Wlan awareness method, apparatus and system

Abstract

Embodiments of the present application provide a WLAN sensing method, device and system. In the embodiments of the present application, a management device in the WLAN sensing system sends a first training trigger message to a training device, and the message carries information associated with WLAN sensing model training, such as model ID, data set ID, AI training algorithm, etc., so that the training device performs model training based on the AI training algorithm and the target data set indicated by the message after determining that it can support the training of the target sensing model, and uploads the trained target sensing model to the management device for storage management to be used for subsequent WLAN sensing processing, thereby realizing AI-assisted WLAN sensing. Compared with the prior art of extracting features using traditional signal processing technology for WLAN sensing, AI technology can effectively extract more subtle features for WLAN sensing, thereby improving the accuracy of WLAN sensing.

Description

Technical Field

[0001] This application relates to the field of communication technology, and in particular to wireless local area network (WLAN) sensing methods, devices and systems. Background Technology

[0002] WLAN sensing is a technology that utilizes the propagation characteristics of wireless signals (such as Wi-Fi signals) in the environment, such as their ability to reflect, penetrate, and bend on object surfaces during propagation, to perform sensing tasks. Its core principle is that when a person or object moves within the wireless signal coverage area, wireless channel data, such as Channel State Information (CSI), estimated based on the wireless signal, changes accordingly. This wireless channel data can be used to achieve WLAN sensing. Sensing tasks can include, for example, gesture control, fall detection, target tracking, intrusion detection, activity recognition, and vital sign monitoring.

[0003] In practical applications, traditional signal processing techniques are typically used to extract features from the aforementioned wireless channel data for WLAN sensing processing. However, traditional signal processing techniques have limited capabilities in extracting subtle features, thus affecting the accuracy of WLAN sensing. Summary of the Invention

[0004] In view of this, this application provides a WLAN sensing method, apparatus and system to improve the accuracy of WLAN sensing.

[0005] This application provides a WLAN sensing method, which is applied to a management device in a WLAN sensing system. The WLAN sensing system further includes a training device. The method includes:

[0006] During the offline training of the target perception model, a first training trigger message is sent to the training device. The first training trigger message includes at least: a model identification (ID) field, a dataset ID field, and an algorithm field. The model ID field is used to indicate the model ID of the target perception model. The dataset ID field is used to indicate the dataset ID of the target dataset to be used in this training. The algorithm field is used to indicate the target artificial intelligence (AI) training algorithm to be used in this training.

[0007] The system receives a first training response message returned by the training device. The first training response message includes at least an Acknowledge (ACK) field. The ACK field indicates whether the response is successful or unsuccessful. A successful response indicates that the training device can support the training of the target perception model, and a unsuccessful response indicates that the training device cannot support the training of the target perception model.

[0008] The system receives a model delivery message sent by the training device and stores the trained target perception model indicated by the model delivery message. The trained target perception model is obtained by the training device training the target perception model based on the target AI training algorithm and the target dataset. The trained target perception model is used for WLAN perception processing.

[0009] This application embodiment also provides a WLAN sensing method, which is applied to a training device in a WLAN sensing system, wherein the WLAN sensing system further includes a management device; the method includes:

[0010] During the offline training of the target perception model, a first training trigger message is received from the management device. The first training trigger message includes at least: a model ID field, a dataset ID field, and an algorithm field. The model ID field is used to indicate the model ID of the target perception model. The dataset ID field is used to indicate the dataset ID of the target dataset to be used in this training. The algorithm field is used to indicate the target AI training algorithm to be used in this training.

[0011] The first training response message is returned to the management device; the first training response message includes at least an ACK field; the ACK field indicates whether the response is successful or failed, the successful response indicates that the training device can support the training of the target perception model, and the failed response indicates that the training device cannot support the training of the target perception model;

[0012] A model delivery message is sent to the management device, the model delivery message being used to indicate a trained target perception model; the trained target perception model is obtained by the training device training the target perception model based on a target AI training algorithm and a target dataset; the trained target perception model is used for WLAN perception processing.

[0013] This application embodiment also provides a WLAN sensing device, which is applied to a management device in a WLAN sensing system, wherein the WLAN sensing system further includes a training device; the device includes:

[0014] The first sending module is used to send a first training trigger message to the training device during the offline training process of the target perception model. The first training trigger message includes at least: a model ID field, a dataset ID field, and an algorithm field. The model ID field is used to indicate the model ID of the target perception model. The dataset ID field is used to indicate the dataset ID of the target dataset to be used in this training. The algorithm field is used to indicate the target AI training algorithm to be used in this training.

[0015] A first receiving module is configured to receive a first training response message returned by the training device; the first training response message includes at least an ACK field; the ACK field indicates whether the response is successful or failed, the successful response indicates that the training device can support the training of the target perception model, and the failed response indicates that the training device cannot support the training of the target perception model;

[0016] The first receiving module is further configured to receive a model delivery message sent by the training device and store the trained target perception model indicated by the model delivery message; the trained target perception model is obtained by the training device training the target perception model based on the target AI training algorithm and the target dataset; the trained target perception model is used for WLAN perception processing.

[0017] This application embodiment also provides a WLAN sensing device, which is applied to a training device in a WLAN sensing system, wherein the WLAN sensing system further includes a management device; the device includes:

[0018] The second receiving module is used to receive a first training trigger message issued by the management device during the offline training process of the target perception model. The first training trigger message includes at least: a model ID field, a dataset ID field, and an algorithm field. The model ID field is used to indicate the model ID of the target perception model. The dataset ID field is used to indicate the dataset ID of the target dataset to be used in this training. The algorithm field is used to indicate the target AI training algorithm to be used in this training.

[0019] The second sending module is used to return a first training response message to the management device; the first training response message includes at least an ACK field; the ACK field indicates whether the response is successful or failed, the successful response indicates that the training device can support the training of the target perception model, and the failed response indicates that the training device cannot support the training of the target perception model;

[0020] The second sending module is further configured to send a model delivery message to the management device, the model delivery message being used to indicate a trained target perception model; the trained target perception model is obtained by the training device training the target perception model based on a target AI training algorithm and a target dataset; the trained target perception model is used for WLAN perception processing.

[0021] This application embodiment also provides a WLAN sensing system, the WLAN sensing system including: a management device and a training device;

[0022] The management device is used to send a first training trigger message to the training device during the offline training process of the target perception model. The first training trigger message includes at least: a model ID field, a dataset ID field, and an algorithm field. The model ID field is used to indicate the model ID of the target perception model. The dataset ID field is used to indicate the dataset ID of the target dataset to be used in this training. The algorithm field is used to indicate the target AI training algorithm to be used in this training.

[0023] The training device is configured to receive a first training trigger message issued by the management device and return a first training response message to the management device; the first training response message includes at least an ACK field; the ACK field indicates whether the response is successful or failed, the successful response indicates that the training device can support the training of the target perception model, and the failed response indicates that the training device cannot support the training of the target perception model;

[0024] The training device is also used to send a model delivery message to the management device, the model delivery message being used to indicate the trained target perception model; the trained target perception model is obtained by the training device training the target perception model based on the target AI training algorithm and the target dataset; the trained target perception model is used for WLAN perception processing;

[0025] The management device is also used to receive the model delivery message sent by the training device and store the trained target perception model indicated by the model delivery message.

[0026] As can be seen from the above technical solutions, in this embodiment of the application, the management device in the WLAN sensing system sends a first training trigger message to the training device. The first training trigger message carries information related to the training of the WLAN sensing model, such as the model ID, dataset ID, and the AI ​​training algorithm to be used in this training. After the training device determines that it can support the training of the target sensing model based on the first training trigger message, it trains the target sensing model based on the AI ​​training algorithm and target dataset indicated by the message, and uploads the trained target sensing model to the management device for storage and management for subsequent WLAN sensing processing. This realizes AI-assisted WLAN sensing. Compared with the existing method of using traditional signal processing technology to extract features for WLAN sensing, AI technology can effectively extract more subtle features for WLAN sensing, thereby improving the accuracy of WLAN sensing.

[0027] Furthermore, this application embodiment also clarifies the signaling interaction (i.e. message interaction) process between various devices in the WLAN sensing system, such as management devices and training devices, as well as the specific signaling structure, which is conducive to the practical implementation of AI-assisted sensing services. Attached Figure Description

[0028] The accompanying drawings, which are incorporated in and form part of this application, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0029] Figure 1 This is a schematic diagram of the method flow provided in the embodiments of this application.

[0030] Figure 2 This is a schematic diagram of another method flow provided for an embodiment of this application.

[0031] Figure 3 This is a schematic diagram of the system structure provided in an embodiment of this application.

[0032] Figure 4 This is a schematic diagram illustrating the offline training implementation provided in an embodiment of this application.

[0033] Figure 5 This is a schematic diagram illustrating the online training implementation provided in an embodiment of this application.

[0034] Figure 6 This is another system structure diagram provided for an embodiment of this application.

[0035] Figure 7 This is a schematic diagram of another offline training implementation provided in an embodiment of this application.

[0036] Figure 8 This is a schematic diagram illustrating another online training implementation provided in an embodiment of this application.

[0037] Figure 9 This is a schematic diagram of the device structure provided in the embodiments of this application.

[0038] Figure 10 This is a schematic diagram of another device structure provided for an embodiment of this application.

[0039] Figure 11 This is a schematic diagram of the electronic device structure provided in an embodiment of this application. Detailed Implementation

[0040] To enable those skilled in the art to better understand the technical solutions provided in the embodiments of this application, and to make the above-mentioned objectives, features and advantages of the embodiments of this application more apparent and understandable, the technical solutions in the embodiments of this application will be further described in detail below with reference to the accompanying drawings.

[0041] See Figure 1 , Figure 1 This is a flowchart illustrating a method provided in an embodiment of this application. This flowchart describes the WLAN sensing method provided in this embodiment from the perspective of a management device in a WLAN sensing system, which also includes a training device. In this embodiment, as one example, the method can be applied to WLAN sensing scenarios, such as gesture control scenarios, fall detection scenarios, intrusion detection scenarios, etc., and this embodiment is not specifically limited to these scenarios.

[0042] like Figure 1 As shown, the process may include the following steps:

[0043] Step 101: During the offline training of the target perception model, a first training trigger message is sent to the training device. The first training trigger message includes at least: a model ID field, a dataset ID field, and an algorithm field. The model ID field is used to indicate the model ID of the target perception model. The dataset ID field is used to indicate the dataset ID of the target dataset to be used in this training. The algorithm field is used to indicate the target AI training algorithm to be used in this training.

[0044] In this embodiment, as one example, when the management device detects a training trigger event, it sends an inference trigger message to the training device.

[0045] In this embodiment, the detection of the training trigger event can be implemented in many ways. For example, as an embodiment, a WLAN sensing application (APP) can be pre-configured. This WLAN sensing APP can interact with the management device. When a user needs to train a target perception model, they can send a training instruction to the management device through the WLAN sensing APP. The training instruction may include: the model ID of the target perception model, the dataset ID of the target dataset to be used in this training, and the AI ​​training algorithm to be used in this training. When the management device receives the training instruction, it will determine that a training trigger event has been detected, and send a first training trigger message to the training device based on the training instruction to instruct the training device to train the target perception model based on the first training trigger message.

[0046] For example, as another embodiment, a user can issue training instructions to the management device by entering command lines on the management device, and the management device determines that an inference trigger event has been detected when it receives the training instructions.

[0047] The model ID in the aforementioned training instructions can be determined based on the application scenario applicable to the target perception model and the AI ​​training algorithm to be used for training. However, the specific method of determination is not limited here. Different model IDs correspond to different application scenarios applicable to the WLAN perception model and / or different AI training algorithms to be used for training.

[0048] Optionally, the above application scenarios may include gesture control scenarios, fall detection scenarios, intrusion detection scenarios, etc.; the above AI training algorithms may include Convolutional Neural Network (CNN) algorithms, Long Short-Term Memory (LSTM) algorithms, Deep Q-Network (DQN) algorithms, and Transformer self-attention network algorithms, etc., and this embodiment is not specifically limited.

[0049] In this embodiment, the management device locally maintains model IDs for pre-trained WLAN sensing models and / or model IDs for WLAN sensing models currently being trained, as well as the states of the WLAN sensing models corresponding to each model ID, such as idle state, training state, and inference state. Based on this, when the management device receives the aforementioned training instruction, it checks whether the model ID in the training instruction exists locally. If it does not exist, it adds the model ID to its local storage.

[0050] Step 102: Receive the first training response message returned by the training device; the first training response message includes at least: an ACK field; the ACK field indicates whether the response is successful or failed, a successful response indicates that the training device can support the training of the target perception model, and a failed response indicates that the training device cannot support the training of the target perception model.

[0051] In this embodiment, as one example, when the training device receives a first training trigger message from the management device, it determines whether it can support the training of the target perception model indicated by the first training trigger message based on the remaining available computing resources in the device. Based on the determination result, it generates a first training response message for the first training trigger message and returns it to the management device. The specific method for determining whether the remaining available computing resources in the device can support the training of the target perception model will be described with examples below and will not be elaborated here.

[0052] In this embodiment, the management device also maintains the states of the datasets required for training the WLAN sensing model locally, such as idle state, training state, and acquisition state, represented in the form of dataset ID-state. Based on this, as an example, after receiving the first training response message returned by the training device in response to the first training trigger message, if the first training response message indicates a successful response, the management device sets the states of both the target sensing model and the target dataset to the training state; if the first training response message indicates a failed response, and the model ID indicated by the first training trigger message is a newly added model ID in the local storage, the management device deletes that model ID from the local storage to avoid mistakenly identifying the WLAN sensing model corresponding to that model ID as a trained WLAN sensing model suitable for inference during subsequent model inference. Furthermore, for example, an alarm message indicating that the WLAN sensing model corresponding to that model ID has failed to train can also be returned to the aforementioned WLAN sensing APP to alert the user.

[0053] Step 103: Receive the model delivery message sent by the training device and store the trained target perception model indicated in the model delivery message. The trained target perception model is obtained by the training device based on the target AI training algorithm and the target dataset. The trained target perception model is used for WLAN sensing processing.

[0054] In this embodiment, a dataset corresponding to the same dataset ID can be used by multiple WLAN sensing models simultaneously, meaning it can serve as the training dataset for multiple WLAN sensing models at the same time. The management device also maintains a local database of WLAN sensing models using each dataset, represented in the form of dataset ID - model ID. Based on this, as an example, the management device receives a model delivery message from the training device and stores the trained target sensing model indicated by the delivery message for later use in WLAN sensing inference. Furthermore, after receiving the model delivery message from the training device, the management device can set the target sensing model to an idle state, and set the target dataset to an idle state when it is not used by any other model besides the target sensing model.

[0055] In this embodiment, as one example, the training device may store a dataset locally. After determining that it can support the training of the target perception model, the training device can obtain the dataset corresponding to the dataset ID indicated by the first training trigger message from the local storage, and train the target perception model based on the obtained dataset and the target AI training algorithm indicated by the first training trigger message. The specific method of training the target perception model is not specifically limited here.

[0056] This concludes the process. Figure 1 The process is shown below.

[0057] pass Figure 1 As shown in the process, in this embodiment of the application, the management device in the WLAN sensing system sends a first training trigger message to the training device. The first training trigger message carries information related to the training of the WLAN sensing model, such as the model ID, dataset ID, and the AI ​​training algorithm to be used in this training. After the training device determines that it can support the training of the target sensing model based on the first training trigger message, it trains the target sensing model based on the AI ​​training algorithm and target dataset indicated by the message. The trained target sensing model is then uploaded to the management device for storage and management for subsequent WLAN sensing processing. This realizes AI-assisted WLAN sensing. Compared with the existing method of using traditional signal processing technology to extract features for WLAN sensing, AI technology can effectively extract more subtle features for WLAN sensing, thereby improving the accuracy of WLAN sensing.

[0058] Furthermore, this application embodiment also clarifies the signaling interaction (i.e. message interaction) process between various devices in the WLAN sensing system, such as management devices and training devices, as well as the specific signaling structure, which is conducive to the practical implementation of AI-assisted sensing services.

[0059] The offline training process of the above-mentioned target perception model will be further described from the perspective of equipment management:

[0060] In this embodiment, the WLAN sensing system also includes a storage device. The storage device can be used to store trained WLAN sensing models and datasets, etc.

[0061] Based on this, as an example, during the offline training of the target perception model, the management device sends a second training trigger message to the training device; the second training trigger message includes at least: a model ID field, a dataset ID field, an algorithm field, and a storage address field; the storage address field is used to indicate the address information of the storage device.

[0062] The management device receives a second training response message returned by the training device in response to the second training trigger message. The second training response message includes at least an ACK field, which indicates whether the response is successful or failed. A successful response indicates that the training device can support the training of the target perception model, while a failed response indicates that the training device cannot support the training of the target perception model.

[0063] If the second training response message indicates a successful response, a third training trigger message is sent to the storage device, enabling the storage device to obtain the target dataset based on the third training trigger message and send a first data reporting message to the training device. The third training trigger message includes at least: a dataset ID field and a training address field; the training address field indicates the address information of the training device. The first data reporting message includes at least: a dataset ID field and a first data payload field; the first data payload field indicates the target dataset obtained based on the third training trigger message.

[0064] The management device receives a training completion message from the training device; the training completion message is sent by the training device to the management device after it has sent the trained target perception model to the storage device for storage.

[0065] In this embodiment, the address information may include, for example, a Media Access Control (MAC) address, an Internet Protocol (IP) address, and a port number, etc., but is not specifically limited here.

[0066] In this embodiment, as one example, the management device receives a third training response message returned by the storage device. The third training response message indicates whether the response was successful or failed. A successful response indicates that the storage device was able to successfully obtain the target dataset, while a failed response indicates that the storage device was unable to successfully obtain the target dataset. If the third training response message indicates a successful response, the states of both the target perception model and the target dataset are set to the training state. If the second training response message indicates a failed response, or if the third training response message indicates a failed response, then if the model ID indicated by the second training trigger message is a newly added model ID in the local storage, that model ID is deleted from the local storage.

[0067] Furthermore, after receiving the training completion message, the management device sets the state of the target perception model to an idle state, and sets the state of the target dataset to an idle state when the target dataset is not used by any other model besides the target perception model.

[0068] The online training process of the aforementioned target perception model is described below from the perspective of equipment management:

[0069] In this embodiment, as one example, the above-described WLAN sensing system further includes a data acquisition device for acquiring training data, such as wireless channel data. Here, wireless channel data can refer to data associated with wireless signals, such as, but not limited to, CSI data, Received Signal Strength Indication (RSSI) data, etc.

[0070] Based on this, as an example, during the online training of the target perception model, the management device sends a fourth training trigger message to the training device and receives a fourth training response message returned by the training device; the fourth training trigger message includes at least: a model ID field, an algorithm field, and a label field; the label field is used to indicate the label of the wireless channel data to be collected in this training; the fourth training response message includes at least: the aforementioned ACK field.

[0071] If the fourth training response message indicates a successful response, the management device sends a collection trigger message to the data acquisition device to be used in this training, so that the data acquisition device can collect wireless channel data based on the collection trigger message and send a second data reporting message to the training device. The collection trigger message includes at least a tag field, a collection configuration field, and a training address field. The second data reporting message includes at least a tag field and a second data payload field. The second data payload field is used to indicate the wireless channel data collected by the data acquisition device based on the collection trigger message. The collection configuration field is used to indicate the configuration parameters for data acquisition by the data acquisition device.

[0072] The management device receives the model delivery message sent by the training device and stores the trained target perception model indicated by the model delivery message; the trained target perception model is obtained by the training device based on the target AI training algorithm and wireless channel data.

[0073] Optionally, the above configuration parameters may include, for example, radio frequency band, bandwidth, number of spatial streams, and number of subcarrier groups, etc., which are not specifically limited here.

[0074] In this embodiment, as one example, the management device receives a collection response message returned by the data acquisition device in response to the collection trigger message; the collection response message indicates whether the response is successful or unsuccessful. A successful response indicates that the data acquisition device can support the collection of wireless channel data indicated by the collection trigger message, while a unsuccessful response indicates that the data acquisition device cannot support the collection of wireless channel data indicated by the collection trigger message.

[0075] If the acquisition response message indicates a successful response, the state of the target perception model is set to the training state, and the states of the target dataset and the corresponding data acquisition device are both set to the acquisition state; otherwise, if the model ID and dataset ID indicated by the second training trigger message are newly added model IDs in the local machine, the model ID and dataset ID are deleted from the local machine, and a training termination message is sent to the training device to instruct the training device to terminate the online training of the target perception model.

[0076] In this embodiment, as one example, after receiving a data acquisition trigger message, the data acquisition device determines whether it can support the acquisition of wireless channel data indicated by the acquisition trigger message based on its remaining available computing resources. Specifically, for example, a preset resource consumption algorithm can be used to calculate the computing resources (such as memory resources and CPU resources) required to acquire wireless channel data once based on the configuration parameters indicated by the acquisition trigger message. Then, it is compared whether the remaining available computing resources of the device are greater than or equal to the calculated computing resources. If so, it is determined that it can support the acquisition of wireless channel data indicated by the acquisition trigger message; otherwise, it is determined that it cannot support the acquisition of wireless channel data indicated by the acquisition trigger message.

[0077] In this embodiment, as an example, if the fourth training response message indicates a failure, the management device will delete the model ID and dataset ID from the local storage if the model ID and dataset ID indicated by the fourth training trigger message are newly added model IDs in the local storage.

[0078] After receiving the model delivery message, the management device sends a data acquisition termination message to all data acquisition devices to be used in this training. When it receives the data acquisition termination completion message returned by each data acquisition device in response to the data acquisition termination message, it sets the state of the target perception model to the idle state. When the target dataset and data acquisition devices are not used by other models other than the target perception model, the state of both the target dataset and the data acquisition devices is set to the idle state.

[0079] Based on the above description, as an embodiment, the WLAN sensing system further includes a storage device. Accordingly, during the online training of the target sensing model, the management device sends a fifth training trigger message to the training device and receives a fifth training response message returned by the training device in response to the fifth training trigger message; the fifth training trigger message includes at least: a model ID field, an algorithm field, a label field, and a storage address field; the fifth training response message includes at least: an ACK field.

[0080] If the fifth training response message indicates a successful response, the management device sends a collection trigger message to all data acquisition devices to be used in this training, so that the data acquisition devices can collect wireless channel data based on the received collection trigger message and send a second data reporting message to the training device.

[0081] The management device receives a training completion message from the training device; the training completion message is sent by the training device to the management device after it has sent the trained target perception model to the storage device for storage.

[0082] In this embodiment, as an example, the fifth training trigger message, the acquisition trigger message, and the second data reporting message all further include a dataset ID field; based on this, after the training device completes the training of the target perception model, it takes the wireless channel data received during this training process as data belonging to the target dataset and sends it to the storage device for storage.

[0083] In this embodiment, as an example, if the fifth training response message indicates a response failure, the management device will delete the model ID and the dataset ID from the local storage if the model ID and dataset ID indicated by the fifth training trigger message are newly added model IDs in the local storage.

[0084] After receiving the training completion message, the management device sends a collection termination message to all data acquisition devices used in this training. Upon receiving the collection termination completion message from each data acquisition device, the management device sets the state of the target perception model to idle. When the target dataset and data acquisition devices are not used by other models besides the target perception model, the management device sets the state of both the target dataset and the data acquisition devices to idle.

[0085] It should be noted that, similar to the first training trigger message in step 101 above, the second, fourth, and fifth training trigger messages in this step are also issued by the management device when a training trigger event is detected. Correspondingly, the second, fourth, and fifth training response messages in this step are also similar to the first training response message above, with the ACK field indicating whether the response was successful or failed. A successful response indicates that the training device can support the training of the target perception model, while a failed response indicates that the training device cannot support the training of the target perception model.

[0086] The WLAN sensing method provided in this application embodiment is described below from the perspective of training equipment:

[0087] See Figure 2 , Figure 2 This is a flowchart illustrating another WLAN sensing method provided in an embodiment of this application. This method is applied to a training device in a WLAN sensing system. Figure 2 As shown, the process may include the following steps:

[0088] Step 201: During the offline training of the target perception model, receive a first training trigger message from the management device; the first training trigger message includes at least: a model ID field, a dataset ID field, and an algorithm field; the model ID field is used to indicate the model ID of the target perception model; the dataset ID field is used to indicate the dataset ID of the target dataset to be used in this training; the algorithm field is used to indicate the target AI training algorithm to be used in this training.

[0089] Step 202: Return a first training response message to the management device; the first training response message includes at least: an ACK field; the ACK field indicates whether the response is successful or failed, a successful response indicates that the training device can support the training of the target perception model, and a failed response indicates that the training device cannot support the training of the target perception model.

[0090] In this embodiment, as an example, when the training device receives the first training trigger message from the management device, it determines whether it can support the training of the target perception model indicated by the first training trigger message based on the remaining available computing resources in the device. These computing resources may include, for example, central processing unit (CPU) resources and memory resources. In this embodiment, the CPU resources required for training each AI training algorithm can be pre-configured on the training device based on actual application needs.

[0091] Based on this, as an example, the above-mentioned determination of whether the training of the target perception model indicated by the first training trigger message can be based on the remaining available computing resources in the device. In specific implementation, it can be as follows: compare whether the remaining available CPU resources in the training device are greater than or equal to the CPU resources required for training the target AI training algorithm, and compare whether the remaining available memory resources in the training device are greater than or equal to the memory resources required for the target dataset; if so, it is determined that the training device can support the training of the target perception model; otherwise, it is determined that the training device cannot support the training of the target perception model.

[0092] Step 203: Send a model delivery message to the management device. The model delivery message indicates the trained target perception model. The trained target perception model is used for WLAN perception processing.

[0093] In this embodiment, as an example, the pre-trained target perception model is obtained by training the target perception model based on the target AI training algorithm and the target dataset after the training device has determined that it can support the training of the target perception model.

[0094] In this embodiment, training configuration information corresponding to each AI training algorithm can be pre-configured on the training device based on actual application needs, for use during model training. Optionally, the training configuration information corresponding to any AI training algorithm may include information such as the algorithm parameters and training stopping conditions of the AI ​​training algorithm; this is not specifically limited, as long as it ensures that model training can be achieved. Based on this, as an example, when the training device trains a target perception model based on a target AI training algorithm and a target dataset, it can call the training configuration information corresponding to the target AI training algorithm locally to train the target perception model using the training configuration information and the target dataset. How specifically the training configuration information and the target dataset are used to train the target perception model is not specifically limited here.

[0095] This concludes the process. Figure 2 The process is shown below.

[0096] pass Figure 2As shown in the process, in this embodiment of the application, the training device in the WLAN sensing system receives a first training trigger message issued by the management device. The first training trigger message carries information related to the training of the WLAN sensing model, such as the model ID, dataset ID, and the AI ​​training algorithm to be used in this training. After determining that it can support the training of the target sensing model based on the first training trigger message, the training device trains the target sensing model based on the AI ​​training algorithm and target dataset indicated by the message, and uploads the trained target sensing model to the management device for storage and management for subsequent WLAN sensing processing. This realizes AI-assisted WLAN sensing. Compared with the existing method of using traditional signal processing technology to extract features for WLAN sensing, AI technology can effectively extract more subtle features for WLAN sensing, thereby improving the accuracy of WLAN sensing.

[0097] Furthermore, this application embodiment also clarifies the signaling interaction (i.e. message interaction) process between various devices in the WLAN sensing system, such as management devices and training devices, as well as the specific signaling structure, which is conducive to the practical implementation of AI-assisted sensing services.

[0098] The offline training process of the above-mentioned target perception model will be further described from the perspective of training equipment:

[0099] In this embodiment, the WLAN sensing system also includes a storage device. The storage device can be used to store trained WLAN sensing models and datasets, etc.

[0100] Based on this, as an example, during the offline training of the target perception model, the training device receives a second training trigger message issued by the management device.

[0101] The training device returns a second training response message to the management device in response to the second training trigger message. When the second training response message indicates a successful response, the management device sends a third training trigger message to the storage device, instructing the storage device to obtain the target dataset based on the third training trigger message and send a first data reporting message to the training device. The third training trigger message includes at least a dataset ID field and a training address field. The training address field is used to indicate the address information of the training device.

[0102] The training device sends a training completion message to the management device; the training completion message is sent by the training device after it has sent the trained target perception model to the storage device for storage.

[0103] The online training process of the above-mentioned target perception model is described below from the perspective of the training equipment:

[0104] In this embodiment, the aforementioned WLAN sensing system also includes a data acquisition device. Based on this, as an example, during the online training of the target sensing model, the training device receives a fourth training trigger message issued by the management device.

[0105] The training device returns a fourth training response message to the management device, so that when the fourth training response message indicates a successful response, the management device sends a data acquisition trigger message to the data acquisition device to be used in this training, instructing the data acquisition device to acquire wireless channel data based on the data acquisition trigger message, and sends a second data reporting message to the training device.

[0106] The training device sends a model delivery message to the management device. The model delivery message indicates the trained target perception model. The trained target perception model is obtained by the training device based on the target AI training algorithm and wireless channel data.

[0107] Based on the above description, as an embodiment, the aforementioned WLAN sensing system also includes a storage device. Accordingly, during the online training process of the target sensing model, the training device receives a fifth training trigger message issued by the management device.

[0108] After receiving the fifth training trigger message, the training device sends a fifth training response message to the management device. When the fifth training response message indicates a successful response, the management device sends a collection trigger message to all data acquisition devices to be used in this training. This enables the data acquisition devices to collect wireless channel data based on the collection trigger message and send a second data reporting message to the training device.

[0109] The training device sends a training completion message to the management device; this training completion message is sent by the training device after it has sent the trained target perception model to the storage device for storage.

[0110] To facilitate understanding of the specific implementation process of the above WLAN sensing method, specific embodiments are described below.

[0111] See Figure 3 , Figure 3 This is a schematic diagram of the structure of a WLAN sensing system provided in an embodiment of this application. Figure 3 As shown, the WLAN sensing system 300 includes: a management device 301, a training device 302, and a data acquisition device 303.

[0112] Optionally, the management device 301, the training device 302, and the data acquisition device 303 can be the same device or different devices; there is no specific limitation here.

[0113] For example, the management device 301 may be a wireless controller (AC) in a WLAN, the training device 302 may be a sensing server, and the data acquisition device 303 may be a wireless access point (AP) or a wireless station (STA).

[0114] In this embodiment, the WLAN sensing method described above may include offline training and online training of the WLAN sensing model. Offline training can be performed using a pre-collected dataset, while online training requires training data collected in real time. In this embodiment, the WLAN sensing method can be implemented through message interaction between devices in the WLAN sensing system.

[0115] As an example, the message format of any message in the above WLAN sensing method flow can be, for example, a Type-Length-Value (TLV) format. The fields related to model training in this message (such as the Model ID field, Dataset ID field, etc.) belong to the V field in the TLV format. It should be noted that the message formats of all messages shown in the following tables are all in TLV format.

[0116] The following is based on Figure 3 The system shown illustrates the offline training process as an example:

[0117] As an example, such as Figure 4 As shown, the offline training process may include the following steps:

[0118] Step 401: When the management device detects a training trigger event, it sends a training trigger message 1 to the training device.

[0119] The message structure of training trigger message 1 is shown in Table 1 below.

[0120] Step 402: When the training device receives the training trigger message 1, it sends a training response message 1 to the management device in response to the training trigger message 1.

[0121] The message structure of training response message 1 is shown in Table 2 below. A successful response in Table 2 indicates that the training device can support the training of the WLAN sensing model corresponding to the model ID indicated by training trigger message 1 (that is, the training device can support this training); a failed response in Table 2 indicates that the training device cannot support the training of the WLAN sensing model corresponding to the model ID indicated by training trigger message 1 (that is, the training device cannot support this training).

[0122] In this embodiment, when the management device receives training response message 1, if the training response message 1 indicates a successful response, it sets the state of the WLAN sensing model corresponding to the model ID indicated by the training response message 1 and the state of the dataset corresponding to the dataset ID indicated by the training response message 1 to the training state. If the training response message 1 indicates a failed response, it terminates the current training. If the model ID indicated by the training response message 1 is a newly added model ID in the local storage, it deletes the model ID from the local storage.

[0123] Step 403: The training device sends a model delivery message to the management device.

[0124] The message structure of the delivery message of the above model is shown in Table 3 below.

[0125] In this embodiment, when the training device determines that it can support this training, it obtains the dataset corresponding to the dataset ID from the local device based on the dataset ID indicated by the training trigger message 1. Then, it trains the WLAN perception model based on the target AI training algorithm indicated by the training trigger message 1 and the obtained dataset, and sends the trained WLAN perception model to the management device after the training is completed.

[0126] Based on this, as an example, if the size of the trained WLAN sensing model is greater than or equal to a set size threshold, the trained WLAN sensing model can be segmented according to the set segmentation size to obtain multiple model data. Each model data is then carried in the corresponding model delivery message and sent to the management device. After that, the management device can generate the trained WLAN sensing model by splicing the model data indicated by the received model delivery messages and store it.

[0127] If the size of the trained WLAN sensing model is smaller than the set size threshold, no segmentation is required. The trained WLAN sensing model can be directly included in the model delivery message and sent to the management device.

[0128] In this embodiment, after the management device has finished storing the WLAN sensing model indicated by the model delivery message, it will also set the state of the WLAN sensing model corresponding to the model ID indicated by the training trigger message 1 to an idle state, and set the state of the dataset corresponding to the dataset ID indicated by the training trigger message 1 to an idle state when the dataset is not used by any other model other than the WLAN sensing model indicated by the training trigger message 1.

[0129] Table 1: Training Trigger Messages

[0130]

[0131]

[0132] Table 2: Training Response Message 1

[0133]

[0134] Table 3: Model Delivery Messages

[0135]

[0136]

[0137] In this embodiment, as an example, for each WLAN sensing model, if it is necessary to change some training parameters of the WLAN sensing model, such as the dataset, the offline training update process of the WLAN sensing model can be triggered. This offline training update process is the same as described above. Figure 4 The process shown is similar, so it will not be repeated here.

[0138] In this embodiment, training termination can be initiated automatically by the training device after training is completed, or it can be actively initiated by the management device, such as when the management device detects a training termination event. It should be noted that the specific method for detecting a training termination event is similar to the implementation of detecting a training trigger event described above, and will not be repeated here.

[0139] For example, as an embodiment, the management device sends a training termination message to the training device. The message structure of the training termination message is shown in Table 4 below.

[0140] When the training device receives the training termination message, it terminates the training of the WLAN sensing model corresponding to the model ID indicated by the training termination message, and then sends a training termination response message to the management device in response to the training termination message.

[0141] The message structure of the training termination response message is shown in Table 5 below. "Success" in Table 5 indicates that the training device successfully terminated the training of the corresponding WLAN sensing model; "Failure" in Table 5 indicates that the training device did not successfully terminate the training of the corresponding WLAN sensing model.

[0142] When the management device receives a training termination response message, if the training termination response message indicates success, it sets the state of the WLAN sensing model corresponding to the model ID indicated by the training termination response message to the idle state, and sets the state of the dataset used by the WLAN sensing model to the idle state when the dataset is not used by other models other than the WLAN sensing model.

[0143] Table 4: Training Termination Messages

[0144]

[0145]

[0146] Table 5: Training Termination Response Messages

[0147]

[0148] The following is based on Figure 3 The system shown illustrates the online training process as an example:

[0149] As an example, such as Figure 5 As shown, the online training process may include the following steps:

[0150] Step 501: The management device sends a training trigger message 2 to the training device.

[0151] The message structure of training trigger message 2 is shown in Table 6 below. "Single report" in Table 6 refers to reporting once for each wireless channel data point collected; "batch report" in Table 6 refers to reporting once for every N wireless channel data points collected, where N can be flexibly set based on actual needs, and N is greater than 1.

[0152] The data format in Table 6 is not specifically limited here and can be flexibly set based on actual application needs. For example, it can be the raw data format of wireless channel data, or it can be other data formats. It should be noted that if the data format used is not the raw data format of wireless channel data, the wireless channel data can be preprocessed based on the preprocessing algorithm agreed upon with the data acquisition equipment before reporting to obtain wireless channel data with the required format, and then reported.

[0153] Step 502: When the training device receives the training trigger message 2, it sends a training response message 2 to the management device in response to the training trigger message 2.

[0154] The message structure of training response message 2 is similar to that in Table 2 above.

[0155] Step 503: When the management device receives the training response message 2, if the training response message 2 indicates success, it sends a collection trigger message 1 to all data acquisition devices indicated by the training trigger message 2.

[0156] The message structure of the collection trigger message 1 is shown in Table 7 below.

[0157] In this embodiment, if the training response message 2 indicates success, the management device sets the state of the WLAN sensing model indicated by the training trigger message 2 to the training state, and sets the states of the dataset and data acquisition device indicated by the training trigger message 2 to the acquisition state; otherwise, if the model ID and dataset ID indicated by the training trigger message 2 are newly added model IDs in the local storage, the management device deletes the model ID and dataset ID from the local storage, and sends a training termination message to the training device to instruct the training device to terminate the online training of the WLAN sensing model indicated by the training termination message.

[0158] Step 504: When the data acquisition device receives the acquisition trigger message 1, it sends an acquisition response message 1 to the management device in response to the acquisition trigger message 1.

[0159] The message structure of acquisition response message 1 is shown in Table 8 below. A "success" message in Table 8 indicates that the data acquisition device can support the acquisition of wireless channel data indicated by the acquisition trigger message; a "failure" message in Table 8 indicates that the data acquisition device cannot support the acquisition of wireless channel data indicated by the acquisition trigger message.

[0160] In this embodiment, when the management device receives the acquisition response message 1, if the acquisition response message 1 indicates success, it sets the state of the WLAN sensing model indicated by the training trigger message 2 to the training state, and sets the state of the dataset and the data acquisition device indicated by the training trigger message 2 to the acquisition state.

[0161] When the management device discovers that all data acquisition devices indicated by training trigger message 2 have returned acquisition response messages 1 indicating failure, and if the model ID and dataset ID indicated by training trigger message 2 are newly added model IDs in the local storage, the management device deletes the model ID and dataset ID from the local storage and sends a training termination message to the training device to instruct the training device to terminate the online training of the WLAN sensing model indicated by the training termination message.

[0162] Step 505: The data acquisition device sends a data reporting message to the training device.

[0163] In this embodiment, after determining that it can support the collection of wireless channel data indicated by the collection trigger message, the data acquisition device will configure itself accordingly based on the configuration parameters of the collection trigger message indication, and then start collecting wireless channel data. The acquired wireless channel data will be processed based on the data format indicated by the collection trigger message to generate a corresponding data reporting message and send it to the training device.

[0164] The message structure of the data reporting message is shown in Table 9 below.

[0165] Step 506: The training device sends a model delivery message to the management device.

[0166] In this embodiment, when the training device receives a data reporting message, it checks whether the target information indicated by the data reporting message matches the target information indicated by the training trigger message 2. If so, it trains a WLAN perception model based on the wireless channel data indicated by the data reporting message and the AI ​​training algorithm indicated by the training trigger message 2. After training is completed (e.g., when it determines that the preset training stop condition is met, or when it receives a training termination message from the management device), the training device sends a model delivery message to the management device based on the trained WLAN perception model, so that the management device stores the trained WLAN perception model indicated by the model delivery message.

[0167] Optionally, the target information mentioned above may include, for example, dataset ID, label, data format, MAC address, etc. There are no specific limitations on this, and it can be flexibly set according to actual needs.

[0168] The target information indicated by the above data reporting message matches the target information indicated by training trigger message 2, which can be understood as the target information indicated by the above data reporting message being the same as the target information indicated by training trigger message 2.

[0169] Here, the specific message structure of the model delivery message can be found in Table 3 above.

[0170] In this embodiment, as an example, when the training device receives a data reporting message, it will also store the wireless channel data indicated by the data reporting message into the storage space corresponding to the dataset indicated by the data reporting message.

[0171] Step 507: After receiving the model delivery message, the management device sends a data acquisition termination message to all data acquisition devices indicated by the training trigger message 2.

[0172] In this embodiment, when the management device receives the model delivery message, it stores the trained WLAN sensing model indicated by the model payment message.

[0173] The message structure for the termination message is shown in Table 10 below.

[0174] Step 508: When the data acquisition device receives the acquisition termination message, it terminates the acquisition of wireless channel data that matches the model ID indicated by the acquisition termination message, and returns an acquisition termination response message to the management device in response to the acquisition termination message.

[0175] It should be noted that if the data acquisition termination response message indicates success, then the data acquisition termination response message can be considered as the aforementioned data acquisition termination completion message.

[0176] The message structure for the termination response message is shown in Table 11 below.

[0177] In this embodiment, when the management device receives the acquisition termination response message, if the acquisition termination response message indicates success, it sets the state of the WLAN sensing model indicated by the training trigger message 2 to the idle state, and sets the state of the dataset and the data acquisition device to the idle state when the dataset indicated by the training trigger message 2 and the data acquisition device are not used by any other model other than the WLAN sensing model indicated by the training trigger message 2.

[0178] Table 6: Training Trigger Message 2

[0179]

[0180]

[0181] Table 7: Collection Trigger Message 1

[0182]

[0183]

[0184] Table 8: Collection Response Message 1

[0185]

[0186] Table 9: Data Reporting Messages

[0187]

[0188]

[0189] Table 10: Data Acquisition Termination Message

[0190]

[0191] Table 11: Collection Termination Response Message

[0192]

[0193] In this embodiment, as an example, for each WLAN sensing model, if it is necessary to change some training parameters of the WLAN sensing model, such as the dataset, the online training update process of the WLAN sensing model can be triggered. This online training update process is the same as described above. Figure 5 The process shown is similar, so it will not be repeated here.

[0194] In this embodiment, as an example, training termination can be automatically initiated by the training device when it determines that the preset training stop conditions are met, or training termination can be initiated by the management device. For example, when the management device receives a training termination command issued by the user through the WLAN sensing APP, it will initiate training termination. The training termination command includes at least the model ID of the WLAN sensing model whose training is to be terminated.

[0195] For example, as an embodiment, the management device sends a training termination message to the training device, and the message structure of the training termination message is shown in Table 4 above.

[0196] When the training device receives a training termination message, it terminates the training of the WLAN sensing model corresponding to the model ID indicated in the training termination message, and then sends a training termination response message to the management device in response to the training termination message. The message structure of the training termination response message is shown in Table 5 above.

[0197] When the management device receives a training termination response message, if the training termination response message indicates success, it sends a data acquisition termination message to all data acquisition devices used by the WLAN sensing model corresponding to the model ID indicated in the training termination response message. The specific implementation of the data acquisition devices after receiving the data acquisition termination message can be found in the relevant description in step 508 above, and will not be repeated here.

[0198] See Figure 6 , Figure 6 This is a schematic diagram of the structure of a WLAN sensing system provided in an embodiment of this application. Figure 6 As shown, the WLAN sensing system 600 includes: a management device 601, a training device 602, a data acquisition device 603, and a storage device 604. Here, the storage device can be used to store the acquired offline dataset and the trained WLAN sensing model.

[0199] Optionally, the management device 601, training device 602, data acquisition device 603, and storage device 604 can be the same device or different devices; there is no specific limitation here.

[0200] The following is based on Figure 6 The system shown illustrates the offline training process as an example:

[0201] As an example, such as Figure 7 As shown, the offline training process may include the following steps:

[0202] Step 701: The management device sends a training trigger message 3 to the training device.

[0203] The message structure of training trigger message 3 is shown in Table 1 above.

[0204] Step 702: When the training device receives the training trigger message 3, it sends a training response message 3 to the management device in response to the training trigger message 3.

[0205] In this embodiment, when the training device receives the training trigger message 3, it determines the training response message 3 based on the remaining available computing resources in the device. For details on how this is determined, please refer to the relevant description above.

[0206] The message structure of training response message 3 is shown in Table 2 above.

[0207] Step 703: When the management device receives the training response message 3, if the training response message 3 indicates success, it sends a training trigger message 4 to the storage device.

[0208] In this embodiment, when the management device receives training response message 3, if the training response message 3 indicates failure, it terminates the current training. If the model ID indicated by the training response message 3 is a newly added model ID in the local storage, it deletes the model ID from the local storage.

[0209] The message structure of training trigger message 4 is shown in Table 12 below.

[0210] Step 704: When the storage device receives the training trigger message 4, it determines whether it can obtain the dataset corresponding to the dataset ID from the local storage device based on the dataset ID indicated by the training trigger message 4, and returns a training response message 4 to the management device in response to the training trigger message 4.

[0211] The specific message structure of training response message 4 can be found in Table 2 above. If training response message 4 indicates success, it means that the dataset corresponding to the dataset ID indicated by training trigger message 4 has been successfully obtained from the local machine; if training response message 4 indicates failure, it means that the dataset corresponding to the dataset ID indicated by training trigger message 4 has not been successfully obtained from the local machine.

[0212] In this embodiment, when the management device receives the training response message 4, if the training response message 4 indicates success, it sets the status of the WLAN sensing model and dataset corresponding to the model ID and dataset ID indicated by the training trigger message 3 to the training state; otherwise, it terminates the current training and deletes the model ID from the local machine if the model ID indicated by the training trigger message 3 is a newly added model ID.

[0213] Step 705: After the storage device successfully obtains the dataset corresponding to the dataset ID from the local storage device based on the dataset ID indicated by the training trigger message 4, it sends a dataset delivery message to the training device.

[0214] In this embodiment, the dataset delivery message may include at least: a model ID field, a dataset ID field, and a dataset load field, wherein the dataset load field is used to indicate the obtained dataset.

[0215] Alternatively, since the dataset is large, the acquired dataset can be sent to the training device via techniques such as Remote Direct Memory Access (RDMA), which is not specifically limited here.

[0216] Step 706: When the training device receives the dataset delivery message, it trains the WLAN perception model based on the dataset indicated by the dataset delivery message and the AI ​​training algorithm indicated by the training response message 3. After the training is completed, it sends a model delivery message to the storage device and a training completion message to the management device.

[0217] The message structure of the training completion message is shown in Table 13 below.

[0218] In this embodiment, when the storage device receives the model delivery message, it stores the trained WLAN sensing model indicated by the model delivery message.

[0219] When the management device receives the training completion message, it sets the state of the WLAN sensing model indicated by training trigger message 3 to the idle state, and sets the state of the dataset indicated by training trigger message 3 to the idle state when it is not used by any other model besides the WLAN sensing model indicated by training trigger message 3.

[0220] Table 12: Training Trigger Message 3

[0221]

[0222] Table 13: Training Completion Messages

[0223] type 2 Indicates TLV type length 2 Indicates the payload length of the message modelID 1 Indicator Model ID Same as above

[0224] It should be noted that, similar to the offline training process without storage devices described above, the offline training and update process for WLAN perception models with storage devices is also similar to the above. Figure 7 The process shown is similar and will not be repeated here. For training termination with storage devices, please refer to the above explanation of training termination without storage devices; it will not be repeated here.

[0225] The following is based on Figure 6The system shown illustrates the online training process as an example:

[0226] As an example, such as Figure 8 As shown, the online training process may include the following steps:

[0227] Step 801: The management device sends a training trigger message 5 to the training device.

[0228] Here, the specific message structure of training trigger message 5 can be found in Table 6 above.

[0229] Step 802: When the training device receives the training trigger message 5, it sends a training response message 5 to the management device in response to the training trigger message 5.

[0230] Here, the specific message structure of training response message 5 can be found in Table 7 above.

[0231] Step 803: When the management device receives the training response message 5, if the training response message 5 indicates success, it sends a training trigger message 6 to all data acquisition devices indicated by the training trigger message 5.

[0232] Here, the specific message structure of training trigger message 6 can be found in Table 13 above.

[0233] In this embodiment, when the management device receives training response message 5, if the training response message 5 indicates failure, it terminates the current training. If the model ID and dataset ID indicated by the training response message 5 are the newly added model ID and dataset ID in the local machine, it deletes the model ID and dataset ID from the local machine.

[0234] Step 804: When the storage device receives the training trigger message 6, it returns a training response message 6 to the management device in response to the training trigger message 6.

[0235] In this embodiment, the specific message structure of training response message 6 is similar to that of training response message 4 described above, and will not be repeated here. A successful training response message 6 indicates that the storage device can support the storage of the WLAN sensing model and dataset indicated by training trigger message 6; a failed training response message 6 indicates that the storage device cannot support the storage of the WLAN sensing model and dataset indicated by training trigger message 6.

[0236] For example, as an embodiment, when the storage device receives the training trigger message 6, it can determine whether the remaining available hard disk storage resources of the device are greater than or equal to a set storage threshold. If so, it determines that the storage device can support the storage of the WLAN sensing model and dataset indicated by the training trigger message 6; otherwise, it determines that the storage device cannot support the storage of the WLAN sensing model and dataset indicated by the training trigger message 6.

[0237] Step 805: When the management device receives the training response message 6, if the training response message 6 indicates success, it sends the acquisition trigger message 2 to all data acquisition devices indicated by the training trigger message 5.

[0238] The specific message structure of collection trigger message 2 can be found in the above collection trigger message 1.

[0239] Step 806: When the data acquisition device receives the acquisition trigger message 2, it sends an acquisition response message 2 to the management device in response to the acquisition trigger message 2.

[0240] The specific message structure of the acquisition response message 2 can be found in the acquisition response message 1 above.

[0241] In this embodiment, when the management device receives the acquisition response message 2, if the acquisition response message 2 indicates success, it sets the state of the WLAN sensing model indicated by the training trigger message 5 to the training state, and sets the state of the dataset and the data acquisition device indicated by the training trigger message 5 to the acquisition state.

[0242] When the management device discovers that all data acquisition devices indicated by training trigger message 5 have returned acquisition response messages 2 indicating failure, and if the model ID and dataset ID indicated by training trigger message 5 are newly added model IDs in the local storage, it deletes the model ID and dataset ID from the local storage and sends a training termination message to the training device to instruct the training device to terminate the online training of the WLAN sensing model indicated by the training termination message.

[0243] Step 807: The data acquisition device sends a data reporting message to the training device.

[0244] Step 808: The training device sends a model delivery message to the storage device.

[0245] The specific implementation of this step is similar to step 506 above, and will not be repeated here.

[0246] Step 809: After the WLAN perception model training indicated by training trigger message 5 is completed, the training device sends a training completion message to the management device and a data acquisition termination message to all data acquisition devices indicated by training trigger message 5.

[0247] Step 810: When the data acquisition device receives the acquisition termination message, it terminates the acquisition of the wireless channel data indicated by the acquisition termination message and returns an acquisition termination response message to the management device in response to the acquisition termination message.

[0248] In this embodiment, when the management device receives the acquisition termination response message, if the acquisition termination response message indicates success, the state of the WLAN sensing model indicated by the training trigger message 5 is set to the idle state. When the data acquisition device and dataset indicated by the training trigger message 5 are not used by other models other than the WLAN sensing model indicated by the training trigger message 5, the state of both the data acquisition device and the dataset is set to the idle state.

[0249] It should be noted that, similar to the online training process without storage devices described above, the online training and update process for WLAN sensing models with storage devices is also similar to the above. Figure 8 The process shown is similar, so it will not be repeated here.

[0250] The message interaction in the above WLAN sensing method process is described in further detail below:

[0251] In this embodiment, as an example, for each set of request messages (such as the training trigger message, acquisition trigger message, training termination message, acquisition termination message, etc.) and response messages in the above WLAN sensing method process, a response timeout can be preset between the request messages and response messages based on actual application requirements. Based on this, a timer is started when a device sends a request message. If the timeout exceeds T... rt If no response message is received for the request message, the request message is resent and the timer is reset until the number of times the request message is sent reaches the set number (e.g., 3 times, 4 times, etc.). If no response message is received for the request message after that, the process is abandoned. The above-mentioned WLAN sensing APP can also output an alarm prompt indicating that the request failed, so that the user can take corresponding actions based on the prompt, such as re-issuing training instructions.

[0252] This concludes the description of the method provided in the embodiments of this application. The system and apparatus provided in the embodiments of this application will now be described:

[0253] As an example, this embodiment also provides a WLAN sensing device. For example, see... Figure 9 , Figure 9 This is a schematic diagram of a WLAN sensing device provided in an embodiment of this application. The device corresponds to… Figure 1 The process is shown below. Figure 9 As shown, the WLAN sensing device 900 is used as a management device in a WLAN sensing system. The WLAN sensing device 900 includes:

[0254] The first sending module 901 is used to send a first training trigger message to the training device during the offline training process of the target perception model. The first training trigger message includes at least: a model ID field, a dataset ID field, and an algorithm field. The model ID field is used to indicate the model ID of the target perception model. The dataset ID field is used to indicate the dataset ID of the target dataset to be used in this training. The algorithm field is used to indicate the target AI training algorithm to be used in this training.

[0255] The first receiving module 902 is used to receive a first training response message returned by the training device; the first training response message includes at least an ACK field; the ACK field indicates whether the response is successful or failed, a successful response indicates that the training device can support the training of the target perception model, and a failed response indicates that the training device cannot support the training of the target perception model.

[0256] The first receiving module 902 is further configured to receive a model delivery message sent by the training device and store the trained target perception model indicated by the model delivery message; the trained target perception model is obtained by the training device based on the target AI training algorithm and the target dataset; the trained target perception model is used for WLAN perception processing.

[0257] As an example, the first receiving module 902 is further configured to set the state of both the target perception model and the target dataset to the training state if the first training response message indicates a successful response; otherwise, if the model ID indicated by the first training trigger message is a newly added model ID in the local machine, the model ID is deleted from the local machine.

[0258] After receiving the model delivery message, the state of the target perception model is set to idle state, and the state of the target dataset is set to idle state when the target dataset is not used by any other model besides the target perception model.

[0259] As one embodiment, the WLAN sensing system further includes a storage device;

[0260] The first sending module 901 is further configured to send a second training trigger message to the training device during the offline training process of the target perception model; the second training trigger message includes at least: a model ID field, a dataset ID field, an algorithm field, and a storage address field; the storage address field is used to indicate the address information of the storage device;

[0261] The first receiving module 902 is further configured to receive a second training response message returned by the training device, the second training response message including at least an ACK field; if the second training response message indicates a successful response, a third training trigger message is sent to the storage device, so that the storage device obtains the target dataset based on the third training trigger message and sends a first data reporting message to the training device; the third training trigger message including at least a dataset ID field and a training address field; the training address field is used to indicate the address information of the training device; the first data reporting message including at least a dataset ID field and a first data payload field; the first data payload field is used to indicate the target dataset;

[0262] The first receiving module 902 is further configured to receive a training completion message sent by the training device; the training completion message is sent by the training device to the management device after the training device has sent the trained target perception model to the storage device for storage.

[0263] As an example, the first receiving module 902 is further configured to receive a third training response message returned by the storage device; the third training response message indicates whether the response is successful or failed, wherein a successful response indicates that the storage device can successfully obtain the target dataset, and a failed response indicates that the storage device cannot successfully obtain the target dataset.

[0264] If the third training response message indicates a successful response, then the states of both the target perception model and the target dataset are set to the training state; if the second training response message indicates a failed response, or if the third training response message indicates a failed response, then if the model ID indicated by the second training trigger message is a newly added model ID in the local machine, then that model ID is deleted from the local machine.

[0265] Upon receiving the training completion message, the state of the target perception model is set to an idle state, and the state of the target dataset is set to an idle state when the target dataset is not used by any other model besides the target perception model.

[0266] As one embodiment, the WLAN sensing system further includes a data acquisition device;

[0267] The first sending module 901 is further configured to send a fourth training trigger message to the training device and receive a fourth training response message returned by the training device during the online training process of the target perception model; the fourth training trigger message includes at least: a model ID field, an algorithm field, and a label field; the label field is used to indicate the label of the wireless channel data to be collected in this training; the fourth training response message includes at least: the ACK field;

[0268] The first sending module 901 is further configured to, if the fourth training response message indicates a successful response, send a collection trigger message to the data acquisition device to be used in this training, so that the data acquisition device collects wireless channel data based on the collection trigger message and sends a second data reporting message to the training device; wherein, the collection trigger message includes at least: a tag field, a collection configuration field, and a training address field; the second data reporting message includes at least: a tag field and a second data payload field; the second data payload field is used to indicate the wireless channel data; the collection configuration field is used to indicate the configuration parameters for the data acquisition device to perform data acquisition;

[0269] The first receiving module 902 is further configured to receive a model delivery message sent by the training device and store the trained target perception model indicated by the model delivery message; wherein the trained target perception model is obtained by the training device training the target perception model based on the target AI training algorithm and the wireless channel data.

[0270] As an example, the first receiving module 902 is further configured to receive a collection response message returned by the data acquisition device; the collection response message indicates whether the response is successful or failed, the successful response indicates that the data acquisition device can support the collection of the wireless channel data, and the failed response indicates that the training device cannot support the training of the target perception model;

[0271] If the acquisition response message indicates a successful response, the state of the target perception model is set to the training state, and the states of the target dataset and the data acquisition device are both set to the acquisition state; otherwise, if the model ID and dataset ID indicated by the fourth training trigger message are newly added model IDs in the local machine, the model ID and dataset ID are deleted from the local machine, and a training termination message is sent to the training device to instruct the training device to terminate the online training of the target perception model.

[0272] As an example, the first receiving module 902 is further configured to delete the model ID and the dataset ID from the local machine if the fourth training response message indicates that the response has failed.

[0273] After receiving the model delivery message, a data acquisition termination message is sent to the data acquisition device; and upon receiving the data acquisition termination completion message returned by the data acquisition device, the state of the target perception model is set to idle state, and when the target dataset and the data acquisition device are not used by any other model besides the target perception model, the states of both the target dataset and the data acquisition device are set to idle state.

[0274] As one embodiment, the WLAN sensing system further includes a storage device;

[0275] The first sending module 901 is further configured to send a fifth training trigger message to the training device and receive a fifth training response message returned by the training device during the online training process of the target perception model; the fifth training trigger message includes at least: a model ID field, an algorithm field, a label field, and a storage address field; the fifth training response message includes at least: the ACK field;

[0276] The first sending module 901 is further configured to send the acquisition trigger message to the data acquisition device if the fifth training response message indicates a successful response, so that the data acquisition device can acquire wireless channel data based on the acquisition trigger message and send the second data reporting message to the training device.

[0277] The first receiving module 902 is further configured to receive a training completion message sent by the training device; the training completion message is sent by the training device to the management device after the training device has sent the trained target perception model to the storage device for storage.

[0278] As an example, the fifth training trigger message, the acquisition trigger message, and the second data reporting message all further include: a dataset ID field;

[0279] After the target perception model is trained, the training device takes the wireless channel data received during the training process as data belonging to the target dataset and sends it to the storage device for storage.

[0280] As an example, the first receiving module 902 is further configured to delete the model ID and the dataset ID from the local machine if the fifth training response message indicates that the response has failed.

[0281] After receiving the training completion message, a data acquisition termination message is sent to the data acquisition device. Upon receiving the data acquisition termination completion message, the state of the target perception model is set to an idle state. When the target dataset and the data acquisition device are not used by any other model besides the target perception model, the states of both the target dataset and the data acquisition device are set to an idle state.

[0282] This concludes the process. Figure 9 Structural description of the device shown.

[0283] As an example, this embodiment also provides a WLAN sensing device. For example, see... Figure 10 , Figure 10 This is a schematic diagram of a WLAN sensing device provided in an embodiment of this application. The device corresponds to… Figure 2 The process is shown below. Figure 10 As shown, the WLAN sensing device 1000 is used in a training device within a WLAN sensing system. The WLAN sensing device 1000 includes:

[0284] The second receiving module 1001 is used to receive a first training trigger message sent by the management device during the offline training process of the target perception model; the first training trigger message includes at least: a model ID field, a dataset ID field, and an algorithm field; the model ID field is used to indicate the model ID of the target perception model; the dataset ID field is used to indicate the dataset ID of the target dataset to be used in this training; the algorithm field is used to indicate the target AI training algorithm to be used in this training.

[0285] The second sending module 1002 is used to return a first training response message to the management device; the first training response message includes at least an ACK field; the ACK field indicates whether the response is successful or failed, the successful response indicates that the training device can support the training of the target perception model, and the failed response indicates that the training device cannot support the training of the target perception model;

[0286] The second sending module 1002 is further configured to send a model delivery message to the management device, the model delivery message being used to indicate a trained target perception model; the trained target perception model is obtained by the training device training the target perception model based on a target AI training algorithm and a target dataset; the trained target perception model is used for WLAN perception processing.

[0287] As one embodiment, the WLAN sensing system further includes a storage device;

[0288] The second receiving module 1001 is further configured to receive a second training trigger message issued by the management device during the offline training process of the target perception model; the second training trigger message includes at least: a model ID field, a dataset ID field, an algorithm field, and a storage address field; the storage address field is used to indicate the address information of the storage device;

[0289] The second sending module 1002 is further configured to return a second training response message to the management device, the second training response message including at least an ACK field; so that when the second training response message indicates a successful response, the management device sends a third training trigger message to the storage device, instructing the storage device to obtain the target dataset based on the third training trigger message, and send a first data reporting message to the training device; the third training trigger message including at least a dataset ID field and a training address field; the training address field is used to indicate the address information of the training device; the first data reporting message including at least a dataset ID field and a first data payload field; the first data payload field is used to indicate the target dataset;

[0290] The second sending module 1002 is further configured to send a training completion message to the management device; the training completion message is sent by the training device after it has sent the trained target perception model to the storage device for storage.

[0291] As one embodiment, the WLAN sensing system further includes a data acquisition device;

[0292] The second receiving module 1001 is further configured to receive a fourth training trigger message issued by the management device during the online training process of the target perception model; the fourth training trigger message includes at least: a model ID field, an algorithm field, and a tag field; the tag field is used to indicate the tag of the wireless channel data to be collected in this training; the fourth training response message includes at least: an ACK field;

[0293] The second sending module 1002 is further configured to return a fourth training response message to the management device, the fourth training response message including at least an ACK field; so that when the fourth training response message indicates a successful response, the management device sends a collection trigger message to the data acquisition device to be used in this training, instructing the data acquisition device to collect wireless channel data based on the collection trigger message, and send a second data reporting message to the training device; wherein, the collection trigger message includes at least a tag field, a collection configuration field, and a training address field; the second data reporting message includes at least a tag field and a second data payload field; the second data payload field is used to indicate the wireless channel data; the collection configuration field is used to indicate the configuration parameters for the data acquisition device to perform data acquisition;

[0294] The second sending module 1002 is further configured to send a model delivery message to the management device, the model delivery message being used to indicate a trained target perception model; wherein the trained target perception model is obtained by the training device training the target perception model based on the target AI training algorithm and the wireless channel data.

[0295] As one embodiment, the WLAN sensing system further includes a storage device; the method further includes:

[0296] The second receiving module 1001 is also used to receive a fifth training trigger message issued by the management device during the online training process of the target perception model; the fifth training trigger message includes at least: a model ID field, an algorithm field, a label field, and a storage address field;

[0297] The second sending module 1002 is further configured to send a fifth training response message to the management device; the fifth training response message includes at least an ACK field; so that when the management device indicates a successful response in the fifth training response message, it sends the acquisition trigger message to the data acquisition device, so that the data acquisition device acquires wireless channel data based on the acquisition trigger message and sends the second data reporting message to the training device.

[0298] The second sending module 1002 is further configured to send a training completion message to the management device; the training completion message is sent by the training device after it has sent the trained target perception model to the storage device for storage.

[0299] This concludes the process. Figure 10 Structural description of the device shown.

[0300] The specific implementation process of the functions and roles of each device in the above-mentioned apparatus can be found in the implementation process of the corresponding steps in the above-mentioned method, and will not be repeated here.

[0301] For the apparatus embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. Those skilled in the art can understand and implement these embodiments without any inventive effort.

[0302] Based on the same application concept as the above method, this application provides a WLAN sensing system, which includes a management device and a training device;

[0303] The management device is used to send a first training trigger message to the training device during the offline training of the target perception model. The first training trigger message includes at least: a model ID field, a dataset ID field, and an algorithm field. The model ID field is used to indicate the model ID of the target perception model. The dataset ID field is used to indicate the dataset ID of the target dataset to be used in this training. The algorithm field is used to indicate the target AI training algorithm to be used in this training.

[0304] The training device is configured to receive a first training trigger message issued by the management device and return a first training response message to the management device; the first training response message includes at least an ACK field; the ACK field indicates whether the response is successful or failed, the successful response indicates that the training device can support the training of the target perception model, and the failed response indicates that the training device cannot support the training of the target perception model;

[0305] The training device is also used to send a model delivery message to the management device, the model delivery message being used to indicate the trained target perception model; the trained target perception model is obtained by the training device training the target perception model based on the target AI training algorithm and the target dataset; the trained target perception model is used for WLAN perception processing;

[0306] The management device is also used to receive the model delivery message sent by the training device and store the trained target perception model indicated by the model delivery message.

[0307] It should be noted that, since the system implementation corresponds to the above method implementation, the offline training process, including the storage device, and the online training process, etc., can be referred to the above relevant descriptions, and will not be repeated here.

[0308] This completes the structural description of the aforementioned system.

[0309] Please see Figure 11This is a schematic diagram of the hardware structure of an electronic device provided as an exemplary embodiment of this application. The electronic device includes a processor and a computer-readable storage medium; the computer-readable storage medium stores a plurality of computer program instructions, which, when executed by the processor, implement the method disclosed in the above example of this application. Depending on the actual function of the electronic device, other hardware may also be included, which will not be elaborated further.

[0310] Based on the same concept as the above method, this application also provides a computer-readable storage medium storing a plurality of computer program instructions, which, when executed by a processor, can implement the method disclosed in the above examples of this application.

[0311] For example, the aforementioned computer-readable storage medium can be any electronic, magnetic, optical, or other physical storage device that can contain or store information, such as executable instructions, messages, etc. For instance, computer-readable storage media can be: RAM (Random Access Memory), volatile memory, non-volatile memory, flash memory, storage drives (such as hard disk drives), solid-state drives, any type of storage disk (such as optical discs, DVDs, etc.), or similar storage media, or combinations thereof.

[0312] The above are merely preferred embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A wireless local area network (WLAN) sensing method, characterized in that, This method is applied to a management device in a WLAN sensing system, which further includes a training device; the method includes: During the offline training of the target perception model, a first training trigger message is sent to the training device. The first training trigger message includes at least: a model ID field, a dataset ID field, and an algorithm field. The model ID field is used to indicate the model ID of the target perception model. The dataset ID field is used to indicate the dataset ID of the target dataset to be used in this training. The algorithm field is used to indicate the target artificial intelligence (AI) training algorithm to be used in this training. The system receives a first training response message returned by the training device. The first training response message includes at least an ACK field. The ACK field indicates whether the response is successful or unsuccessful. A successful response indicates that the training device can support the training of the target perception model, and a unsuccessful response indicates that the training device cannot support the training of the target perception model. The system receives a model delivery message sent by the training device and stores the trained target perception model indicated by the model delivery message. The trained target perception model is obtained by the training device training the target perception model based on the target AI training algorithm and the target dataset. The trained target perception model is used for WLAN perception processing. The WLAN sensing system further includes a storage device; the method further includes: During the offline training of the target perception model, a second training trigger message is sent to the training device; the second training trigger message includes at least: a model ID field, a dataset ID field, an algorithm field, and a storage address field; the storage address field is used to indicate the address information of the storage device; The system receives a second training response message returned by the training device, the second training response message including at least an ACK field; if the second training response message indicates a successful response, a third training trigger message is sent to the storage device, so that the storage device obtains the target dataset based on the third training trigger message and sends a first data reporting message to the training device; the third training trigger message including at least a dataset ID field and a training address field; the training address field is used to indicate the address information of the training device; the first data reporting message including at least a dataset ID field and a first data payload field; the first data payload field is used to indicate the target dataset; The system receives a training completion message from the training device; the training completion message is sent by the training device to the management device after the training device has sent the trained target perception model to the storage device for storage.

2. The method according to claim 1, characterized in that, The method further includes: If the first training response message indicates a successful response, then the state of both the target perception model and the target dataset is set to the training state; otherwise, if the model ID indicated by the first training trigger message is a newly added model ID in the local machine, then the model ID is deleted from the local machine. After receiving the model delivery message, the state of the target perception model is set to idle state, and the state of the target dataset is set to idle state when the target dataset is not used by any other model besides the target perception model.

3. The method according to claim 1, characterized in that, The method further includes: Receive a third training response message returned by the storage device; the third training response message indicates whether the response is successful or failed, the successful response indicates that the storage device can successfully obtain the target dataset, and the failed response indicates that the storage device cannot successfully obtain the target dataset; If the third training response message indicates a successful response, then the states of both the target perception model and the target dataset are set to the training state; if the second training response message indicates a failed response, or if the third training response message indicates a failed response, then if the model ID indicated by the second training trigger message is a newly added model ID in the local machine, then that model ID is deleted from the local machine. Upon receiving the training completion message, the state of the target perception model is set to an idle state, and the state of the target dataset is set to an idle state when the target dataset is not used by any other model besides the target perception model.

4. The method according to claim 1, characterized in that, The WLAN sensing system further includes a data acquisition device; the method further includes: During the online training of the target perception model, a fourth training trigger message is sent to the training device, and a fourth training response message is received from the training device. The fourth training trigger message includes at least: a model ID field, an algorithm field, and a label field. The label field is used to indicate the label of the wireless channel data to be collected in this training. The fourth training response message includes at least: the ACK field. If the fourth training response message indicates a successful response, a collection trigger message is sent to the data acquisition device to be used in this training, so that the data acquisition device collects wireless channel data based on the collection trigger message and sends a second data reporting message to the training device; wherein, the collection trigger message includes at least: a tag field, a collection configuration field, and a training address field; the second data reporting message includes at least: a tag field and a second data payload field; the second data payload field is used to indicate the wireless channel data; the collection configuration field is used to indicate the configuration parameters for the data acquisition device to perform data acquisition; The system receives a model delivery message sent by the training device and stores the trained target perception model indicated by the model delivery message; wherein the trained target perception model is obtained by the training device training the target perception model based on the target AI training algorithm and the wireless channel data.

5. The method according to claim 4, characterized in that, The method further includes: Receive a collection response message returned by the data acquisition device; the collection response message indicates whether the response is successful or failed, the successful response indicates that the data acquisition device can support the collection of the wireless channel data, and the failed response indicates that the training device cannot support the training of the target perception model; If the acquisition response message indicates a successful response, the state of the target perception model is set to the training state, and the states of the target dataset and the data acquisition device are both set to the acquisition state; otherwise, if the model ID and dataset ID indicated by the fourth training trigger message are newly added model IDs in the local machine, the model ID and dataset ID are deleted from the local machine, and a training termination message is sent to the training device to instruct the training device to terminate the online training of the target perception model.

6. The method according to claim 5, characterized in that, The method further includes: If the fourth training response message indicates a failure, then if the model ID and dataset ID indicated by the fourth training trigger message are newly added model IDs in the local machine, then the model ID and dataset ID will be deleted from the local machine. After receiving the model delivery message, a data acquisition termination message is sent to the data acquisition device; and upon receiving the data acquisition termination completion message returned by the data acquisition device, the state of the target perception model is set to idle state, and when the target dataset and the data acquisition device are not used by any other model besides the target perception model, the states of both the target dataset and the data acquisition device are set to idle state.

7. The method according to claim 4, characterized in that, The WLAN sensing system further includes a storage device; the method further includes: During the online training of the target perception model, a fifth training trigger message is sent to the training device, and a fifth training response message is received from the training device; the fifth training trigger message includes at least: a model ID field, an algorithm field, a label field, and a storage address field; the fifth training response message includes at least: the ACK field. If the fifth training response message indicates a successful response, the acquisition trigger message is sent to the data acquisition device so that the data acquisition device can acquire wireless channel data based on the acquisition trigger message and send the second data reporting message to the training device. The system receives a training completion message from the training device; the training completion message is sent by the training device to the management device after the training device has sent the trained target perception model to the storage device for storage.

8. The method according to claim 7, characterized in that, The fifth training trigger message, the acquisition trigger message, and the second data reporting message all further include: a dataset ID field; After the target perception model is trained, the training device takes the wireless channel data received during the training process as data belonging to the target dataset and sends it to the storage device for storage.

9. The method according to claim 8, characterized in that, The method further includes: If the fifth training response message indicates a failure, then if the model ID and dataset ID indicated by the fifth training trigger message are newly added model IDs in the local machine, then delete the model ID and dataset ID from the local machine. After receiving the training completion message, a data acquisition termination message is sent to the data acquisition device. Upon receiving the data acquisition termination completion message, the state of the target perception model is set to an idle state. When the target dataset and the data acquisition device are not used by any other model besides the target perception model, the states of both the target dataset and the data acquisition device are set to an idle state.

10. A wireless local area network (WLAN) sensing method, characterized in that, This method is applied to a training device in a WLAN sensing system, which further includes a management device; the method includes: During the offline training of the target perception model, a first training trigger message is received from the management device. The first training trigger message includes at least: a model ID field, a dataset ID field, and an algorithm field. The model ID field is used to indicate the model ID of the target perception model. The dataset ID field is used to indicate the dataset ID of the target dataset to be used in this training. The algorithm field is used to indicate the target artificial intelligence (AI) training algorithm to be used in this training. The first training response message is returned to the management device; the first training response message includes at least: an ACK field; the ACK field indicates whether the response is successful or failed, the successful response indicates that the training device can support the training of the target perception model, and the failed response indicates that the training device cannot support the training of the target perception model; A model delivery message is sent to the management device, the model delivery message indicating a trained target perception model; the trained target perception model is obtained by the training device training the target perception model based on a target AI training algorithm and a target dataset; the trained target perception model is used for WLAN perception processing; The WLAN sensing system further includes a storage device; the method further includes: During the offline training of the target perception model, a second training trigger message is received from the management device; the second training trigger message includes at least: a model ID field, a dataset ID field, an algorithm field, and a storage address field; the storage address field is used to indicate the address information of the storage device. The management device returns a second training response message, which includes at least an ACK field. When the second training response message indicates a successful response, the management device sends a third training trigger message to the storage device, instructing the storage device to obtain the target dataset based on the third training trigger message and send a first data reporting message to the training device. The third training trigger message includes at least a dataset ID field and a training address field. The training address field indicates the address information of the training device. The first data reporting message includes at least a dataset ID field and a first data payload field. The first data payload field indicates the target dataset. The training device sends a training completion message to the management device; the training completion message is sent by the training device after it has sent the trained target perception model to the storage device for storage.

11. The method according to claim 10, characterized in that, The WLAN sensing system further includes a data acquisition device; the method further includes: During the online training of the target perception model, a fourth training trigger message is received from the management device; the fourth training trigger message includes at least: a model ID field, an algorithm field, and a label field; the label field is used to indicate the label of the wireless channel data to be collected in this training; the fourth training response message includes at least: an ACK field; The management device returns a fourth training response message, which includes at least an ACK field. When the fourth training response message indicates a successful response, the management device sends a collection trigger message to the data acquisition device to be used in this training, instructing the data acquisition device to collect wireless channel data based on the collection trigger message, and sends a second data reporting message to the training device. The collection trigger message includes at least a tag field, a collection configuration field, and a training address field. The second data reporting message includes at least a tag field and a second data payload field. The second data payload field indicates the wireless channel data. The collection configuration field indicates the configuration parameters for data acquisition by the data acquisition device. A model delivery message is sent to the management device, the model delivery message indicating the trained target perception model; wherein the trained target perception model is obtained by the training device training the target perception model based on the target AI training algorithm and the wireless channel data.

12. The method according to claim 11, characterized in that, The WLAN sensing system further includes a storage device; the method further includes: During the online training of the target perception model, a fifth training trigger message is received from the management device; the fifth training trigger message includes at least: a model ID field, an algorithm field, a label field, and a storage address field; Send a fifth training response message to the management device; the fifth training response message includes at least an ACK field; so that when the fifth training response message indicates a successful response, the management device sends the acquisition trigger message to the data acquisition device, so that the data acquisition device acquires wireless channel data based on the acquisition trigger message and sends the second data reporting message to the training device; The training device sends a training completion message to the management device; the training completion message is sent by the training device after it has sent the trained target perception model to the storage device for storage.

13. A wireless local area network (WLAN) sensing device, characterized in that, This device is used in the management equipment of a WLAN sensing system, which also includes a training device; the device includes: The first sending module is used to send a first training trigger message to the training device during the offline training process of the target perception model. The first training trigger message includes at least: a model identifier ID field, a dataset ID field, and an algorithm field. The model ID field is used to indicate the model ID of the target perception model. The dataset ID field is used to indicate the dataset ID of the target dataset to be used in this training. The algorithm field is used to indicate the target artificial intelligence AI training algorithm to be used in this training. A first receiving module is configured to receive a first training response message returned by the training device; the first training response message includes at least an ACK field; the ACK field indicates whether the response is successful or failed, the successful response indicates that the training device can support the training of the target perception model, and the failed response indicates that the training device cannot support the training of the target perception model; The first receiving module is further configured to receive a model delivery message sent by the training device and store the trained target perception model indicated by the model delivery message; the trained target perception model is obtained by the training device training the target perception model based on the target AI training algorithm and the target dataset; the trained target perception model is used for WLAN perception processing. The WLAN sensing system also includes a storage device; The first sending module is further configured to send a second training trigger message to the training device during the offline training process of the target perception model; the second training trigger message includes at least: a model ID field, a dataset ID field, an algorithm field, and a storage address field; the storage address field is used to indicate the address information of the storage device; The first receiving module is further configured to receive a second training response message returned by the training device, the second training response message including at least an ACK field; if the second training response message indicates a successful response, a third training trigger message is sent to the storage device, so that the storage device obtains the target dataset based on the third training trigger message and sends a first data reporting message to the training device; the third training trigger message including at least a dataset ID field and a training address field; the training address field is used to indicate the address information of the training device; the first data reporting message including at least a dataset ID field and a first data payload field; the first data payload field is used to indicate the target dataset; The first receiving module is further configured to receive a training completion message sent by the training device; the training completion message is sent by the training device to the management device after the training device has sent the trained target perception model to the storage device for storage.

14. A wireless local area network (WLAN) sensing device, characterized in that, This device is used in a training device within a WLAN sensing system, which also includes a management device; the device comprises: The second receiving module is used to receive a first training trigger message issued by the management device during the offline training process of the target perception model. The first training trigger message includes at least: a model identifier ID field, a dataset ID field, and an algorithm field. The model ID field is used to indicate the model ID of the target perception model. The dataset ID field is used to indicate the dataset ID of the target dataset to be used in this training. The algorithm field is used to indicate the target artificial intelligence AI training algorithm to be used in this training. The second sending module is used to return a first training response message to the management device; the first training response message includes at least an ACK field; the ACK field indicates whether the response is successful or failed, the successful response indicates that the training device can support the training of the target perception model, and the failed response indicates that the training device cannot support the training of the target perception model; The second sending module is further configured to send a model delivery message to the management device, the model delivery message being used to indicate a trained target perception model; the trained target perception model is obtained by the training device training the target perception model based on a target AI training algorithm and a target dataset; the trained target perception model is used for WLAN perception processing; The WLAN sensing system also includes a storage device; The second receiving module is further configured to receive a second training trigger message issued by the management device during the offline training process of the target perception model; the second training trigger message includes at least: a model ID field, a dataset ID field, an algorithm field, and a storage address field; the storage address field is used to indicate the address information of the storage device; The second sending module is further configured to return a second training response message to the management device, the second training response message including at least an ACK field; so that when the second training response message indicates a successful response, the management device sends a third training trigger message to the storage device, instructing the storage device to obtain the target dataset based on the third training trigger message, and send a first data reporting message to the training device; the third training trigger message including at least a dataset ID field and a training address field; the training address field is used to indicate the address information of the training device; the first data reporting message including at least a dataset ID field and a first data payload field; the first data payload field is used to indicate the target dataset; The second sending module is also used to send a training completion message to the management device; the training completion message is sent by the training device after it has sent the trained target perception model to the storage device for storage.

15. A wireless local area network (WLAN) sensing system, characterized in that, The WLAN sensing system includes: management devices, storage devices, and training devices; The management device is used to send a first training trigger message to the training device during the offline training process of the target perception model. The first training trigger message includes at least: a model identifier ID field, a dataset ID field, and an algorithm field. The model ID field is used to indicate the model ID of the target perception model. The dataset ID field is used to indicate the dataset ID of the target dataset to be used in this training. The algorithm field is used to indicate the target artificial intelligence AI training algorithm to be used in this training. The training device is configured to receive a first training trigger message issued by the management device and return a first training response message to the management device; the first training response message includes at least an ACK field; the ACK field indicates whether the response is successful or failed, the successful response indicates that the training device can support the training of the target perception model, and the failed response indicates that the training device cannot support the training of the target perception model; The training device is also used to send a model delivery message to the management device, the model delivery message being used to indicate a trained target perception model; the trained target perception model is obtained by the training device training the target perception model based on a target AI training algorithm and a target dataset; the trained target perception model is used for WLAN perception processing; The management device is also used to receive the model delivery message sent by the training device and store the trained target perception model indicated by the model delivery message. The management device is also used to send a second training trigger message to the training device during the offline training process of the target perception model; the second training trigger message includes at least: a model ID field, a dataset ID field, an algorithm field, and a storage address field; the storage address field is used to indicate the address information of the storage device; The training device is further configured to, during the offline training of the target perception model, receive a second training trigger message issued by the management device; return a second training response message to the management device, the second training response message including at least an ACK field; and, when the second training response message indicates a successful response, cause the management device to issue a third training trigger message to the storage device, instructing the storage device to obtain the target dataset based on the third training trigger message, and send a first data reporting message to the training device; the third training trigger message including at least a dataset ID field and a training address field; the training address field indicating the address information of the training device; and the first data reporting message including at least a dataset ID field and a first data payload field; the first data payload field indicating the target dataset. The management device is also configured to receive a second training response message returned by the training device; if the second training response message indicates a successful response, a third training trigger message is sent to the storage device. The storage device is configured to obtain the target dataset based on a third training trigger message and send a first data reporting message to the training device; the third training trigger message includes at least a dataset ID field and a training address field; the training address field is used to indicate the address information of the training device; the first data reporting message includes at least a dataset ID field and a first data payload field; the first data payload field is used to indicate the target dataset; The training device is also used to send a training completion message to the management device; the training completion message is sent by the training device after it has sent the trained target perception model to the storage device for storage; The management device is also used to receive a training completion message sent by the training device; the training completion message is sent by the training device to the management device after the training target perception model has been sent to the storage device for storage.

Images