Wireless adaptive rate control method and related apparatus

Abstract

The present disclosure provides a wireless adaptive rate control method and related device, the wireless adaptive rate control method comprising: obtaining channel state information and network state information at data transmission; inputting the channel state information and the network state information into a long short-term memory neural network model to obtain a transmission rate output by the long short-term memory neural network model; and performing subsequent data transmission based on the transmission rate. The present disclosure can enable the transmission rate to be accurately adapted to the channel quality in a timely manner, thereby improving network throughput.

Description

Technical Field

[0001] This disclosure relates to the field of communication technology, specifically to a wireless adaptive rate control method and related apparatus. Background Technology

[0002] The IEEE 802.11 wireless LAN communication standard offers various transmission rates for its different communication modes. The purpose of adaptive rate control (ARPC) is to automatically select a transmission rate that adapts to the wireless channel quality, ensuring that high transmission rates are matched to good channel quality and low transmission rates to poor channel quality. Existing ARPC methods are categorized into two types based on different channel quality measurement schemes: those based on direct channel quality measurement and those based on statistical calculations of channel quality. The former, without additional feedback mechanisms, cannot guarantee a proper match between transmission rate and wireless channel quality, while the latter struggles to adapt the transmission rate to channel quality in a timely manner when the channel changes rapidly, thus consistently negatively impacting network throughput. Summary of the Invention

[0003] To address the aforementioned technical issues, this disclosure provides a wireless adaptive rate control method and related apparatus, aiming to enable the transmission rate to adapt to channel quality in a timely and accurate manner, thereby improving network throughput.

[0004] According to a first aspect of this disclosure, a wireless adaptive rate control method is provided, comprising:

[0005] Acquire channel status information and network status information during data transmission;

[0006] The channel state information and the network state information are input into the long short-term memory neural network model to obtain the transmission rate output by the long short-term memory neural network model;

[0007] Subsequent data transmission will be performed based on the aforementioned transmission rate.

[0008] Optionally, the long short-term memory neural network model is trained through the following steps:

[0009] Obtain the current sample set, wherein each sample in the current sample set has a rate label, and the rate in the rate label is adapted to the wireless channel quality after the corresponding sample collection period.

[0010] Input the training samples used for training in the current sample set into the current long short-term memory neural network model to obtain the transmission rate output by the current long short-term memory neural network model;

[0011] The accuracy of the current long short-term memory neural network model is determined based on the transmission rate and rate label corresponding to the training samples.

[0012] If the accuracy does not converge, the weights and biases of one or more gates in the current long short-term memory neural network model are adjusted, and the adjusted long short-term memory neural network model is determined as the current long short-term memory neural network model.

[0013] Return to the step of inputting the training samples used for training in the current sample set into the current long short-term memory neural network model, until the accuracy of the current long short-term memory neural network model converges.

[0014] Optionally, the current long short-term memory neural network model is further optimized through the following steps after the accuracy converges:

[0015] Input the test samples used for testing in the current sample set into the current long short-term memory neural network model to obtain the transmission rate output by the current long short-term memory neural network model;

[0016] The accuracy of the current long short-term memory neural network model is determined based on the transmission rate and rate label corresponding to the test sample.

[0017] If the accuracy is not greater than the threshold, then update the structural parameters of the current long short-term memory neural network model and determine the updated long short-term memory neural network model as the current long short-term memory neural network model, and / or, after resetting the types of information included in the sample, obtain multiple samples and determine the sample set composed of the multiple obtained samples as the current sample set;

[0018] Return to the step of inputting the training samples used for training in the current sample set into the current long short-term memory neural network model, until the accuracy of the current long short-term memory neural network model is greater than the threshold.

[0019] Optionally, the types of information included in the sample can be reset, including: re-determining one or more pieces of information from multiple candidate channel state information and multiple candidate network state information as the information included in the sample;

[0020] When the accuracy of the current long short-term memory neural network model is greater than the threshold, the information included in each sample in the current sample set is the target information. The channel state information and network state information during data transmission are obtained, including: obtaining the target information during data transmission.

[0021] Optionally, the multiple candidate channel state information includes: signal and noise plus interference ratio, signal-to-noise ratio, received signal strength, signal receiving sensitivity, transmit power, error vector magnitude, transmit gain, and receive gain;

[0022] The multiple candidate network status information includes: frame loss rate, frame error rate, number of acknowledgment characters, throughput, channel number, communication mode, and modulation method.

[0023] Optionally, the transmission rate output by the current long short-term memory neural network model is one of a plurality of preset rates;

[0024] The multiple preset rates are pre-configured according to the communication mode used for data transmission before the training samples used for training in the current sample set are input into the current long short-term memory neural network model.

[0025] Optionally, obtaining the current sample set includes: selecting a candidate sample set corresponding to the environment in which the current data transmission occurs from a plurality of candidate sample sets as the current sample set, wherein one candidate sample set in the plurality of candidate sample sets corresponds to one environment.

[0026] Optionally, obtain the current sample set, including:

[0027] Multiple original samples are convolutionally processed to obtain multiple processed samples, and the set of the multiple processed samples is determined as the current sample set;

[0028] Alternatively, a moving weighted average can be applied to multiple original samples collected within a preset time period to obtain a processed sample corresponding to the preset time period, and the set of multiple processed samples can be determined as the current sample set, wherein the samples collected later within the preset time period have a larger weight in the moving weighted average processing.

[0029] According to a second aspect of this disclosure, an electronic device is provided, comprising: a processor, a memory, and a program stored in the memory and executable on the processor, wherein the program, when executed by the processor, implements the steps of any of the wireless adaptive rate control methods described in the first aspect.

[0030] According to a third aspect of this disclosure, a computer-readable storage medium is provided, on which a computer program or instructions are stored, which, when executed by a processor, implement the steps of any of the wireless adaptive rate control methods described in the first aspect.

[0031] The beneficial effects of this disclosure are:

[0032] The wireless adaptive rate control method disclosed herein, after acquiring channel state information and network state information during data transmission, inputs the channel state information and network state information into a Long Short-Term Memory (LSTM) neural network model to obtain the transmission rate output by the LTM neural network model, and then performs subsequent data transmission based on the transmission rate. Because the gating mechanism and memory cells of the LTM neural network model can well adapt to the transitivity of network state information in determining the transmission rate and the forgetting nature of channel state information in determining the transmission rate, adaptive selection of the transmission rate can be achieved, enabling the transmission rate to be timely and accurately adapted to channel quality, thereby improving network throughput.

[0033] It should be noted that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Attached Figure Description

[0034] Figure 1 This diagram illustrates a flowchart of a wireless adaptive rate control method provided in an embodiment of the present disclosure;

[0035] Figure 2 A schematic diagram illustrating an exemplary structure of a long short-term memory neural network is shown.

[0036] Figure 3 This diagram illustrates the structure of a cell in a long short-term memory neural network.

[0037] Figure 4 The flowchart illustrates the execution steps of the training process for the Long Short-Term Memory Neural Network in an embodiment of this disclosure.

[0038] Figure 5 The flowchart illustrates the execution steps of the optimization process of the Long Short-Term Memory Neural Network in an embodiment of this disclosure.

[0039] Figure 6 A schematic diagram of the structure of an electronic device provided in an embodiment of this disclosure is shown. Detailed Implementation

[0040] To facilitate understanding of this disclosure, a more complete description will now be given with reference to the accompanying drawings, in which preferred embodiments of the present disclosure are shown. However, this disclosure may be implemented in various forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to provide a thorough and complete understanding of the disclosure.

[0041] The IEEE 802.11 wireless LAN communication standard offers various transmission rates for its different communication modes. For example, IEEE 802.11a supports eight transmission rates from 6 Mbit / s to 54 Mbit / s; IEEE 802.11g extends 802.11b to support 12 transmission rates at a frequency of 2.4 GHz; the transmission rate supported by IEEE 802.11n depends on multiple factors such as modulation scheme, coding rate, number of spatial streams, and whether 40MHz bonding is used. These factors combine to produce a wide variety of transmission rates to choose from. Therefore, 802.11n uses a Modulation and Coding Scheme (MCS) to configure the radio frequency transmission rate.

[0042] The purpose of wireless adaptive rate control is to automatically select a transmission rate from various options within the current communication mode that is appropriate for the wireless channel quality, ensuring that high transmission rates are matched to good wireless channel quality and low transmission rates are matched to poor wireless channel quality. Existing wireless adaptive rate control methods are classified into two categories based on different channel quality measurement schemes: methods based on direct channel quality measurement and methods based on statistical calculations of channel quality.

[0043] Specifically, methods based on direct channel quality measurement, such as those using physical layer metrics like Signal-to-Interference and Noise Ratio (SNIR) and Received Signal Strength (RSS), estimate wireless channel quality. However, existing literature has shown that SNIR and RSS do not have a strong correlation with frame drop rate, and the cross-layer design approach of these methods faces difficulties in practical applications. Another method based on direct channel quality measurement involves sending one or more test frames at different rates at regular intervals to estimate the performance of different numbers of test frames in the channel. This method improves the real-time estimation capability of transmission rate, but because it cannot send too many test frames, the success or failure of a few test frames determines the rate selection, which can easily lead to misjudgments, as seen in the Sample Rate method. Therefore, methods based on direct channel quality measurement require additional feedback mechanisms to ensure a proper match between transmission rate and wireless channel quality.

[0044] Methods based on channel quality statistics include estimating channel quality based on the number of consecutive successful and consecutive failed transmissions, such as the Automatic Rate Fallback (ARF) method. In this method, if two consecutive acknowledgment characters (ACKs) are not received, the transmission rate is reduced by one level, and a timer is started. If ten consecutive ACKs are received or the timer expires, the transmission rate is increased. Another method uses the statistical characteristics of data frames over a recent period to determine the transmission rate, such as the Robust Rate Adaptation Algorithm (RRAA). This algorithm statistically analyzes frame drop rates and other information over a period (a window) and compares them with corresponding thresholds, effectively mitigating misjudgments caused by random frame drops. However, methods based on channel quality statistics struggle to respond promptly to channel changes, resulting in a technical problem where the transmission rate cannot adapt to channel quality in a timely manner.

[0045] In view of the above reasons, this disclosure provides a wireless adaptive rate control method, which uses a long short-term memory neural network model to achieve adaptive selection of transmission rate, thereby improving network throughput.

[0046] Figure 1 A flowchart of a wireless adaptive rate control method provided in an embodiment of this disclosure is shown. (Refer to...) Figure 1 The wireless adaptive rate control method includes:

[0047] Step S110: Obtain channel state information and network state information during data transmission;

[0048] Step S120: Input the channel state information and network state information into the long short-term memory neural network model to obtain the transmission rate output by the long short-term memory neural network model;

[0049] Step S130: Perform subsequent data transmission based on the transmission rate.

[0050] Specifically, the Long Short-Term Memory (LSTM) neural network model is a classic algorithm of Recurrent Neural Networks (RNNs). Figure 2 The diagram shown is an exemplary structural schematic of a Long Short-Term Memory (LSTM) neural network. (Refer to...) Figure 2The Long Short-Term Memory (LSTM) neural network 200 includes multiple cells 210 connected sequentially. Each cell 210 processes its own input information based on the output information of the previous cell to obtain its own output information. To perform this information processing, each cell 210 is equipped with a tanh layer, a sigmoid function unit, a pointwise multiplication unit, and a pointwise addition unit. Specifically, the tanh layer executes the tanh function, the sigmoid function unit executes the sigmoid function, the pointwise multiplication unit performs pointwise multiplication operations, and the pointwise addition unit performs pointwise addition operations.

[0051] Cell A, located in the middle of the three cells shown in the diagram. t (i.e., located at the mark A) t-1 Cells and labeled A t+1 Taking the relationship between cells as an example, the output information of the previous cell is C. t-1 and h t-1 The current cell's input information is X. t The current cell's output information is C. t and h t In this embodiment, the Long Short-Term Memory (LSTM) neural network 200 is used for adaptive selection of the transmission rate. Therefore, the input information X consists of channel state information and network state information, and the output information h is the transmission rate. It should be understood that the channel state information and network state information are not constant during data transmission. Therefore, the channel state information and network state information need to be collected once during each time-ordered acquisition period. The channel state information and network state information collected in different acquisition periods are input into multiple cells 210 of the LSM neural network 200 in chronological order. For example, cell A... t Input information X t Given the channel state information and network state information collected during the nth collection period, then cell A t-1 Input information X t-1 For the channel state information and network state information collected during the (n-1)th collection period, cell A t+1 Input information X t+1 This refers to the channel state information and network state information collected during the (n+1)th acquisition period. For cell 210, which inputs the channel state information and network state information collected during the nth acquisition period, its output transmission rate must be adapted to the wireless channel quality during the (n+1)th acquisition period to guide data transmission during that period. To ensure that the transmission rate can adapt to the channel quality in a timely manner even in application scenarios where the channel quality is constantly changing, the above acquisition period needs to be set to a short period of time, such as 100ms.

[0052] In practical applications, after acquiring the channel state and network state information for data transmission during the current acquisition period, this information can be input into the Long Short-Term Memory (LSTM) neural network model as the input information for the last cell in the model. The input information for other cells in the LTM model consists of the channel state and network state information acquired during the acquisition periods prior to the current acquisition period. Subsequently, the LTM model outputs the transmission rate used for subsequent data transmission through its last cell.

[0053] Figure 3 The image shows cell A in a Long Short-Term Memory (LSTM) neural network of 200 cells. t A structural diagram, below using cell A as an example. t This example illustrates the information processing principle of any cell in a 200-cell Long Short-Term Memory (LSTM) neural network. (Refer to...) Figure 3 Cell A t Using a horizontally running C-line (i.e., the long-term memory C-line, also known as the cell state or memory cell) 211 as the main line, forgetting gate 212, input gate 213, and output gate 214 are added on this basis to control cell A. t Output information C t and h t .

[0054] Among them, the task of the forgetting gate 212 is to receive cell A t-1 Transmitted long-term memory C t-1 And through the forgetting factor f shown in formula (1) t To decide whether to retain or forget cell A t-1 Long-term memory C t-1 Which part?

[0055] f t =σ(W f [h t-1 ,x t ]+b f (1)

[0056] The function of input gate 213 is to transmit the information i shown in formula (2) t And the information C' shown in formula (3) t Determine what new information is added to the cell state to supplement the information discarded by the forget gate.

[0057] i t =σ(W i [h t-1 ,x t ]+b i (2)

[0058] C't =σ(W j [h t-1 ,x t ]+b j (3)

[0059] Thus, cell A t Output information C t As shown in the following formula (4).

[0060] C t =f t *C t-1 +i t *C' t (4)

[0061] Next, the output gate outputs information o as shown in formula (5). t To determine which part of the cell state needs to be output.

[0062] o t =σ(W o [h t-1 ,x t ]+b o (5)

[0063] Finally, cell A t Output information h t As shown in formula (6).

[0064] h t =o t *tanh(C t (6)

[0065] Since the channel state information input into a cell represents the channel condition and is instantaneous information collected, this type of information should ideally be forgotten during the process of predicting and adapting the transmission rate to the wireless channel quality; while the network state information input into the cell represents the network condition and is information obtained through statistics during the collection period or some information that remains unchanged, this type of information should ideally be transmitted during the process of predicting and adapting the transmission rate to the wireless channel quality. In this way, the cell structure described above can effectively integrate channel state information and network state information through gating mechanisms and memory cells, thereby achieving adaptive selection of transmission rate and improving network throughput.

[0066] In formulas (1) to (6) above, σ represents the sigmoid function; W f W i W j and W o b represents the weights in a Long Short-Term Memory (LSTM) neural network. f bi b j and b o As a bias in the Long Short-Term Memory (LSTM) neural network, the LSM network adjusts these weights and biases during training to ensure that channel state information is appropriately forgotten and network state information is appropriately transmitted, thereby ultimately outputting transmission information that adapts to the quality of the wireless channel.

[0067] In one optional embodiment, the long short-term memory neural network model is trained through a training step. (See also...) Figure 4 Training step S20 includes:

[0068] Step S21: Obtain the current sample set. The samples in the current sample set have corresponding rate labels. The rate in the rate label is matched with the wireless channel quality after the corresponding sample collection period.

[0069] Step S22: Input the training samples used for training in the current sample set into the current long short-term memory neural network model to obtain the transmission rate output by the current long short-term memory neural network model;

[0070] Step S23: Determine the accuracy of the current long short-term memory neural network model based on the transmission rate and rate label corresponding to the training samples (accuracy is the quotient of the number of training samples with the same rate in the transmission rate and rate label to the total number of training samples).

[0071] Step S24: Determine whether the accuracy of the current long short-term memory neural network model has converged. If the accuracy has not converged, proceed to step S25; otherwise, end the training.

[0072] Step S25: Adjust the weights and biases of one or more gates in the current long short-term memory neural network model and determine the adjusted long short-term memory neural network model as the current long short-term memory neural network model. Then return to step S22 (the execution process of steps S25 to S22 is illustrated by step A in the figure) until the accuracy of the current long short-term memory neural network model converges and the training ends.

[0073] Furthermore, current long short-term memory neural network models undergo optimization steps after accuracy convergence. (See reference...) Figure 5 Optimization step S30 includes:

[0074] Step S31: Input the test samples used for testing in the current sample set into the current long short-term memory neural network model to obtain the transmission rate output by the current long short-term memory neural network model;

[0075] Step S32: Determine the accuracy of the current long short-term memory neural network model based on the transmission rate and rate label corresponding to the test sample (accuracy is the quotient of the number of test samples with the same rate in the transmission rate and rate label to the total number of test samples).

[0076] Step S33: Determine whether the accuracy of the current long short-term memory neural network model is not greater than the threshold. If it is not greater than the threshold, proceed to step S34; otherwise, end the optimization.

[0077] Step S34: Update the structural parameters of the current long short-term memory neural network model and determine the updated long short-term memory neural network model as the current long short-term memory neural network model. And / or, after resetting the types of information included in the samples, obtain multiple samples and determine the sample set composed of the multiple samples as the current sample set. Then return to execute step S22 (the execution process of steps S34 to S22 is shown in the figure through step A) until the accuracy of the current long short-term memory neural network model is greater than the threshold and the optimization ends.

[0078] Specifically, current long short-term memory neural network models have several structural parameters that cannot be changed through training, such as the number of cells, the number of input nodes, and the number of hidden nodes. It should be noted that... Figure 2 The Long Short-Term Memory (LSTM) neural network model shown is a relatively simple LSM model, mainly used to illustrate the information processing principle of each cell in the LSM model, and does not imply a limitation on the LSM model used in this disclosure. In practice, the structure of the LSM model is optimized through the above optimization step S30 to achieve an accuracy greater than the threshold. Of course, the above optimization step S30 can also be used to speed up the execution speed of the model while ensuring the model accuracy.

[0079] Furthermore, in step S34 above, the types of information included in the sample are reset, including: re-determining one or more pieces of information from multiple candidate channel state information and multiple candidate network state information as the information included in the sample. This can be to add one or more pieces of information to try to improve the accuracy of the model, or to replace one or more pieces of information to try to improve the accuracy of the model, or even to delete a few types of information to try to improve the model execution efficiency without affecting the accuracy of the model. If the accuracy of the current long short-term memory neural network model is greater than the threshold, the information included in each sample in the current sample set is the target information. Step S110, obtaining the channel state information and network state information during data transmission, includes: obtaining the target information during data transmission. Through numerous experiments, the inventors determined that the channel state information that has a significant impact on transmission rate includes signal-to-noise-plus-interference ratio, signal-to-noise ratio, received signal strength, signal receiving sensitivity, transmit power, error vector magnitude (EVM), transmit gain, and receive gain. Therefore, the aforementioned candidate channel state information can include these information. Similarly, the network state information that has a significant impact on transmission rate includes frame loss rate, frame error rate, acknowledgment character count, throughput, channel number, communication mode, and modulation scheme. Therefore, the aforementioned candidate network state information can include these information.

[0080] Furthermore, in steps S22 and S31, the transmission rate output by the current Long Short-Term Memory (LSTM) neural network model is one of multiple preset rates, which are pre-configured before step S22 based on the communication mode used for data transmission. Specifically, the multiple preset rates can be set to all transmission rates included in the communication mode used for data transmission, or they can be set to a subset of the transmission rates included in the communication mode used for data transmission. For example, if the communication mode used for data transmission is IEEE 802.11b, then the multiple preset rates can be set to a total of four, ordered from smallest to largest as 1Mb / s, 2Mb / s, 5.5Mb / s, and 11Mb / s.

[0081] It should be noted that the Long Short-Term Memory (LSTM) neural network model, after training and optimization, is applied to scenarios where the communication mode is the same as that used during training and optimization. In practice, after determining the communication mode of the scenario in which the LTM neural network model is applied, the aforementioned multiple preset rates are pre-configured according to the determined communication mode. Then, steps S20 and S30 are executed to obtain an LTM neural network model with an accuracy greater than a threshold. Subsequently, the obtained LTM neural network model is used for adaptive selection of the transmission rate.

[0082] Step S20 above continuously adjusts the weights and biases to ensure that the current long short-term memory neural network model achieves the highest possible stable accuracy after receiving the current input information. Step S30 above continuously adjusts the model's structural parameters and the types of input information to optimize the model's structural parameters and input information, thereby improving model accuracy and execution efficiency. The final long short-term memory neural network model applied in step S120 is a model with the required accuracy achieved through steps S20 and S30.

[0083] In another optional embodiment, step S21, obtaining the current sample set, includes: selecting a candidate sample set corresponding to the environment where the current data transmission occurs from multiple candidate sample sets as the current sample set, where one candidate sample set in the multiple candidate sample sets corresponds to one environment. In practice, the current sample set can be generated using existing Matlab selection models. Specifically, the environment where the data transmission occurs may be a clean environment with less interference, such as a shielded room, or a complex environment with more interference, such as a staircase. Interference in different environments can also affect the quality of the wireless channel. Therefore, using the candidate sample set corresponding to the environment where the current data transmission occurs as the current sample set for training and optimizing the Long Short-Term Memory Neural Network Model can ensure that the Long Short-Term Memory Neural Network Model achieves high accuracy in use.

[0084] In another optional embodiment, step S21, obtaining the current sample set, may include: performing convolution processing on multiple original samples to obtain multiple processed samples, and determining the set of multiple processed samples as the current sample set. In this way, since the data processing method of convolution enables each processed sample to integrate the information of the corresponding original sample and its surrounding original samples, each processed sample can avoid being affected by sudden interference as much as possible.

[0085] Step S21, obtaining the current sample set, may also include: performing a moving weighted average on multiple original samples collected within a preset time period to obtain a processed sample corresponding to the preset time period, and determining the set of multiple processed samples as the current sample set, wherein samples collected later within the preset time period have a larger weight in the moving weighted average processing. For example, the preset time period is from 100ms to 200ms, which is divided into 10 sub-time periods. One original sample is collected in each sub-time period. The weights of the 10 sub-time periods arranged in chronological order are 1%, 3%, 5%, 7%, 9%, 11%, 13%, 15%, 17%, and 19%, respectively. Finally, the original samples of the 10 sub-time periods are weighted and summed based on these weights to obtain a processed sample corresponding to 100ms to 200ms. Since the samples in the current sample set are used to predict future transmission rates, the processed sample obtained through moving weighting not only avoids the impact of sudden interference as much as possible, but also closely reflects the facts represented by the overall samples within the corresponding preset time period.

[0086] The wireless adaptive rate control method provided in this disclosure, through pre-configuration of multiple preset rates in a long short-term memory neural network model and training and optimization corresponding to actual application scenarios, can adaptively select the transmission rate under different communication models and in different environments, achieving high accuracy and meeting timeliness requirements. This wireless adaptive rate control method has been compared with traditional wireless adaptive rate control methods implemented on the same hardware board in the same environment. Network throughput measurements ultimately proved that the wireless adaptive rate control method provided in this disclosure has greater network throughput in both clean and complex environments.

[0087] To reduce the implementation difficulty of the aforementioned wireless adaptive rate control method, chips capable of directly acquiring channel state information can be used. This eliminates the need for a complex process combining data collection and statistics to obtain network state information. On the software side, a Long Short-Term Memory (LSTM) neural network model implemented in C can be incorporated into the chip's Software Development Kit (SDK) to implement the wireless adaptive rate control method.

[0088] Figure 6 The image shows an exemplary electronic device. (As shown) Figure 6As shown, the electronic device 1300 includes a memory 1310 and a processor 1320, as well as a program stored in the memory 1310 and executable on the processor 1320. When executed by the processor 1320, the program can implement the various processes in the embodiments of the wireless adaptive rate control method described above, and achieve the same technical effect. Of course, the electronic device may also include auxiliary sub-devices such as a power supply component 1330, a network interface 1340, and an input / output interface 1350.

[0089] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by instructions, or by instructions controlling related hardware. These instructions can be stored in a computer-readable storage medium and loaded and executed by a processor. Therefore, this disclosure also provides a computer-readable storage medium storing a computer program. When executed by a processor, the computer program can implement the various processes of the embodiments of the above-described wireless adaptive rate control method. The computer-readable storage medium can be, for example, a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk—any medium capable of storing program code.

[0090] Since the program stored in the readable storage medium can execute the steps of any of the wireless adaptive rate control methods provided in the embodiments of this disclosure, the beneficial effects achievable by any of the wireless adaptive rate control methods provided in the embodiments of this disclosure can be realized, as detailed in the preceding embodiments, and will not be repeated here. The specific implementation of each of the above operations can be found in the preceding embodiments, and will not be repeated here.

[0091] It should be noted that in describing the various embodiments in this specification, the focus is on the differences from other embodiments, while the same or similar parts between the various embodiments can be understood by referring to each other. Regarding the electronic device embodiment, since it is basically similar to the method embodiment, the relevant parts can be referred to the description of the method embodiment. Because this electronic device embodiment has the beneficial effects that can be achieved by the above-described method embodiments, as detailed in the preceding embodiments, it will not be repeated here.

[0092] Furthermore, it should be noted that in the apparatus and method of this disclosure, it is obvious that the components or steps can be decomposed and / or recombined. These decompositions and / or recombinations should be considered equivalent solutions of this disclosure. Moreover, the steps performing the above series of processes can naturally be executed in the order described, but are not necessarily required to be executed in chronological order; some steps can be executed in parallel or independently of each other. Those skilled in the art will understand that all or any step or component of the method and apparatus of this disclosure can be implemented in any computing device (including processors, storage media, etc.) or network of computing devices, in hardware, firmware, software, or a combination thereof, which can be achieved by those skilled in the art using their basic programming skills after reading the description of this disclosure.

[0093] Finally, it should be noted that the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance. Furthermore, the embodiments described above are merely specific implementations of this disclosure, used to illustrate the technical solutions of this disclosure, and not to limit it. The scope of protection of this disclosure is not limited thereto. Although this disclosure has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments, or make equivalent substitutions for some of the technical features, within the scope of the technology disclosed in this disclosure; and these modifications, changes, 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 disclosure, and should all be covered within the scope of protection of this disclosure. Therefore, the scope of protection of this disclosure should be determined by the scope of the claims.

Claims

1. A wireless adaptive rate control method, comprising: Acquire channel status information and network status information during data transmission; The channel state information and the network state information are input into the long short-term memory neural network model to obtain the transmission rate output by the long short-term memory neural network model; Subsequent data transmission will be based on the aforementioned transmission rate. The long short-term memory neural network model is trained through the following steps: Obtain the current sample set, wherein each sample in the current sample set has a rate label, and the rate in the rate label is adapted to the wireless channel quality after the corresponding sample collection period. Input the training samples used for training in the current sample set into the current long short-term memory neural network model to obtain the transmission rate output by the current long short-term memory neural network model; The accuracy of the current long short-term memory neural network model is determined based on the transmission rate and rate label corresponding to the training samples. If the accuracy does not converge, the weights and biases of one or more gates in the current long short-term memory neural network model are adjusted, and the adjusted long short-term memory neural network model is determined as the current long short-term memory neural network model. Return to the step of inputting the training samples used for training in the current sample set into the current long short-term memory neural network model, until the accuracy of the current long short-term memory neural network model converges.

2. The wireless adaptive rate control method according to claim 1, wherein, The current long short-term memory neural network model is further optimized through the following steps after the accuracy converges: Input the test samples used for testing in the current sample set into the current long short-term memory neural network model to obtain the transmission rate output by the current long short-term memory neural network model; The accuracy of the current long short-term memory neural network model is determined based on the transmission rate and rate label corresponding to the test sample. If the accuracy is not greater than the threshold, then update the structural parameters of the current long short-term memory neural network model and determine the updated long short-term memory neural network model as the current long short-term memory neural network model, and / or, after resetting the types of information included in the sample, obtain multiple samples and determine the sample set composed of the multiple obtained samples as the current sample set; Return to the step of inputting the training samples used for training in the current sample set into the current long short-term memory neural network model, until the accuracy of the current long short-term memory neural network model is greater than the threshold.

3. The wireless adaptive rate control method according to claim 2, wherein, The types of information included in the sample are reset, including: re-determining one or more pieces of information from multiple candidate channel state information and multiple candidate network state information as the information included in the sample; When the accuracy of the current long short-term memory neural network model is greater than the threshold, the information included in each sample in the current sample set is the target information. The channel state information and network state information during data transmission are obtained, including: obtaining the target information during data transmission.

4. The wireless adaptive rate control method according to claim 3, wherein, The multiple candidate channel state information includes: signal and noise plus interference ratio, signal-to-noise ratio, received signal strength, signal receiving sensitivity, transmit power, error vector magnitude, transmit gain, and receive gain; The multiple candidate network status information includes: frame loss rate, frame error rate, number of acknowledgment characters, throughput, channel number, communication mode, and modulation method.

5. The wireless adaptive rate control method according to claim 1 or 2, wherein, The transmission rate output by the current long short-term memory neural network model is one of several preset rates; The multiple preset rates are pre-configured according to the communication mode used for data transmission before the training samples used for training in the current sample set are input into the current long short-term memory neural network model.

6. The wireless adaptive rate control method according to claim 1, wherein, Obtaining the current sample set includes: selecting a candidate sample set corresponding to the environment where the current data transmission occurs from multiple candidate sample sets as the current sample set, wherein one candidate sample set in the multiple candidate sample sets corresponds to one environment.

7. The wireless adaptive rate control method according to claim 1, wherein, Retrieve the current sample set, including: Multiple original samples are convolutionally processed to obtain multiple processed samples, and the set of the multiple processed samples is determined as the current sample set; Alternatively, a moving weighted average is applied to multiple original samples collected within a preset time period to obtain a processed sample corresponding to the preset time period, and the set of multiple processed samples is determined as the current sample set, wherein, within the preset time period, the weight of samples collected at later times in the moving weighted average process is greater than the weight of samples collected at earlier times in the moving weighted average process; the later times are later than the earlier times.

8. An electronic device, characterized in that, include: A processor, a memory, and a program stored in the memory and executable on the processor, wherein the program, when executed by the processor, implements the steps of the wireless adaptive rate control method as described in any one of claims 1-7.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program or instructions that, when executed by a processor, implement the steps of the wireless adaptive rate control method as described in any one of claims 1-7.

Images