Congestion control system and control method for jointly adjusting sending rate based on self intelligent queue-time delay in wireless local area network

Abstract

The invention provides a congestion control system and a control method for jointly adjusting a sending rate based on an intelligent queue-time delay under a wireless local area network, relates to the technical field of congestion control of the wireless local area network, and aims to solve the problems that a traditional congestion control algorithm depends on single signal sensing, is lagged in response, cannot adapt to bottleneck dynamic switching of the wireless local area network, and is low in cost. And the transmission delay and the link throughput are difficult to balance. According to the method, after each round of data transmission, a network card driving queue residual length RQ is obtained, a data packet round trip time RTT is monitored, a congestion scene is cooperatively judged, a congestion window cwnd is dynamically adjusted to optimize the transmission rate, and corresponding formulas or strategies are adopted for adjustment according to scenes. According to the method, the problems of one-sided sensing and response lag of a traditional congestion control algorithm can be solved, a scene is judged through RQ and RTT dual-sensing cooperation, cwnd is dynamically adjusted, time delay is reduced when the network is good, throughput is guaranteed when the network is poor, and the performance is superior to Cubic and BBR.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of wireless local area network congestion control, and particularly relates to a congestion control system and method for adjusting sending rate based on body-aware intelligent queue-delay combination under a wireless local area network. BACKGROUND

[0002] The existing traditional end-to-end congestion control methods include BBR, Cubic and Vegas, wherein, (1) The core idea of Cubic algorithm is to dynamically adjust the size of congestion window of TCP connection according to network congestion. Cubic algorithm uses a time-based cubic curve function to calculate the size of congestion window. The congestion window grows at a slower speed when the network is not congested, but quickly decreases when encountering congestion.

[0003] (2) The core idea of BBR algorithm is to actively probe the bottleneck bandwidth in the network. It identifies the bottleneck link in the network by sending data and measuring the transmission rate of data packets and round-trip time (RTT). BBR tries to find the maximum bandwidth without causing network congestion. It does not use congestion window to control the sending rate, but dynamically adjusts the data sending rate according to the bandwidth estimation to gradually occupy the network bottleneck bandwidth.

[0004] (3) The core idea of Vegas algorithm is to detect network congestion by monitoring the change of network round-trip time (RTT). It compares the actual RTT with the baseline RTT (RTT under normal conditions) to determine whether the network is congested. When the actual RTT is significantly higher than the baseline RTT, Vegas algorithm considers that congestion may occur, and uses the congestion window growth rate to control the data transmission rate.

[0005] However, the above congestion control methods have the following problems: (1) Cubic algorithm is the default congestion control algorithm of Linux system at present. Its core idea is to dynamically adjust the size of the congestion window of TCP connection according to the network congestion situation. Cubic algorithm uses a time-based cubic curve function to calculate the size of the congestion window. Cubic algorithm can fully utilize the link capacity to fill the bottleneck queue as much as possible before packet loss occurs, so it can usually achieve higher throughput and link utilization. However, when the bandwidth utilization is high, it still increases the congestion window and thus increases the packet loss rate and exacerbates network jitter. With the development of technology and the decline of hardware cost, when the network is just congested, packet loss does not occur, but the data is stored in the buffer. When the buffer overflow caused by packet loss is detected, the buffer has accumulated a large amount of data, which means that the network congestion has been very serious. This leads to a certain lag in the sender's judgment of network congestion, and lacks the ability of "multi-dimensional real-time perception" of embodied intelligence, that is, it only relies on the single feedback signal of "packet loss", cannot actively capture the dynamic changes of the first hop queue accumulation, can only passively wait for the packet loss signal after the congestion is exacerbated, and the perception of the link is long and the response is not timely. In addition, the buffer inflation problem caused by this approach will cause large queuing delay, resulting in excessive queuing delay.

[0006] (2) BBR dynamically adjusts the data transmission rate according to the bandwidth and round-trip delay of the network to avoid network congestion and maximize data transmission efficiency. However, due to the rapid changes of wireless links, periodic probing may lead to inaccurate bandwidth measurement and performance degradation. In addition, this method takes the instantaneous sending rate as its main control object, although it also controls cwnd, but Cardwell believes that its control of cwnd is only an auxiliary role. More importantly, BBR does not embody the "environmental dynamic response" feature of embodied intelligence: the periodic probing mechanism cannot adjust the perception frequency with the real-time fluctuations of wireless links, which means that the "perception rhythm" of the agent does not match the "environmental change rhythm", leading to bandwidth measurement deviation and inability to adapt to the dynamic environment of wireless local area networks.

[0007] (3) Delay-based congestion control algorithms, which dynamically adjust the data transmission rate by monitoring the changes in round-trip time (RTT). They can detect signs of network congestion in advance before packets are lost due to buffer overflow, and thus can more effectively maintain lower end-to-end delay, which is crucial for real-time media transmission. However, this delay-based regulation method may be too conservative in adjusting the rate due to excessive sensitivity to congestion, resulting in suboptimal utilization of network links. On the other hand, in a network mixed with multiple traffic types, these delay-based algorithms controlled flows may be at a disadvantage and even starved when competing with those based on packet loss mechanisms. Essentially, such algorithms lack the embodied intelligence of "multi-dimensional perception coordination" logic, i.e., relying only on a single environmental signal of "delay", and cannot balance local and global link states, with a one-sided perception dimension, making it difficult to balance "link utilization-competitive advantage" in complex traffic scenarios.

[0008] The traditional algorithms have the common problem of lack of embodied intelligence: they all use the "single signal perception-passive behavior adjustment" mode, do not build an intelligent agent with "congestion control module" as the core, and cannot realize the closed loop of "perception (multi-dimensional environmental interaction)-decision (quantitative judgment of congestion type)-action (precise adjustment of sending rate)". The core feature of the wireless local area network environment is "dynamic switching of bottlenecks (first hop / non-first hop), fast link state fluctuation", and the passive perception and lag response of traditional algorithms are difficult to adapt to the environment requirements. SUMMARY

[0009] The present application provides a wireless local area network under the congestion control system and control method for adjusting the sending rate based on the embodied intelligent queue-delay joint adjustment, which takes the congestion control module as the core of embodied intelligence, locally perceives the remaining length of the card driver queue (RQ) to obtain the local link load, globally monitors the round-trip time (RTT) of data packets to capture the global congestion state, and dynamically adjusts the congestion window (cwnd) based on the dual-dimensional coordination of first hop / non-first hop / no congestion scenarios, solving the problem that the traditional congestion control algorithm in the prior art relies on single signal perception, response lag, cannot adapt to the dynamic switching of wireless local area network bottlenecks, and is difficult to balance the transmission delay and link throughput.

[0010] A congestion control system for adjusting the sending rate based on the embodied intelligent queue-delay joint adjustment in a wireless local area network, comprising an embodied intelligent control module, a local perception unit, a global perception unit and a rate adjustment unit, the local perception unit and the global perception unit are connected with the embodied intelligent control module, the embodied intelligent control module is connected with the rate adjustment unit, wherein, The body intelligence control module is used for cooperative judgment based on the residual queue length RQ and the round trip time RTT, and sends a judgment instruction to the rate adjustment unit; The local perception unit is used for collecting the residual queue length RQ of the network card drive queue in the wireless local area network, so as to obtain the load state of the local link and upload to the body intelligence control module; The global perception unit is used for monitoring the round trip time RTT of the data packet, so as to obtain the congestion state of the global link and upload to the body intelligence control module; The rate adjustment unit is used for dynamically adjusting the congestion window cwnd based on the cooperative judgment result of the residual queue length RQ and the round trip time RTT, so as to adapt to the first hop or non-first hop bottleneck scene of the wireless local area network.

[0011] Further, the network card drive queue collected by the local perception unit is located between the TCP / IP protocol and the network interface controller NIC, and is a first-in first-out FIFO ring buffer; The residual queue length RQ is the number of data packets in the ring buffer which are not sent by the network interface controller after each round of data transmission.

[0012] Further, the round trip time RTT monitored by the global perception unit includes the current average round trip time RTT average And the minimum round trip time RTT min The minimum round trip time RTT min is the minimum RTT value measured in a period of time, reflecting the inherent transmission time of the network in the ideal state.

[0013] A congestion control method for adjusting the sending rate based on the body intelligence queue-delay joint of the wireless local area network, based on the above-mentioned congestion control system for adjusting the sending rate based on the body intelligence queue-delay joint of the wireless local area network, comprising the following steps: S1: After each round of data transmission, the residual queue length RQ of the network card drive queue in the wireless local area network is obtained, and the round trip time RTT of the data packet is monitored; S2, based on the cooperative judgment of the residual queue length RQ and the round trip time RTT, the congestion scene of the current wireless local area network is determined, and the congestion scene includes the first hop bottleneck and the non-first hop bottleneck or no congestion; S3, according to the congestion scene, dynamically adjusting the congestion window cwnd to adapt to the current link state, realizing the optimization of the sending rate.

[0014] Further, in S2, when the residual queue length RQ is greater than the preset threshold δ1, the adjustment mode of S3 includes: If cwnd≥(RQ- ), then cwnd is reduced as follows to make the residual queue length decrease: ; If cwnd < (RQ- , the congestion window cwnd is halved.

[0015] Further, in S2, when the residual queue length RQ is equal to 0, the judgment of the congestion scenario and the adjustment mode of S3 include: If the current average round-trip time RTT average is greater than the minimum round-trip time RTT min by more than a preset threshold , the congestion window cwnd is adjusted according to the following formula:

[0016] wherein, represents the corresponding accumulated queue length calculated according to the part of RTT exceeding the threshold; If the current average round-trip time RTT average is not greater than the minimum round-trip time RTT min by more than a preset threshold , the congestion window cwnd is increased to probe the available bandwidth.

[0017] Further, in S2, when the residual queue length RQ is equal to 0 and the current average round-trip time RTT average is not greater than the minimum round-trip time RTT min by more than a preset threshold , if it is detected that the residual queue length RQ is 0 for n consecutive rounds, the congestion window cwnd is increased according to the following formula:

[0018] wherein, δ2 is a preset window increasing step.

[0019] Further, in S2, when the residual queue length RQ is greater than 0 and less than a preset threshold δ1, the judgment of the congestion scenario and the adjustment mode of S3 include: If it is detected that the global link is not congested according to the round-trip time RTT, the current congestion window cwnd is kept unchanged; If it is detected that there is a non-first-hop bottleneck in the global link according to the round-trip time RTT, the cwnd is immediately reduced according to the following formula:

[0020] A storage medium, the storage medium has a computer program stored thereon, the computer program is executed by a processor to implement the above-mentioned wireless local area network under the body intelligence queue-delay joint adjustment of sending rate based on the congestion control method of the queue.

[0021] A computer device comprises a memory, a processor and a computer program stored on the memory and executable on the processor, the processor executes the program to implement the above-mentioned wireless local area network-based queue-delay joint adjustment of sending rate based on embodied intelligence congestion control method.

[0022] Compared with the prior art, the above technical scheme has the following remarkable beneficial effects: The application introduces the "environment-agent interaction" paradigm with embodied intelligence, relies on the real-time perception of the local network card driving residual queue length (RQ) and the comprehensive monitoring of global round-trip time (RTT), realizes the full coverage of all bottleneck scenarios of the first hop / non-first hop of the wireless local area network, effectively solves the problem of the one-sidedness of the perception dimension of the traditional algorithm, the real-time interaction mechanism with each round of data transmission as the period greatly shortens the perception-response interval compared with the traditional scheme relying on ACK end-to-end feedback, and improves the adaptation speed to the wireless link fluctuations; the congestion window (cwnd) is dynamically adjusted through the quantitative threshold and the scenario-based decision logic, the transmission delay is significantly reduced (reduced by 14% compared with Cubic and 11.3% compared with BBR) when the network state is good, the throughput is stably guaranteed when the network state is poor, and the dynamic balance of the delay and the throughput is realized; the quantitative design of the decision logic of the application avoids the extreme adjustment of the traditional algorithm, improves the transmission stability, and the performance is better than that of the mainstream algorithms such as Cubic and BBR through the engineering landing verification of the Linux kernel and the actual router, and the application scene is widely applicable, thereby providing a more efficient, stable and adaptable technical scheme for the congestion control of the wireless local area network. BRIEF DESCRIPTION OF DRAWINGS

[0023] Figure 1 A method flowchart of a wireless local area network-based queue-delay joint adjustment of sending rate based on embodied intelligence congestion control method of the application; Figure 2 A working principle diagram of a network card driving queue; Figure 3 A comparison diagram of the data packet delay of the application and Cubic and BBR, wherein, Figure 3 (a) is a comparison diagram of the data packet delay of the application and Cubic; Figure 3 (b) is a comparison diagram of the data packet delay of the application and BBR; Figure 4 A delay and throughput scatter plot of several different congestion control algorithms; Figure 5 The throughput of several different congestion control algorithms. DETAILED DESCRIPTION

[0024] With reference to the drawings of the embodiments of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described, obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative work belong to the scope of protection of the present application.

[0025] With reference to Figure 1 As shown in the figure, a congestion control system for adjusting sending rate based on body-aware intelligent queue-delay joint adjustment under a wireless local area network, comprising a body-aware intelligent control module, a local perception unit, a global perception unit and a rate adjustment unit, the local perception unit and the global perception unit are connected with the body-aware intelligent control module, and the body-aware intelligent control module is connected with the rate adjustment unit, wherein, The body-aware intelligent control module is used for cooperative judgment based on the residual queue length RQ and the round-trip delay RTT, and sends a judgment instruction to the rate adjustment unit. The local perception unit is used for collecting the residual queue length RQ of the network card drive queue in the wireless local area network to obtain the load state of the local link, and uploading to the body-aware intelligent control module. The global perception unit is used for monitoring the round-trip delay RTT of the data packet to obtain the congestion state of the global link, and uploading to the body-aware intelligent control module. The rate adjustment unit is used for dynamically adjusting the congestion window cwnd based on the cooperative judgment result of the residual queue length RQ and the round-trip delay RTT to adapt to the first hop or non-first hop bottleneck scene of the wireless local area network.

[0026] Specifically, the present application takes the congestion control system as the body-aware intelligent agent under the wireless local area network scene, takes the network card drive queue (RQ) as the local environment perception interface, takes the round-trip delay (RTT) as the global environment perception interface, and through real-time interaction with the network environment (collecting RQ after each round of sending and dynamically monitoring RTT), a perception-decision-action closed loop is constructed: based on the double-dimensional perception result, the congestion type (first hop / non-first hop) is judged, the congestion window (cwnd) is dynamically adjusted to realize the sending rate adaptation, and the problems of one-sided perception and response lag of the traditional algorithm are solved.

[0027] Further, the network card drive queue collected by the local perception unit is located between the TCP / IP protocol and the network interface controller NIC, and is a first-in-first-out FIFO ring buffer. The residual queue length RQ is the number of data packets in the ring buffer which are not sent by the network interface controller after each round of data sending.

[0028] Further, the round-trip delay RTT monitored by the global perception unit includes the current average round-trip delay RTT averageand minimum round-trip time RTT min , the minimum round-trip time RTT min is the minimum RTT value measured in a period of time, reflecting the inherent transmission time of the network in an ideal state.

[0029] A congestion control method for adjusting the sending rate based on the embodiment of the intelligent queue-delay joint under the wireless local area network, based on the above-mentioned congestion control system for adjusting the sending rate based on the embodiment of the intelligent queue-delay joint under the wireless local area network, characterized in that it comprises the following steps: S1: After each round of data transmission is completed, the remaining queue length RQ of the network card driver queue in the wireless local area network is obtained, and the round-trip time RTT of the data packet is monitored; S2, based on the cooperative judgment of the remaining queue length RQ and the round-trip time RTT, determine the congestion scenario of the current wireless local area network, the congestion scenario includes the first hop bottleneck and the non-first hop bottleneck or no congestion; S3, according to the congestion scenario, dynamically adjust the congestion window cwnd to adapt to the current link state, and realize the optimization of the sending rate.

[0030] Specifically, when the data transmission application of the wireless terminal faces the rapid change of the link congestion state, the adjustment of the sending rate may not be timely, which may cause the quality of experience to decrease. In order to alleviate this problem, the application designs a congestion control method for adjusting the sending rate based on the joint of the driver queue length and the delay, which mainly adjusts the cwnd according to the remaining queue length after each round of wireless network card transmission. When the queue length is large, the cwnd will be quickly reduced to alleviate the congestion, and when the queue length is small, the cwnd will gradually increase to improve the link utilization. Since the local congestion information is obtained at the sending side, the congestion feedback path is much shorter than the end-to-end ACK feedback, and the application can track the real-time change of the wireless link capacity more quickly and accurately.

[0031] The core of the scheme is the dual-dimension perception-dynamic decision-precise action logic of the embodiment of the intelligent agent: through the cooperation of local perception (RQ) and global perception (RTT), the specific congestion scenario of the network environment is judged, and then the adaptive cwnd adjustment action is outputted, so as to ensure that the behavior of the intelligent agent is highly matched with the environment state.

[0032] Although the bottleneck is usually the first hop, it can also occur on the link where the router forwards to the receiver. Therefore, the method of relying solely on the driver queue is not accurate and comprehensive, so the method proposes a regulation strategy that integrates RTT. RTT refers to the round-trip time of each data packet, which is an important indicator in network communication. Although the sender cannot accurately determine the link congestion state according to the driver queue length information when the accumulation does not occur at the first hop, the change in the data packet queuing delay caused by the accumulation can be captured by the sender according to the change in RTT. When the system detects that the RTT is too large, it means that a large number of data packets are accumulated at a node in the link, thus causing a large queuing delay. Therefore, the congestion window needs to be reduced immediately to alleviate the link congestion.

[0033] The network card driver queue (RQ) is located between the TCP / IP protocol and the network interface controller (NIC), and is usually implemented as a first-in-first-out (FIFO) ring buffer. When the host transmits a data packet, the data packet is first injected into the ring buffer, and then the NIC reads the data packet and sends it out, and the ring buffer is emptied after the data packet is successfully sent. The specific working process is as shown in Figure 2 .

[0034] RTT (Round-Trip Time) refers to the total time for a data packet to be sent from the sender, transmitted through the network to the receiver, and then returned to the sender with an acknowledgment (ACK) from the receiver. When the number of data packets in the network exceeds the processing capacity of the link, the data packets will wait in the queue at the congestion point for transmission, which will cause a large amount of queuing delay of the data packets, thereby causing the RTT to become large.

[0035] RTT min in the following refers to the minimum RTT value measured in a period of time. The data packet corresponding to RTT min usually has almost no queuing delay in the link, and it reflects the inherent transmission time of the network in an ideal state. Therefore, RTT min is a relatively stable reference value that can be used to assist in evaluating the delay performance of the network. That is, by comparing the current RTT with the minimum RTT, the TCP protocol can more accurately determine whether the network is in a congested state, and if the current RTT is significantly higher than the minimum RTT, it indicates that there may be congestion in the network, and the sending rate needs to be adjusted.

[0036] In the embodiment of the embodied intelligent framework, RTT is a global environment perception unit: when the local perception (RQ) cannot cover the non-first-hop bottleneck, the change in RTT can capture the cumulative effect of the agent's sending behavior in the global link environment, i.e. the accumulation of data packets at non-first-hop nodes will directly cause the RTT to increase. By comparing the current RTT with RTT minBy using the ratio of [value], the intelligent agent can quantify the congestion level of the global environment, making up for the limitations of single local perception and achieving full-scene environment coverage perception of both local and global scenarios.

[0037] Specifically, when entering the congestion avoidance phase, this scheme mainly adjusts the congestion window based on the following three situations: Long Remaining Queue (First Hop Bottleneck): When the embodied agent detects through its Local Sense Unit (RQ) that the number of data packets sent exceeds the Delay-Bandwidth Product (BDP, i.e., the product of the highest throughput when the pipe is full and the lowest RTT when the pipe is not full), it indicates that the link is congested and the sending rate should be reduced. Therefore, this algorithm sets a remaining queue threshold. When the remaining queue length (RQ) is detected to be greater than this threshold after each round of network card transmission, the sender will reduce cwnd accordingly based on the difference between RQ and the threshold to reduce the remaining queue length:

[0038] If cwnd is less than (RQ- (This can be solved by simply halving the congestion window.)

[0039] Empty Remaining Queue (Global Congestion / Resource Idle Scenario): When the embodied agent detects a remaining queue length of 0 via the Local Awareness Unit (RQ), it indicates that congestion may not have occurred or the bottleneck link is not in the first hop. In this case, the Global Awareness Unit needs to be activated to compare the current RTT. average The ratio to the minimum RTT. If the current RTT average The ratio of the minimum RTT exceeds the threshold. If the RTT is high, it indicates the existence of a bottleneck beyond the first hop. In this case, it's necessary to calculate the number of packets piling up on the bottleneck link (the packets causing the increased RTT) based on the RTT value and throughput (converted into packet count), and adjust the congestion window accordingly. If the RTT... average If the ratio to the minimum RTT does not exceed the threshold, it indicates that congestion has not yet occurred, and available bandwidth can be further probed. Therefore, when the remaining queue length is equal to 0, there are two cases: (1) When RTT average If the ratio of the minimum RTT to the minimum RTT exceeds the threshold, then cwnd needs to be adjusted according to the following formula:

[0040] in This represents the length of the stacking queue calculated based on the portion of the RTT that exceeds the threshold.

[0041] (2) When RTT average If the ratio of the minimum RTT to the minimum RTT does not exceed the threshold, the congestion window needs to be appropriately increased to fully utilize the link. This algorithm uses a window increment step. This is used to detect available capacity. However, when the dequeue rate is much greater than the inqueue rate, a fixed increment will cause cwnd to grow too slowly. Therefore, if the remaining queue length is detected to be 0 for n consecutive rounds, the increment of cwnd can be set to... Adjust cwnd according to the following formula:

[0042] Appropriate Remaining Queue (Local Balancing + Global Verification Scenario): When the embodied agent detects through the Local Perception Unit (RQ) that the remaining queue length is greater than 0 and less than a threshold... At this time, the sending end judges that the first hop link is fully utilized and there will be no long queuing delay. However, at this time, the data packets may accumulate on other nodes. Therefore, in this case, it is still necessary to judge whether data accumulation occurs on other nodes of the link based on the current RTT. Therefore, the congestion window will be adjusted according to the following two situations: (1) When it is detected that there is no congestion on other links according to the RTT, the congestion window (cwnd) can be kept unchanged; (2) When it is detected that there is a bottleneck on the external link according to the RTT, the congestion window (cwnd) will be reduced immediately.

[0043] A storage medium storing a computer program, which, when executed by a processor, implements the congestion control method described above for adjusting the transmission rate based on embodied intelligent queues and delay in a wireless local area network.

[0044] Specifically, the storage medium of this invention stores a corresponding computer program, enabling the program to fully leverage the core technological advantages of intelligent queue-latency joint adjustment when executed by a processor. Relying on the local network card to drive real-time load perception of the remaining queue length (RQ) and global round-trip time (RTT) congestion state monitoring, it achieves comprehensive coverage of all bottleneck scenarios in both first-hop and non-first-hop wireless LANs, effectively solving the problem that traditional congestion control algorithms rely solely on single signal perception and have limited adaptability. The real-time interaction mechanism, with each round of data transmission as a cycle during program execution, significantly shortens the interval between perception and adjustment. It can quickly respond to the dynamic fluctuations of wireless links, avoiding the congestion judgment lag problem caused by the reliance on end-to-end feedback in traditional algorithms. By dynamically optimizing the congestion window (cwnd) through quantified thresholds and scenario-based decision logic, it can effectively reduce transmission latency when the network condition is good and stably guarantee link throughput when the network condition is poor, achieving a dynamic balance between latency and throughput. Moreover, after engineering verification in the Linux kernel and actual routers, its performance is better than mainstream algorithms such as Cubic and BBR. It can be quickly deployed without additional complex configuration, providing efficient, stable and highly adaptable technical support for wireless LAN congestion control.

[0045] A computer device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the congestion control method for adjusting the transmission rate based on embodied intelligent queues and delays in a wireless local area network as described above.

[0046] Specifically, the computer device of this invention stores the corresponding computer program in the memory. Relying on the collaborative work of the processor and the memory, it provides a reliable hardware carrier for the implementation of congestion control schemes. When the processor executes the program, it can efficiently realize the core logic of intelligent queue-latency joint adjustment: it can not only complete the local load perception of the remaining queue length (RQ) driven by the network card after each round of data transmission in real time, but also accurately monitor the round-trip time (RTT) of data packets to capture the global congestion status. Then, through collaborative judgment, it can identify the first-hop bottleneck, non-first-hop bottleneck, and congestion-free scenarios of the wireless LAN, and dynamically adjust the congestion window (…). (cwnd) adapts to link conditions; during the execution of the program, this device effectively solves the problems of traditional congestion control algorithms relying on a single signal perception and having a delayed response. It avoids the high latency caused by the buffer expansion of the Cubic algorithm and the poor adaptability of the periodic probe of the BBR algorithm. It can reduce transmission latency when the network condition is good and can stably guarantee throughput when the network condition is poor, achieving a dynamic balance between latency and throughput. Moreover, after engineering verification in the Linux kernel and actual routers, its performance is better than mainstream algorithms such as Cubic and BBR. It can be integrated into the wireless LAN environment without additional complex configuration.

[0047] This invention was deployed in the Linux kernel and tested in different network environments, referring to... Figure 3 As shown, the test results demonstrate that the solution proposed in this invention is superior to the current default congestion control algorithm Cubic in Linux systems, which is based on Google's BBR.

[0048] Since there is no difference between 90Mbps and 88Mbps for users when network conditions are good, this solution prioritizes reducing latency when network conditions are favorable. It accurately determines congestion-free states through local and global dual-sensoring, avoiding queuing delays caused by blindly increasing window size in traditional algorithms. Figure 2 The graph shows the latency results of this scheme compared to Cubic and BBR in a single test. It is clear from the graph that the latency per data packet is significantly lower when using this invention compared to the Cubic and BBR algorithms. We conducted 10 experiments on the data packet latency produced by each of the three schemes and calculated the average RTT. Under relatively good network conditions, the average data packet latency of this scheme is reduced by 14% and 11.3% compared to Cubic and BBR, respectively. Figure 4The diagram shows scatter plots of delay versus throughput at the P95 quantile for several different congestion control algorithms under this network condition. It can be observed that, combined, the performance of this invention outperforms Cubic, BBR, and other congestion control algorithms. Essentially, the embodied intelligent perception-decision-action closed loop achieves a dynamic balance between delay and throughput.

[0049] However, when network conditions are poor, user experience is often directly related to network throughput. Figure 5 The diagram illustrates a scenario where chload=900, indicating extremely poor network conditions and heavy router activity. Throughput graphs for three different congestion control algorithms show that the transmission rate using this scheme is significantly higher than that of Cubic and BBR.

[0050] The advantages of this invention stem from the "multi-dimensional perception-dynamic interaction" characteristics of embodied intelligence, specifically manifested in three aspects: Embodied intelligence enhances adaptability across all scenarios: Compared to traditional algorithms with single perception (packet loss, single latency), this solution uses dual perception of RQ (local) + RTT (global) to cover all bottleneck scenarios in the first hop / non-first hop of the wireless LAN, reducing latency when the network is good and maintaining throughput when the network is poor. Embodied intelligence enhances response speed through real-time interaction: Environmental data (RQ, RTT) is collected in cycles of each round of transmission. Compared with the end-to-end lag feedback that relies on ACK in traditional algorithms, embodied intelligence has a shorter perception-action interval and can quickly adjust when the wireless link fluctuates. Embodied intelligence-based quantitative decision-making enhances stability: By quantifying environmental states through thresholds such as δ1, α, and δ2, perceived signals are transformed into precise actions, avoiding the conservative / aggressive extreme adjustments of traditional algorithms and achieving a performance balance.

[0051] In summary, the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit it. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A congestion control system for wireless local area networks based on embodied intelligent queues and time delay joint adjustment of transmission rate, characterized in that, It includes an embodied intelligent control module, a local sensing unit, a global sensing unit, and a rate adjustment unit. The local sensing unit and the global sensing unit are both connected to the embodied intelligent control module, which is connected to the rate adjustment unit. The embodied intelligent control module is used to make collaborative judgments based on the remaining queue length RQ and round-trip time RTT, and send the judgment instructions to the rate adjustment unit; The local sensing unit is used to collect the remaining queue length RQ of the network card driver queue in the wireless local area network to obtain the load status of the local link and upload it to the embedded intelligent control module. The global awareness unit is used to monitor the round-trip time (RTT) of data packets to obtain the congestion status of the global link and upload it to the embodied intelligent control module. The rate adjustment unit is used to dynamically adjust the congestion window (cwnd) based on the collaborative judgment results of the remaining queue length (RQ) and round-trip time (RTT) to adapt to the first-hop or non-first-hop bottleneck scenarios of the wireless local area network.

2. The congestion control system based on embodied intelligent queue-delay joint adjustment of transmission rate in a wireless local area network according to claim 1, characterized in that, The network card driver queue collected by the local sensing unit is located between the TCP / IP protocol and the network interface controller (NIC), and is a first-in-first-out (FIFO) circular buffer. The remaining queue length RQ is the number of data packets in the circular buffer that have not been sent by the network interface controller after each round of data transmission.

3. The congestion control system based on embodied intelligent queue-delay joint adjustment of transmission rate in a wireless local area network according to claim 1, characterized in that, The round-trip time (RTT) monitored by the global sensing unit includes the current average round-trip time (RTT). average and minimum round-trip time (RTT) min The minimum round-trip time (RTT) min It is the minimum RTT value measured over a period of time, reflecting the inherent transmission time of the network under ideal conditions.

4. A congestion control method for wireless local area networks based on embodied intelligent queues and delay-based joint adjustment of transmission rate, based on the congestion control system for wireless local area networks based on embodied intelligent queues and delay-based joint adjustment of transmission rate as described in claims 1-3, characterized in that, Includes the following steps: S1: After each round of data transmission is completed, obtain the remaining queue length RQ of the network card driver queue in the wireless LAN, and monitor the round-trip time (RTT) of the data packets; S2. Based on the collaborative judgment of the remaining queue length RQ and round-trip time RTT, determine the current congestion scenario of the wireless LAN. The congestion scenario includes first-hop bottleneck, non-first-hop bottleneck or no congestion. S3. Based on the congestion scenario, dynamically adjust the congestion window (cwnd) to adapt to the current link state and optimize the transmission rate.

5. The congestion control method for wireless local area networks based on embodied intelligent queues and delay joint adjustment of transmission rate according to claim 4, characterized in that, In S2, when the remaining queue length RQ is greater than the preset threshold δ1, the adjustment methods in S3 include: If cwnd≥(RQ- If the remaining queue length decreases, then cwnd will be reduced according to the following formula: ； If cwnd < (RQ- If the congestion window (cwnd) is halved, then the congestion window (cwnd) will be halved.

6. The congestion control method for wireless local area networks based on embodied intelligent queues and delay joint adjustment of transmission rate according to claim 4, characterized in that, In S2, when the remaining queue length RQ equals 0, the congestion scenario assessment and adjustment methods in S3 include: If the current average round-trip time (RTT) is average With minimum round-trip time (RTT) min The ratio exceeds the preset threshold Then adjust the congestion window cwnd according to the following formula: in, This represents the length of the backlog queue calculated based on the portion of the RTT that exceeds the threshold. If the current average round-trip time (RTT) is average With minimum round-trip time (RTT) min The ratio did not exceed the preset threshold. If so, increase the congestion window (cwnd) to detect available bandwidth.

7. The congestion control method for wireless local area networks based on embodied intelligent queues and delay joint adjustment of transmission rate according to claim 6, characterized in that, In S2, when the remaining queue length RQ is equal to 0 and the current average round-trip time RTT is... average With minimum round-trip time (RTT) min The ratio did not exceed the preset threshold. If the remaining queue length RQ is detected to be 0 for n consecutive rounds, the congestion window cwnd is increased according to the following formula: Where δ2 is the preset window increment step size.

8. The congestion control method for wireless local area networks based on embodied intelligent queues and delay joint adjustment of transmission rate according to claim 4, characterized in that, In S2, when the remaining queue length RQ is greater than 0 and less than the preset threshold δ1, the congestion scenario judgment and adjustment method in S3 include: If global link congestion is detected based on round-trip time (RTT), then the current congestion window (cwnd) remains unchanged. If a non-first-hop bottleneck is detected in the global link based on the round-trip time (RTT), then the concurrency limit (cwnd) should be reduced immediately according to the following formula: 。 9. A storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the congestion control method for adjusting the transmission rate based on embodied intelligent queues and delay in a wireless local area network as described in any one of claims 4-8.

10. A computer device, characterized in that, include: The method includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the congestion control method for adjusting the transmission rate based on embodied intelligent queues and delay in a wireless local area network as described in any one of claims 4-8.

Images