A router rate switching method for low latency services

Abstract

This invention discloses a router rate switching method for low-latency services, comprising the following steps: Step S1: Real-time acquisition and normalization processing of multi-dimensional network environment feature parameters; Step S2: Low-latency service flow feature parsing and prioritization; Step S3: Construction and online training of a reinforcement learning rate decision model; Step S4: Optimal rate matching decision for low-latency services; Step S5: Execution of smooth low-latency rate switching for Wi-Fi 6 / 7 dual-mode; Step S6: Post-switching performance acquisition and iterative optimization of model feedback. This invention designs a rate switching method based on the Wi-Fi 6 / 7 dual-mode protocol stack, which can autonomously switch between dual-mode according to the network environment and terminal capabilities. It is compatible with older generation Wi-Fi standards such as IEEE 802.11a / b / g / n / ac, solving the problem of insufficient compatibility of single-generation Wi-Fi routers and covering all types of terminal access scenarios.

Description

Technical Field

[0001] This invention relates to the field of wireless local area network communication technology, specifically to a router rate switching method for low-latency services. Background Technology

[0002] With the large-scale deployment of next-generation information technologies such as 5G, IoT, high-definition video, cloud gaming, VR / AR, and industrial wireless control, the demand for low latency, high bandwidth, and high stability in wireless LANs from users and industrial scenarios is growing exponentially. As the core access technology for wireless LANs, Wi-Fi 6 has become widespread, achieving peak speeds of up to 9.6Gbps and supporting high-concurrency device access. Wi-Fi 7, as a next-generation technology, boasts peak speeds exceeding 30Gbps, air interface latency as low as 1ms, and supports 4096-QAM high-order modulation and 320MHz ultra-wide channel bandwidth, becoming a core technology for low-latency services.

[0003] The current router speed switching mechanism has significant technical problems: First, most routers only support a single Wi-Fi generation, lacking compatibility with Wi-Fi 6 / 7 dual-mode and unable to flexibly switch between dual-mode based on terminal capabilities and network environment; Second, existing speed switching mostly adopts passive switching with fixed thresholds, triggering adjustments based solely on signal strength, without considering dynamic factors such as channel interference, network load, and service type, resulting in lagging and frequent speed switching; Third, no dedicated switching strategies have been designed for low-latency sensitive services such as VR / AR, cloud gaming, and industrial control, making it easy for low-latency services to experience latency spikes, increased packet loss rates, transmission stutters, or even disconnections during speed switching; Fourth, there is no smooth buffering mechanism for speed switching, resulting in excessively high additional latency during dual-mode switching, failing to meet the millisecond-level transmission requirements of low-latency services.

[0004] In existing technologies, rate switching methods lack a joint decision-making model considering both network environment and service requirements, lack machine learning-based adaptive optimization capabilities, and fail to establish a closed-loop control system encompassing environment awareness, service identification, rate decision-making, smooth switching, and feedback iteration. These deficiencies lead to a mismatch between wireless transmission rates and network environment / service requirements, resulting in compromised low-latency service transmission quality and severely restricting the application of Wi-Fi 6 / 7 dual-mode routers in high-end homes, smart parks, and industrial control scenarios. Summary of the Invention

[0005] The purpose of this invention is to provide a router rate switching method for low-latency services, so as to solve the problems existing in the prior art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a router rate switching method for low-latency services, comprising the following steps:

[0007] Step S1: The router collects multi-dimensional characteristic parameters of the wireless transmission environment in real time at fixed intervals. The multi-dimensional characteristic parameters include Received Signal Strength Indicator (RSSI), Signal-to-Noise Ratio (SNR), Channel Interference Power, Network Device Access Load, and Channel Bandwidth Utilization. The collected raw characteristic parameters are linearly normalized to obtain standardized network environment parameters.

[0008] Step S2: The router performs deep analysis on the service flow of the access terminal and extracts three major characteristic parameters: service latency sensitivity, bandwidth requirement, and packet loss tolerance; based on the analysis results, it identifies the low-latency service type, assigns a dedicated priority label to the low-latency service, and distinguishes between low-latency services and ordinary data services.

[0009] Step S3: Construct a rate decision model based on deep reinforcement learning, using standardized network environment parameters and low-latency service priority labels as the model's state inputs, and Wi-Fi 6 / 7 dual-mode selection, transmission rate level, modulation method, and channel bandwidth as the model's executable actions; design a reward function with network throughput, service latency, and packet loss rate as optimization objectives, and optimize the model's decision weights through online interactive training.

[0010] Step S4: The rate decision model inputs and outputs decision results based on the current state, and combines them with the priority constraints of low-latency services to select the Wi-Fi 6 / 7 dual-mode, transmission rate, modulation method and channel bandwidth that are suitable for the current network environment, forming a dedicated rate matching scheme for low-latency services.

[0011] Step S5: Based on the rate matching scheme, a smooth switching mechanism with segmented buffering and frame scheduling optimization is adopted to perform rate switching between Wi-Fi 6 and Wi-Fi 7 modes; during the switching process, the transmission context of low-latency service flows is preserved, and the additional delay of dual-mode switching is controlled to not exceed a preset threshold.

[0012] Step S6: Collect network performance parameters after rate switching in real time. The network performance parameters include throughput, service latency, packet loss rate, and rate stability. Feed the performance parameters back to the reinforcement learning rate decision model, update the model's decision parameters and weight coefficients, and complete the online iterative optimization of the model.

[0013] Furthermore, in step S1, after linearly normalizing the multi-dimensional feature parameters, the comprehensive evaluation value of the network environment is calculated through weighted fusion. The calculation formula is:

[0014]

[0015] in, The weighting coefficients are and satisfy the following conditions: RSSI maxRouter maximum received signal strength, SNR max I is the router's maximum signal-to-noise ratio. max L is the maximum permissible interference power of the channel. max This is the maximum load that the router can support for device access.

[0016] Furthermore, in step S2, the priority weight P of low-latency services is quantified based on service characteristic parameters, and the calculation formula is as follows:

[0017]

[0018] in, Weighting coefficients The following conditions must be met: T represents the service latency sensitivity, B represents the service bandwidth requirement, and D represents the service packet loss tolerance. All three values ​​are in the range of [0,1].

[0019] Furthermore, in step S3, the reward function R of the reinforcement learning rate decision model is calculated as follows:

[0020]

[0021] in, The weighting coefficients are and satisfy the following conditions: Low-latency service scenarios Value greater than and Th represents the standardized network throughput, Td represents the real-time service latency, and Pl represents the real-time service packet loss rate.

[0022] Furthermore, in step S1, the fixed period for collecting multi-dimensional feature parameters is 10ms. During the collection process, the parameters of three consecutive periods are subjected to moving average filtering to remove abnormal data.

[0023] Furthermore, in step S2, the low-latency services include VR / AR transmission services, cloud gaming services, and industrial wireless control services; the deep analysis of service flows adopts a combination of port identification and payload feature matching, with an identification accuracy of no less than 99%.

[0024] Furthermore, in step S4, when selecting the rate matching scheme, priority is given to meeting the latency constraints of low-latency services, and then the transmission rate and channel bandwidth are matched. In Wi-Fi 7 mode, 320MHz channel bandwidth and 4096-QAM modulation method are selected first.

[0025] Furthermore, in step S5, the preset threshold for the additional latency of dual-mode switching is 50ms; the smooth switching mechanism adopts service flow segmentation caching, prioritizing the scheduling of low-latency service data frames during switching, while ordinary data service frames enter the waiting queue.

[0026] Furthermore, in step S6, the model feedback iteration period is 1 second. In each iteration, the neural network weights of the model are adjusted according to the reward function value. The higher the reward function value, the smaller the adjustment range of the model decision parameters.

[0027] Compared with the prior art, the beneficial effects of the present invention are:

[0028] This invention is based on a rate switching method designed for Wi-Fi 6 / 7 dual-mode protocol stack. It can autonomously switch between dual-mode according to the network environment and terminal capabilities, and is compatible with older generation Wi-Fi standards such as IEEE 802.11a / b / g / n / ac. It solves the problem of insufficient compatibility of single-generation Wi-Fi routers and covers all types of terminal access scenarios.

[0029] For low-latency sensitive services such as VR / AR, cloud gaming, and industrial control, we have designed a dedicated priority marking and rate decision strategy to prioritize the transmission resources of low-latency services, control the handover latency to within 50ms, reduce the air interface latency to as low as 5ms, improve the rate stability by more than 30%, and completely solve the problems of stuttering and disconnection in low-latency services.

[0030] Breaking through the limitations of traditional fixed threshold switching, it senses changes in the network environment through multi-dimensional parameters such as RSSI, SNR, interference, and load, and combines reinforcement learning models to autonomously learn changes in the network environment, thereby achieving precise matching between the rate and the network environment and service requirements, and avoiding frequent rate jumps and delayed switching.

[0031] Employing a segmented buffering and frame scheduling optimization smooth switching mechanism, it preserves the low-latency service transmission context, ensuring seamless switching between Wi-Fi 6 / 7 dual modes without data packet loss or sudden increase in latency, thus guaranteeing service transmission continuity.

[0032] A closed-loop control system of decision-making, switching, feedback, and optimization is constructed. Through real-time performance acquisition and online model iteration, the rate decision weights are continuously optimized, enabling the router to adapt to complex and ever-changing network environments and maintain optimal transmission performance over the long term.

[0033] By combining dynamic channel selection and rate adaptive adjustment, the impact of channel interference on low-latency services is reduced. It can maintain stable transmission rate and low latency characteristics even in multi-device concurrency and complex electromagnetic environments, making it suitable for various scenarios such as home, enterprise, industry, and public venues. Attached Figure Description

[0034] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0035] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0036] Please see Figure 1 This invention provides a router rate switching method for low-latency services, comprising the following steps:

[0037] Step S1: Real-time acquisition and normalization of multi-dimensional network environment feature parameters

[0038] The router collects multi-dimensional characteristic parameters of the wireless transmission environment in real time at fixed intervals, including Received Signal Strength Indicator (RSSI), Signal-to-Noise Ratio (SNR), Channel Interference Power, Network Device Access Load, and Channel Bandwidth Utilization. The collected raw parameters are linearly normalized to eliminate dimensional differences and obtain standardized network environment parameters.

[0039] Step S2: Low-latency service flow feature analysis and prioritization

[0040] The level-marking router performs in-depth analysis of the service flow of the access terminal, extracting three major features: service latency sensitivity, bandwidth requirements, and packet loss tolerance; based on the analysis results, it identifies the type of low-latency service, assigns a dedicated priority mark to the low-latency service, and distinguishes low-latency services from ordinary data services.

[0041] Step S3: Construction and online training of the reinforcement learning rate decision model

[0042] A rate decision model based on deep reinforcement learning is constructed. Standardized network environment parameters and low-latency service priority labels are used as the model's state inputs, while Wi-Fi 6 / 7 dual-mode selection, transmission rate level, modulation method, and channel bandwidth are used as the model's executable actions. With network throughput, service latency, and packet loss rate as the core optimization objectives, a reward function is designed, and the model's decision weights are optimized through online interactive training to achieve autonomous learning of the optimal rate strategy.

[0043] Step S4: Optimal matching decision for dedicated rates for low-latency services

[0044] Based on the current state input and output decision results, combined with the low-latency service priority constraints, the model selects the Wi-Fi 6 / 7 dual-mode, transmission rate, modulation method and channel bandwidth that are suitable for the current network environment, forming a dedicated rate matching scheme for low-latency services.

[0045] Step S5: Wi-Fi 6 / 7 Dual-Mode Low-Latency Smooth Rate Switching Execution

[0046] Based on the rate matching scheme, a smooth switching mechanism with segmented buffering and frame scheduling optimization is adopted to perform rate switching between Wi-Fi 6 and Wi-Fi 7 modes. During the switching process, the transmission context of low-latency service streams is preserved, and the additional latency of dual-mode switching is controlled to not exceed a preset threshold to avoid service interruption.

[0047] Step S6: Post-Switch Performance Acquisition and Model Feedback Iterative Optimization

[0048] The system collects network performance parameters after rate switching in real time, including throughput, service latency, packet loss rate, and rate stability. These performance parameters are then fed back to the reinforcement learning rate decision model to update the model's decision parameters and weight coefficients, enabling online iterative optimization of the model and improving the accuracy of subsequent rate decisions.

[0049] This invention incorporates three core calculation formulas, used for comprehensive network environment evaluation, low-latency service priority quantification, and reinforcement learning reward function calculation, respectively. The parameters of each formula are clearly defined and their physical meanings are explicit, as detailed below:

[0050] Formula 1: Calculation of Comprehensive Network Environment Evaluation Value

[0051] The standardized network environment parameters are weighted and fused to obtain a comprehensive network environment evaluation value. The calculation formula is:

[0052] (1)

[0053] in, Let be the weighting coefficient, satisfying The RSSI is dynamically adjusted according to the needs of low-latency services; RSSI is a real-time received signal strength indicator. max The maximum received signal strength of the router; SNR is the real-time signal-to-noise ratio. max I is the router's maximum signal-to-noise ratio; I is the real-time channel interference power. max L represents the maximum permissible interference power of the channel; L represents the real-time network device access load. max This is the maximum load that the router can support for device access.

[0054] Formula 2: Calculation of Priority Weight for Low-Latency Services

[0055] The priority weight P of low-latency services is quantified based on business characteristics, and the calculation formula is as follows:

[0056] (2)

[0057] in, Let be the weighting coefficient, satisfying The parameters are preset according to the service type; T is the service latency sensitivity, with a value range of [0,1]. The higher the value, the higher the latency sensitivity; B is the service bandwidth requirement, with a value range of [0,1]. The higher the value, the greater the bandwidth requirement; D is the service packet loss tolerance, with a value range of [0,1]. The higher the value, the lower the packet loss tolerance.

[0058] Formula 3: Calculation of Reinforcement Learning Reward Function

[0059] A reward function R is designed with network performance as the objective, guiding the model to optimize decisions. The calculation formula is as follows:

[0060] (3)

[0061] in, Let be the weighting coefficient, satisfying Low latency services The maximum value is Th; the normalized network throughput is Th, with a value range of [0,1]; the real-time service latency is Td, in milliseconds; and the real-time service packet loss rate is Pl, with a value range of [0,1].

[0062] Example:

[0063] This embodiment provides an adaptive rate switching method for Wi-Fi 6 / 7 routers for low-latency services. It is applied to smart routers compatible with both Wi-Fi 6 and 7 modes, and aims to achieve adaptive rate switching and smooth transition between modes for low-latency services such as VR / AR in home scenarios, low-latency cloud gaming, and low-latency industrial wireless control in enterprise scenarios. The specific implementation steps are as follows:

[0064] Step S1: Real-time acquisition and normalization of multi-dimensional network environment feature parameters

[0065] The core of this step is to achieve full-dimensional perception of the wireless transmission environment, providing accurate data support for subsequent rate decisions. The specific execution details are as follows:

[0066] Parameter acquisition period and dimension settings:

[0067] The router is set to a fixed data collection period of 10ms. This period is adapted to the millisecond-level response requirements of low-latency services, avoiding decision delays caused by data collection lag. The collected multi-dimensional characteristic parameters include: Received Signal Strength Indicator (RSSI), Signal-to-Noise Ratio (SNR), Channel Interference Power (I), Network Device Access Load (L), and Channel Bandwidth Utilization (C), totaling five core parameters, covering three core dimensions: signal quality, interference level, and network load.

[0068] Abnormal data filtering processing:

[0069] To avoid parameter anomalies caused by transient interference, a moving average filter is applied to the raw parameters from three consecutive acquisition cycles to remove abnormal data exceeding the normal threshold, resulting in a stable raw parameter sequence. For example, the normal range for RSSI is -30dBm to -80dBm. If an abnormal value of -100dBm is acquired, it is directly discarded, and the average value of the previous two cycles is used instead.

[0070] Linear normalization processing:

[0071] The filtered original parameters are linearly normalized to map all parameters to the [0,1] interval, eliminating dimensional differences. The calculation formula is as follows:

[0072]

[0073] Where X is the original parameter, X min X is the minimum value of the parameter. max X is the maximum value of the parameter. norm These are standardized parameters.

[0074] Calculation of comprehensive network environment assessment value:

[0075] Substituting the standardized parameters into Formula 1, the comprehensive network environment evaluation value E is calculated. In this embodiment, a weighting coefficient is set for low-latency services:

[0076] =0.3、 =0.3、 =0.25、 =0.15, which satisfies the condition. .

[0077] For example, the parameters collected at a certain moment are:

[0078] RSSI = −50dBm (RSSI max =−30dBm), SNR=40dB (SNR) max =50dB), I=10dBm (I =50dB), max =20dBm), L=50 units (L max =200 units), substituting into the formula, we get: E=0.5775.

[0079] The closer the value is to 1, the better the network environment is, and the more suitable it is to enable Wi-Fi 7 high-speed mode;

[0080] The lower the value, the worse the network environment, and you need to switch to Wi-Fi 6 stable mode.

[0081] Step S2: Low-latency service flow feature parsing and priority marking

[0082] The core of this step is to accurately identify low-latency services, allocate dedicated transmission priorities to them, and ensure that rate resources are tilted towards low-latency services. The specific execution details are as follows:

[0083] In-depth analysis of business flow:

[0084] The router employs a dual parsing method of port identification and load feature matching to perform full parsing of both uplink and downlink traffic flows from access terminals.

[0085] Port identification: Dedicated transmission ports are matched for VR / AR services, cloud gaming services, and industrial wireless control services to quickly locate low-latency service flows;

[0086] Payload feature matching: Analyze the payload features of service data frames, identify the transmission protocol and data format of low-latency services, avoid identification errors caused by port spoofing, and achieve an identification accuracy of no less than 99%.

[0087] Business feature parameter extraction:

[0088] Extract three core feature parameters from the parsed business flow:

[0089] Business latency sensitivity VR / AR business T=0.95, cloud gaming business T=0.9, industrial control business T=1.0, and ordinary web page business T=0.2;

[0090] Service bandwidth requirements B: VR / AR services B=0.9, cloud gaming services B=0.8, industrial control services B=0.3, general download services B=1.0;

[0091] Packet loss tolerance D for services: VR / AR services D=0.95, cloud gaming services D=0.9, industrial control services D=1.0, and ordinary video services D=0.5.

[0092] Low-latency service priority weight calculation

[0093] Substitute the feature parameters into Formula 2 to calculate the business priority weight P. This embodiment sets the weight coefficient as follows:

[0094] .

[0095] Taking industrial wireless control services as an example,

[0096]

[0097] A higher P-value indicates a higher service priority, meaning that the transmission resources for that service will be prioritized during rate decisions.

[0098] Priority marking and business classification:

[0099] according to Services are categorized by priority value: P ≥ 0.8 represents the highest priority low-latency service, 0.5 ≤ P < 0.8 represents the medium priority ordinary service, and P < 0.5 represents the low priority data service. The router adds a dedicated priority tag to the highest priority service and stores it at the head of the service scheduling queue.

[0100] Step S3: Construction and online training of the reinforcement learning rate decision model

[0101] The core of this step is to build an intelligent decision-making model that autonomously adapts to the network environment and business needs through reinforcement learning, replacing the traditional fixed threshold switching. Specific execution details are as follows:

[0102] Model structure design:

[0103] A reinforcement learning rate decision model based on a deep Q-network (DQN) is constructed. The model consists of an input layer, a hidden layer, and an output layer.

[0104] Input layer: The input dimension is 6, including standardized network environment parameters (5 items) and low latency service priority weight P (1 item).

[0105] Hidden layer: Employs a two-layer fully connected neural network with ReLU activation function to extract joint features of the network environment and business operations;

[0106] Output layer: The output dimension is 4, including Wi-Fi 6 / 7 mode selection, transmission rate level, modulation method, and channel bandwidth.

[0107] Executable action definition:

[0108] The model's executable actions cover the full-rate configuration of Wi-Fi 6 / 7 dual-mode:

[0109] Mode selection: Wi-Fi 6 mode, Wi-Fi 7 mode;

[0110] Transmission speed levels: 9.6Gbps, 4.8Gbps, and 2.4Gbps under Wi-Fi 6; 24Gbps, 12Gbps, and 6Gbps under Wi-Fi 7.

[0111] Modulation methods: 1024-QAM (Wi-Fi 6), 4096-QAM (Wi-Fi 7);

[0112] Channel bandwidth: 80MHz, 160MHz (Wi-Fi 6); 320MHz (Wi-Fi 7).

[0113] Reward function design and calculation:

[0114]

[0115] With latency, throughput, and packet loss rate of low-latency services as the core optimization objectives, the parameters are substituted into Formula 3 to calculate the reward function. In this embodiment, the weighting coefficients are set as follows:

[0116] Low latency services The largest value is selected to prioritize latency metrics.

[0117] For example, under a certain decision-making action, .

[0118] The higher the value, the better the current decision action, and the model will strengthen the weight of that decision.

[0119] Online interactive training:

[0120] The model interacts with the network environment in 10ms increments. Each time a rate decision is made, a reward function value is collected, and the neural network weights are updated via backpropagation. Initially, the model makes random decisions. After 1000 training interactions, the model converges and outputs the optimal rate decision.

[0121] Step S4: Optimal matching decision for dedicated rates for low-latency services

[0122] The core of this step is to combine the model output with the priority of low-latency services to select the optimal rate configuration, ensuring that service transmission requirements match the network environment. Specific execution details are as follows:

[0123] Preliminary screening of decision results:

[0124] The rate decision model outputs three candidate rate configuration schemes based on the current state input. Each scheme includes dual-mode, transmission rate, modulation method, and channel bandwidth.

[0125] Low-latency service priority constraint filtering:

[0126] From the three candidate schemes, the scheme that meets the latency constraints of low-latency services is selected first: latency ≤5ms in Wi-Fi7 mode and latency ≤10ms in Wi-Fi6 mode; if multiple schemes meet the latency constraints, the scheme with the highest speed is selected.

[0127] The final rate matching scheme was determined as follows:

[0128] In this embodiment, the comprehensive evaluation value of the network environment is E=0.5775, the low latency service of industrial control is P=0.86, and the optimal solution output by the model is: Wi-Fi7 mode + 24Gbps rate + 4096-QAM modulation + 320MHz channel bandwidth. This solution can meet the 1ms latency requirement of industrial control services while ensuring transmission stability.

[0129] Step S5: Wi-Fi 6 / 7 Dual-Mode Low-Latency Smooth Rate Switching Execution

[0130] The core of this step is to avoid service interruptions and latency spikes caused by rate switching through a smooth switching mechanism. The specific execution details are as follows:

[0131] Preparations before switching:

[0132] The router locks the transmission context of low-latency service flows, including data frame sequence numbers, transmission buffers, and terminal connection information, to ensure that service data is not lost during the handover process.

[0133] Segmented buffering and frame scheduling optimization:

[0134] A service flow segmentation caching mechanism is adopted to divide low-latency service data into 10ms long data segments. During the switching period, low-latency service data frames are scheduled first, while ordinary service data frames enter the waiting queue. At the same time, the frame scheduling order is optimized to avoid data frame collisions and retransmissions.

[0135] Dual-mode switching execution:

[0136] According to the optimal rate matching scheme, the Wi-Fi 6 / 7 mode switching is performed. During the switching process, the protocol stack automatically completes frame structure adaptation, modulation and demodulation mode switching, and channel access mechanism coordination, without requiring the terminal to re-authenticate.

[0137] Switching to delay control:

[0138] In this embodiment, the additional latency of dual-mode switching is 35ms, which is lower than the preset threshold of 50ms. During the switching process, the latency of industrial control services is stable at 4ms, with no packet loss or lag, meeting the requirements for low-latency service transmission.

[0139] Step S6: Post-Switch Performance Acquisition and Model Feedback Iterative Optimization

[0140] The core of this step is to build a closed-loop control system, continuously optimize the model through performance feedback, and improve long-term adaptive capability. The specific execution details are as follows:

[0141] Real-time collection of performance parameters:

[0142] After the switchover is completed, the router collects network performance parameters every 1 second, including throughput, service latency, packet loss rate, and rate stability. In this embodiment, the collection results are: throughput 23.8Gbps, latency 4ms, packet loss rate 0.005%, and rate stability improved by 35%.

[0143] The reward function value is recalculated:

[0144] Substitute the collected performance parameters into Formula 3 and recalculate the reward function value. =0.482, an improvement of 0.004 compared to before the switch, indicating that the current decision-making scheme is effective.

[0145] Online iterative optimization of the model:

[0146] Adjust the weights of the model's neural network based on the reward function value. The higher the value, the smaller the weight adjustment range; if If the value falls below the threshold, the model will increase the weight adjustment range to optimize the next decision. In this embodiment, the model completes one iteration every 1 second, continuously adapting to changes in the network environment.

[0147] Long-term performance optimization:

[0148] Through continuous iteration, the model can autonomously learn the rate decision rules of different network environments and different low-latency services. Even in environments with multiple concurrent devices and strong interference, it can still maintain stable transmission of low-latency services, and the rate stability is improved by more than 30% compared with traditional routers.

[0149] Example effect verification:

[0150] This embodiment compares the method of the present invention with the traditional fixed threshold rate switching method. The test scenario is a home environment with 200 devices accessing the system concurrently, including three types of low-latency services: VR / AR, cloud gaming, and industrial control. The test results are as follows:

[0151]

[0152] Test results show that the method of the present invention is significantly superior to the traditional method in terms of low-latency service transmission, smooth rate switching, dual-mode compatibility, and environmental adaptability, and fully meets the transmission requirements of low-latency services.

[0153] Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for low latency traffic oriented router rate switching, the method comprising: Includes the following steps: Step S1: The router collects multi-dimensional characteristic parameters of the wireless transmission environment in real time at fixed intervals. The multi-dimensional characteristic parameters include Received Signal Strength Indicator (RSSI), Signal-to-Noise Ratio (SNR), Channel Interference Power, Network Device Access Load, and Channel Bandwidth Utilization. The collected raw characteristic parameters are linearly normalized to obtain standardized network environment parameters. Step S2: The router performs deep analysis on the service flow of the access terminal and extracts three major characteristic parameters: service latency sensitivity, bandwidth requirement, and packet loss tolerance; based on the analysis results, it identifies the low-latency service type, assigns a dedicated priority label to the low-latency service, and distinguishes between low-latency services and ordinary data services. Step S3: Construct a rate decision model based on deep reinforcement learning, using standardized network environment parameters and low-latency service priority labels as the model's state inputs, and Wi-Fi 6 / 7 dual-mode selection, transmission rate level, modulation method, and channel bandwidth as the model's executable actions; design a reward function with network throughput, service latency, and packet loss rate as optimization objectives, and optimize the model's decision weights through online interactive training. Step S4: The rate decision model inputs and outputs decision results based on the current state, and combines them with the priority constraints of low-latency services to select the Wi-Fi 6 / 7 dual-mode, transmission rate, modulation method and channel bandwidth that are suitable for the current network environment, forming a dedicated rate matching scheme for low-latency services. Step S5: Based on the rate matching scheme, a smooth switching mechanism with segmented buffering and frame scheduling optimization is adopted to perform rate switching between Wi-Fi 6 and Wi-Fi 7 modes; during the switching process, the transmission context of low-latency service flows is preserved, and the additional delay of dual-mode switching is controlled to not exceed a preset threshold. Step S6: Collect network performance parameters after rate switching in real time. The network performance parameters include throughput, service latency, packet loss rate, and rate stability. Feed the performance parameters back to the reinforcement learning rate decision model, update the model's decision parameters and weight coefficients, and complete the online iterative optimization of the model.

2. The router rate switching method for low-latency services according to claim 1, characterized in that: In step S1, after linear normalization of the multi-dimensional feature parameters, a network environment comprehensive evaluation value is calculated through weighted fusion The calculation formula is: ； wherein, is a weight coefficient and satisfies ; RSSI max router maximum received signal strength, SNR max is a router maximum signal-to-noise ratio, I max is a channel maximum allowed interference power, L max is a router maximum supported device access load.

3. The method of claim 1, wherein the method further comprises: In step S2, the priority weight P of low-latency services is quantified based on service characteristic parameters, and the calculation formula is as follows: ； wherein, is a weight coefficient ; satisfies T is a service delay sensitivity, B is a service bandwidth requirement, D is a service packet tolerance, and the value range of the three is [0, 1].

4. The router rate switching method for low-latency services according to claim 1, characterized in that: In step S3, the reward function R of the reinforcement learning rate decision model is calculated as follows: ； in, The weighting coefficients are and satisfy the following conditions: Low-latency business scenarios Value greater than and Th represents the standardized network throughput, Td represents the real-time service latency, and Pl represents the real-time service packet loss rate.

5. A router rate switching method for low-latency services according to claim 1, characterized in that: In step S1, the fixed period for collecting multi-dimensional feature parameters is 10ms. During the collection process, the parameters of three consecutive periods are subjected to moving average filtering to remove abnormal data.

6. A router rate switching method for low-latency services according to claim 1, characterized in that: In step S2, the low-latency services include VR / AR transmission services, cloud gaming services, and industrial wireless control services; the deep analysis of service flows adopts a combination of port identification and payload feature matching, with an identification accuracy of no less than 99%.

7. A router rate switching method for low-latency services according to claim 1, characterized in that: In step S4, when selecting the rate matching scheme, the latency constraints of low-latency services are prioritized, and then the transmission rate and channel bandwidth are matched. In Wi-Fi 7 mode, 320MHz channel bandwidth and 4096-QAM modulation method are preferred.

8. A router rate switching method for low-latency services according to claim 1, characterized in that: In step S5, the preset threshold for the additional delay of dual-mode switching is 50ms; the smooth switching mechanism adopts service flow segmentation caching, and during the switching period, low-latency service data frames are scheduled first, while ordinary data service frames enter the waiting queue.

9. A router rate switching method for low-latency services according to claim 1, characterized in that: In step S6, the model feedback iteration period is 1 second. In each iteration, the neural network weights of the model are adjusted according to the reward function value. The higher the reward function value, the smaller the adjustment range of the model decision parameters.

Images