steps S21 to S25.

[0082] Step S21: Obtain the basic MCS value of the inner loop.

[0083] The base station sends a downlink reference signal to the terminal, which then measures the signal quality and sends back a CQI report. Upon receiving the CQI report, the base station determines the base MCS value used by the inner-loop AMC by querying a predefined CQI-MC mapping table. This step provides a benchmark for subsequent outer-loop intelligent adjustments.

[0084] Step S22: Extract multidimensional contextual features.

[0085] This step is the data preparation phase for intelligent decision-making, and it specifically includes two parallel sub-steps: Step S22-1: Obtain physical layer parameters. Based on the historical CQI sequence and ACK / NACK feedback reported by the terminal, calculate four-dimensional physical layer parameters, specifically including: average CQI value, used to assess the overall channel quality; standard deviation of CQI, used to quantify channel volatility; slope of CQI linear fit, used to predict channel change trends; and instantaneous BLER value, used to reflect recent transmission reliability. These features together provide a comprehensive diagnosis of the channel state.

[0086] Step S22-2: Obtain throughput parameters. Based on historical successfully transmitted block size data, calculate four-dimensional throughput parameters at different time scales, specifically including: short-term throughput, used to perceive instantaneous transmission capacity; long-term throughput, used to evaluate average performance benchmark; throughput change rate, used to identify abnormal performance fluctuations; and throughput fitting slope, used to determine the long-term performance trend of the system.

[0087] Step S23: Intelligent decision-making generates MCS offset.

[0088] The physical layer features and throughput features obtained in step S22 are combined into a complete context feature vector, which is then input into an intelligent decision-making model based on the LinUCB algorithm. This model calculates a UCB score for each candidate value in the pre-defined MCS offset action space, and finally selects the offset with the highest score as the target ΔMCS for this decision cycle. This mechanism achieves an optimal balance between exploration and utilization.

[0089] Step S24: Calculate the final MCS and perform the transfer.

[0090] The target ΔMCS obtained from intelligent decision-making is added to the basic MCS value determined by the inner loop to obtain the final MCS, which is then applied to the selection of modulation and coding schemes for downlink data transmission.

[0091] Step S25: Online learning and model updates.

[0092] A reward signal is generated based on the ACK / NACK feedback received after downlink data transmission. This signal, along with the feature vector used in this decision cycle, is then used to update the parameters in the LinUCB model corresponding to the selected ΔMCS action. This step completes the learning loop, enabling the model to continuously optimize future decisions based on actual transmission performance.

[0093] Through the cyclic execution of the above process, this method achieves a fundamental transformation of the outer-loop AMC from simple rule adjustment to intelligent self-learning, which can maintain the accuracy and stability of MCS adjustment in complex wireless environments, ultimately improving system throughput and transmission reliability.

[0094] This application also provides a storage medium storing a computer program, which is configured to execute the steps in any of the above method embodiments when running.

[0095] This application also provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to execute the steps in any of the above method embodiments by running the computer program.

[0096] This application also provides a base station that includes the aforementioned electronic device.

[0097] The complete process of the method, including inner ring basic MCS determination, outer ring feature extraction, intelligent decision-making, inner and outer ring fusion, and online learning model update, is all completed by the base station processing unit.

[0098] The above methods are mainly aimed at 5G application scenarios with high requirements for downlink throughput, stability and reliability, and are especially suitable for mobile communication environments with complex and dynamically changing channel environments.

[0099] For example, in enhanced mobile broadband scenarios, including 4K / 8K ultra-high-definition video streaming, virtual reality / augmented reality, and internet access during high-speed travel (such as high-speed rail), user experience directly depends on consistently high throughput. This solution, through smooth and precise MCS adjustment, can effectively avoid problems such as video stuttering and VR dizziness, maximizing the utilization of available spectrum resources while ensuring connection reliability.

[0100] For example, in ultra-reliable low-latency communication scenarios, including industrial automation, telemedicine, smart grid control, and vehicle-to-everything (V2X) communication, such applications have extremely high requirements for transmission reliability and latency. This solution, by introducing instantaneous BLER and multi-dimensional channel trend judgment, can proactively and gradually adopt a conservative MCS strategy when channel quality deteriorates due to interference or fading, thereby avoiding packet errors and retransmissions and ensuring ultra-high communication reliability.

[0101] For example, in massive machine-type communication scenarios, such as large-scale IoT device connections (including smart meter reading, environmental monitoring, etc.), although the data rate requirement for a single device is not high, the base station needs to serve a massive number of connections simultaneously, each with different channel conditions. The intelligent decision-making model in this solution can personalizedally allocate appropriate MCS offsets to UEs under different channel conditions, thereby optimizing the overall resource utilization efficiency of the system.

[0102] In summary, by introducing an intelligent outer loop controller on the base station side, this solution provides a highly efficient downlink adaptive transmission solution that can self-optimize and continuously learn for various typical 5G applications, especially services in complex and dynamic wireless environments, effectively improving the overall performance of the system in real networks.

[0103] It will be understood by those skilled in the art that all or some of the steps, systems, or apparatuses disclosed above, and their functional modules / units, can be implemented as software, firmware, hardware, or suitable combinations thereof. In hardware implementations, the division between functional modules / units mentioned above does not necessarily correspond to the division of physical components; for example, a physical component may have multiple functions, or a function or step may be performed collaboratively by several physical components. Some or all components may be implemented as software executed by a processor, such as a digital signal processor or microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit (ASIC). Such software may be distributed on a computer-readable medium, which may include computer storage media (or non-transitory media) and communication media (or transient media). As is known to those skilled in the art, the term "computer storage medium" includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules, or other data). Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disc (DVD) or other optical disc storage, magnetic cartridges, magnetic tape, disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and can be accessed by a computer. Furthermore, it is well known to those skilled in the art that communication media typically contain computer-readable instructions, data structures, program modules, or other data in modulated data signals such as carrier waves or other transmission mechanisms, and may include any information delivery medium.

Claims

1. A method for adjusting the downlink outer loop AMC in a 5G wireless system, characterized in that, include: Obtain the feature vector within the current decision-making cycle; The feature vector within the current decision period is input into the intelligent decision model, and the target offset of the downlink outer ring AMC within the current decision period is determined from the preset MCS offset value range. The feature vector within the current decision period includes the values ​​of physical layer parameters and throughput parameters. The physical layer parameters are determined based on the CQI feedback from the UE in the most recent two TTI periods, and the throughput parameters are determined based on the TB successfully transmitted by the UE within a preset throughput window.

2. The method according to claim 1, characterized in that: The physical layer parameters include at least one of the following: CQI average value, CQI standard deviation, and CQI fitting slope; wherein the CQI fitting slope is obtained by linearly fitting the CQI of the UE in the most recent at least two TTI periods. The throughput parameter includes at least one of short-term throughput, long-term throughput, throughput change rate, and throughput fitting slope; the throughput window includes a short-term window and a long-term window, wherein the length of the short-term window is less than the length of the long-term window; wherein: The short-term throughput = (the sum of the sizes of all successful TBs of the UE within the short-term window) / (the length of the short-term window); The long-term throughput = (the sum of the sizes of all successful TBs of the UE within the long-term window) / (the length of the long-term window); The throughput change rate = (the short-term throughput) / (the long-term throughput). The throughput fitting slope is obtained by linearly fitting the short-term throughput of the UE within multiple consecutive short-term windows.

3. The method according to claim 1 or 2, characterized in that, The physical layer parameters also include BLER instantaneous values, where: The instantaneous BLER value = (the total number of NACKs for HARQs reported by the UE within the preset BLER window) / (the total number of NACKs for HARQs reported by the UE within the BLER window + the total number of ACKs for HARQs reported by the UE within the BLER window).

4. The method according to claim 1, characterized in that, The range of values ​​is a discrete set of integer values, represented as {-3, -2, -1, 0, +1, +2, +3}.

5. The method according to claim 1 or 4, characterized in that, The intelligent decision-making model is a model based on the Linear Upper Bound Confidence Interval (LinUCB) algorithm. Its action space is the range of values ​​for the MCS offset. The intelligent decision-making model calculates a UCB score for each value in the range and obtains the target offset based on the UCB score for each value.

6. The method according to claim 1, characterized in that, The method further includes: The target offset is added to the base MCS value determined by the inner loop AMC to obtain the final MCS value used for downlink data transmission in the current decision cycle.

7. The method according to claim 6, characterized in that, The method further includes: After controlling the downlink transmission operation using the final value of the MCS, the intelligent decision-making model is updated based on the ACK / NACK feedback information of the HARQ from the UE, wherein: If the feedback information corresponding to the final value of MCS is ACK, a reward signal with a positive reward value is generated; otherwise, a reward signal with zero or negative reward is generated. The reward signal is used to update the model parameters used in the intelligent decision-making model when determining the target offset based on the feature vector within the current decision period.

8. A storage medium, characterized in that, The storage medium stores a computer program, wherein the computer program is configured to execute the method described in any one of claims 1 to 7 when it is run.

9. An electronic device comprising a memory and a processor, characterized in that, The memory stores a computer program, and the processor is configured to run the computer program to perform the method as described in any one of claims 1 to 7.

10. A base station, characterized in that, Includes the electronic device as described in claim 9.

Images