Sensing model training method, user equipment, base station, and server

Abstract

Provided in embodiments of the present disclosure are a sensing model training method, a user equipment, a base station, and a server. The sensing model training method comprises: receiving a training data set sent by one or more base stations participating in sensing; and training a sensing model on the basis of the training data set.

Description

Method for training perception model, user equipment, base station and server

[0001] Cross-reference to Related Applications

[0002] The present disclosure claims priority to the Chinese patent application No. 202411116215.X, filed on August 14, 2024, and entitled “Method for training perception model, user equipment, base station and server”, the entire content of which is incorporated herein by reference. TECHNICAL FIELD

[0003] The present disclosure relates to the field of communications, and more particularly, to a method for training a perception model, a user equipment, a base station and a server. BACKGROUND

[0004] The principle of the traditional communication perception method is that the user equipment or the base station obtains the perception target detection result through the detection of the perception signal. However, the traditional communication perception method often has a large false alarm or a large missed detection, for example, when there are multiple different types of perception targets, the perception success rate is often not high. Therefore, it is necessary to provide a more effective way for implementing target perception in a mobile communication network, such as target identification, trajectory tracking, etc.

[0005] In addition, the existing mobile communication network can provide perception / positioning services. However, in the existing mobile communication network, there is no special network element to provide perception / positioning services, and therefore it is necessary to define a new network element for providing perception / positioning services, so as to improve the service range of the mobile communication network. SUMMARY

[0006] The embodiments of the present disclosure provide a method for training a perception model, a user equipment, a base station and a server, which are used to train a perception model to perceive a perception target.

[0007] According to a first aspect of the present disclosure, a method for training a perception model is provided. The method is applied to a perception training server, and the method comprises: receiving a training data set sent by one or more base stations participating in perception; and training a perception model based on the training data set.

[0008] In some embodiments, the method further comprises: receiving configuration information of a perception signal sent by the one or more base stations participating in perception, wherein the configuration information of the perception signal comprises one or more of sequence characteristics, frequency information, time domain information of the perception signal; and wherein the time domain information comprises one or more of the number of time slots, the number of symbols, silence characteristics, beam direction, and beam.

[0009] In some embodiments, the training data set comprises one or more of: time domain sampled signals of reflected signals and / or scattered signals of a perception signal, joint data of transformed reflected signals and / or scattered signals of a perception signal into frequency domain data or time domain data, joint data of transformed reflected signals and / or scattered signals of a perception signal into frequency domain data and time domain data, channel impulse response time domain data of reflected signals and / or scattered signals of a perception signal, channel impulse response frequency domain data of reflected signals and / or scattered signals of a perception signal; wherein the perception signal is a signal transmitted by the one or more participating base stations to the target area.

[0010] In some embodiments, the training data set comprises one or more of: signal strength, time delay, frequency offset, target features of the perception signal of reflected signals and / or scattered signals of a perception signal; the target features of the perception signal comprises one or more of: Doppler shift, signal time difference of arrival, angle information, channel impulse response, received power, perception signal transmission delay, perception signal reception delay; wherein the perception signal is a signal transmitted by the one or more participating base stations to the target area.

[0011] In some embodiments, the method further comprises: receiving labeling information transmitted by one or more participating user equipment, wherein the labeling information comprises one or more of: identification of the one or more user equipment, moving speed, time corresponding to the moving speed, position, time corresponding to the position, shape of the one or more user equipment, volume of the one or more user equipment, maximum cross-sectional area of the one or more user equipment.

[0012] In some embodiments, the method further comprises: training the perception model based on the training data set and the labeling information.

[0013] In some embodiments, the method further comprises: transmitting the perception model to a perception server; wherein the perception model comprises one or more of: model architecture, model loss function, model type, model input, model output.

[0014] According to a second aspect of the present disclosure, a method for training a perception model is provided. Wherein the method is applied to a base station, the method comprises: transmitting a perception signal to a target area, and receiving reflected signals and / or scattered signals of the perception signal; generating a training data set based on one or more of: the perception signal, the reflected signals of the perception signal, the scattered signals of the perception signal; transmitting the training data set to a perception training server.

[0015] In some embodiments, the method further comprises: sending, to the perception training server, configuration information of the perception signal, wherein the configuration information of the perception signal comprises one or more of sequence features of the perception signal, frequency information, time domain information, and the time domain information comprises one or more of a number of time slots, a number of symbols, silence features, beam directions, and beams.

[0016] In some embodiments, the training data set comprises one or more of time domain sampled signals of the reflection signal and / or the scattering signal of the perception signal, joint data of the reflection signal and / or the scattering signal of the perception signal transformed into frequency domain data or time domain data, joint data of the reflection signal and / or the scattering signal of the perception signal transformed into frequency domain data and time domain data, channel impulse response time domain data of the reflection signal and / or the scattering signal of the perception signal, and channel impulse response frequency domain data of the reflection signal and / or the scattering signal of the perception signal.

[0017] In some embodiments, the training data set further comprises one or more of signal strength, time delay, frequency offset, target features of the perception signal of the reflection signal and / or the scattering signal of the perception signal; wherein the target features of the perception signal comprises one or more of Doppler shift, signal time difference of arrival, angle information, channel impulse response, received power, perception signal sending delay, and perception signal receiving delay.

[0018] According to a third aspect of the present disclosure, a method for training a perception model is provided. The method is applied to a user equipment, and the method comprises: sending, to a perception training server, labeling information for training a perception model; the labeling information comprises one or more of an identity of the user equipment, a moving speed, a time corresponding to the moving speed, a position, a time corresponding to the position, a form of the user equipment, a volume of the user equipment, and a maximum cross-sectional area of the user equipment.

[0019] In some embodiments, the perception model comprises one or more of a model architecture, a model loss function, a model type, a model input, and a model output.

[0020] According to a fourth aspect of the present disclosure, a method for obtaining a perception model is provided. The method is applied to a perception server, and the method comprises:

[0021] receiving a perception model sent by a perception training server, wherein the perception model comprises one or more of a model architecture, a model loss function, a model type, a model input, and a model output.

[0022] In some embodiments, the perception model is trained based on a training data set sent by one or more base stations participating in perception and labeled information sent by one or more user equipment participating in perception.

[0023] In some embodiments, the training data set comprises one or more of time domain sampled signals of reflected signals and / or scattered signals of a perception signal, joint data of reflected signals and / or scattered signals of the perception signal transformed into frequency domain data or time domain data, joint data of reflected signals and / or scattered signals of the perception signal transformed into frequency domain data and time domain data, channel impulse response time domain data of reflected signals and / or scattered signals of the perception signal, channel impulse response frequency domain data of reflected signals and / or scattered signals of the perception signal; wherein the perception signal is a signal sent by the one or more base stations participating in perception to a target area.

[0024] In some embodiments, the training data set further comprises one or more of signal strength, time delay, frequency offset, target feature of the perception signal of reflected signals and / or scattered signals of the perception signal; wherein the target feature of the perception signal comprises one or more of Doppler shift, signal time difference of arrival, angle information, channel impulse response, received power, perception signal sending delay, perception signal receiving delay; wherein the perception signal is a signal sent by the one or more base stations participating in perception to a target area.

[0025] In some embodiments, the labeled information comprises one or more of an identity of the user equipment, a moving speed, a time corresponding to the moving speed, a position, a time corresponding to the position, a form of the user equipment, a volume of the user equipment, a maximum cross-sectional area of the user equipment.

[0026] According to a fifth aspect of the present disclosure, a perception training server is provided. The perception training server comprises: a receiving module configured to receive a training data set sent by one or more base stations participating in perception; and a training module configured to train a perception model based on the training data set.

[0027] According to a sixth aspect of the present disclosure, a base station is provided. The base station comprises: a sending module configured to send a perception signal to a target area; a receiving module configured to receive reflected signals and / or scattered signals of the perception signal; and a processing module configured to generate a training data set based on one or more of the perception signal, the reflected signals of the perception signal, and the scattered signals of the perception signal; and the sending module is further configured to send the training data set to a perception training server.

[0028] According to a seventh aspect of the present disclosure, a user equipment is provided. The user equipment comprises: a sending module configured to send, to a perception training server, annotation information used for training a perception model, wherein the annotation information comprises one or more of an identity of the user equipment, a moving speed, a time corresponding to the moving speed, a position, a time corresponding to the position, a form of the user equipment, a volume of the user equipment, and a maximum cross-sectional area of the user equipment.

[0029] According to an eighth aspect of the present disclosure, a perception server is provided. The perception server comprises: a receiving module configured to receive a perception model sent by a perception training server, wherein the perception model comprises one or more of a model architecture, a model loss function, a model type, a model input, and a model output.

[0030] According to a ninth aspect of the present disclosure, a server is provided. The server comprises: a memory and a processor; the memory is configured to store program instructions; and the processor is configured to invoke the program instructions stored in the memory, and execute any method described above based on the program instructions.

[0031] According to a tenth aspect of the present disclosure, a base station is provided. The base station comprises: a memory and a processor; the memory is configured to store program instructions; and the processor is configured to invoke the program instructions stored in the memory, and execute any method described above based on the program instructions.

[0032] According to an eleventh aspect of the present disclosure, a user equipment is provided. The user equipment comprises: a memory and a processor; the memory is configured to store program instructions; and the processor is configured to invoke the program instructions stored in the memory, and execute any method described above based on the program instructions.

[0033] According to a twelfth aspect of the present disclosure, a computer readable storage medium is provided. The computer readable storage medium stores computer instructions, and when the computer instructions are run on a computer, the computer is caused to execute any method described above.

[0034] The method for training a sensing model provided in this disclosure can directly obtain reasonable annotation information from the user equipment, reducing the workload of manual annotation and improving computational efficiency and accuracy. Furthermore, this method can directly process the training dataset and / or configuration information of the sensing signals sent by the base station to obtain the final sensing model, eliminating the need for intermediate feature extraction, reducing intermediate steps, and improving processing efficiency. Moreover, the trained sensing model is applicable to different environmental conditions and channel variations in communication sensing networks, exhibiting higher generalization ability compared to traditional sensing detection methods. During the training of the sensing model, multiple base stations simultaneously share and collaboratively process sensing data, thereby improving the final sensing accuracy. Attached Figure Description

[0035] To more clearly illustrate the technical solutions of the embodiments of this disclosure, the accompanying drawings used in the embodiments of this disclosure will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this disclosure. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0036] Figure 1 illustrates an exemplary block diagram of a perception architecture 100 according to some embodiments of the present disclosure;

[0037] Figure 2 is a flowchart illustrating a positioning method flow 200 according to some embodiments of the present disclosure;

[0038] Figure 3 is a flowchart illustrating a method flow 300 for training a perception model according to some embodiments of the present disclosure;

[0039] Figure 4 is a flowchart illustrating a method by which a perception training server can be used to train a perception model according to some embodiments of the present disclosure;

[0040] Figure 5 is a flowchart illustrating a method for training a sensing model using a base station according to some embodiments of the present disclosure;

[0041] Figure 6 is a flowchart of a method for training a perception model using a user device according to some embodiments of the present disclosure;

[0042] Figure 7 is an exemplary block diagram illustrating a perception architecture 700 according to some embodiments of the present disclosure;

[0043] Figure 8 is a schematic diagram illustrating the structure of a user equipment 800 provided according to some embodiments of the present disclosure;

[0044] Figure 9 is a schematic diagram showing the structure of a base station 900 provided according to some embodiments of the present disclosure;

[0045] FIG. 10 is a structural schematic diagram of a perception training server 1000 according to some embodiments of the present disclosure;

[0046] FIG. 11 is a structural schematic diagram of a perception server 1100 according to some embodiments of the present disclosure;

[0047] FIG. 12 is a structural schematic diagram of a computer device according to some embodiments of the present disclosure. DETAILED DESCRIPTION

[0048] In order to make the purposes, technical solutions and advantages of the present disclosure clearer, the technical solutions in the embodiments of the present disclosure will be described clearly and completely below with reference to the drawings in the embodiments of the present disclosure. Obviously, the described embodiments are only part of the embodiments of the present disclosure, not all the embodiments. Based on the embodiments in the present disclosure, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of protection of the present disclosure.

[0049] Some terms in the embodiments of the present disclosure are explained below to facilitate understanding by those of ordinary skill in the art.

[0050] (1) In the embodiments of the present disclosure, the nouns "network" and "system" are often used alternately, but those of ordinary skill in the art can understand their meanings.

[0051] (2) In the embodiments of the present disclosure, the term "multiple" refers to two or more, and other quantifiers are similar.

[0052] (3) "And / or", which describes the association relationship of the associated objects, means that there can be three relationships, for example, A and / or B can represent: A exists alone, A and B exist together, and B exists alone. The character " / " generally represents that the associated objects before and after it are in an "or" relationship.

[0053] The embodiments of the present disclosure provide a method, a user equipment, a base station and a server for training a perception model. The technical solutions provided by the embodiments of the present disclosure can be applied to various systems. For example, the systems to which the technical solutions can be applied can be a long term evolution (LTE) system, an LTE frequency division duplex (FDD) system, an LTE time division duplex (TDD) system, a long term evolution advanced (LTE-A) system, a universal mobile system (UMTS), a worldwide interoperability for microwave access (WiMAX) system, a 5G new radio (NR) system and an evolved communication system thereof, a 6G (sixth generation mobile communication technology) system, and the like.

[0054] The method, the user equipment, the base station and the server for training a perception model provided by the embodiments of the present disclosure can directly obtain reasonable labeling information from the user equipment, thereby reducing the workload of manual labeling and improving the calculation efficiency and accuracy. In addition, the method can directly process based on the training data set and / or the configuration information of the perception signal sent by the network equipment, thereby obtaining the final perception model, without intermediate feature extraction, thereby reducing the intermediate steps and improving the processing efficiency. Furthermore, the trained perception model can be applied to different environmental conditions and channel changes in a communication perception network, and has higher generalization ability compared with the traditional perception detection method. In the process of training the perception model, multiple base stations are used to share and cooperatively process the perception data, thereby improving the final perception accuracy.

[0055] The method and the device are based on the same application concept. Since the principles of the method and the device for solving problems are similar, the implementation of the device and the method can be referred to each other, and the repeated parts will not be described again.

[0056] FIG. 1 illustrates an exemplary block diagram of a perception architecture 100 according to some embodiments of the present disclosure. As shown in FIG. 1, the perception architecture 100 includes a user equipment 101, a base station 102, a server 103 and / or any other suitable component that can be used for perception in a mobile communication network.

[0057] In some embodiments, the user equipment 101 can be a device that provides voice and / or data connectivity to a user, a handheld device having wireless connection capability, or other processing devices connected to a wireless modem, etc. In different systems, the name of the user equipment can also be different, for example, in a 5G system or a 6G system, the wireless terminal device can be a USB storage device, other personal computer memory devices and dongles, and can also communicate with one or more core networks (CN) through a radio access network (RAN). The wireless terminal device can be a mobile terminal device, such as a mobile phone (or called "cellular" phone) and a computer with a mobile terminal device, for example, it can be a portable, pocket, handheld, computer built-in or vehicle-mounted mobile device, which exchanges voice and / or data with a radio access network. For example, personal communication service (PCS) phones, cordless phones, session initiated protocol (SIP) phones, wireless local loop (WLL) stations, personal digital assistants (PDAs), personal computers, tablet computers, machine-type communication (MTC) terminal devices, etc. The wireless terminal device can also be called system, subscriber unit, subscriber station, mobile station, mobile, remote station, access point, remote terminal, access terminal, user terminal, user agent, user device, and wireless access router / modem that meets the definition limit, etc. The embodiments of the present disclosure are not limited.

[0058] In some embodiments, the user equipment 101 can be used to send a perception reference signal, or perform perception based on auxiliary information. In some embodiments, the user equipment can be used to calculate the final result and accuracy of perception based on the perception result.

[0059] In some embodiments, the network device 102 can be a base station. The base station can be a cell including multiple serving user equipment. According to different specific application scenarios, the base station can also be called an access point, or can be a device in an access network that communicates with wireless terminal equipment through one or more sectors over an air interface, or other names. The network device can be used to exchange the received air frame and Internet Protocol (IP) packet as a router between the wireless terminal equipment and the rest of the access network, which can include an Internet Protocol (IP) communication network. The network device can also coordinate the management of the properties of the air interface. For example, the network device involved in the embodiments of the present disclosure can be an evolved network device (eNB or e-NodeB) in a long term evolution (LTE) system, a 5G base station (gNB) in a 5G network architecture, etc., and can also be a home evolved base station (HeNB), a relay node, a home base station (femto), a pico base station, a network test device, etc., and the embodiments of the present disclosure are not limited. In some network structures, the network device can include a centralized unit (CU) node and a distributed unit (DU) node, and the centralized unit and the distributed unit can also be arranged geographically apart.

[0060] In some embodiments, taking the network device 102 as an example of a 5G wireless access network, the network device 102 can include one or more gNBs and ng-eNBs. In some embodiments, the gNB is configured to provide a node of NR (New Radio) user plane and control plane protocol terminal to user equipment, and is connected to the core network through the NG interface. The gNB can support frequency division duplex (FDD), time division duplex (TDD) and / or dual mode. In some embodiments, the ng-eNB can be used to be compatible with the 4G network in the 5G network. In some embodiments, the ng-eNB can provide a node of E-UTRA (Evolved Universal Terrestrial Radio Access) user plane and control plane protocol terminal to user equipment, and is connected to the core network through the NG interface.

[0061] In some embodiments, the network device 102 can include an NG interface. The NG interface can be an interface between a base station and a core network, which includes a control plane interface (NG-C) and a user plane interface (NG-U). In some embodiments, the network device 102 can include an Xn interface. The Xn interface can be a network interface between multiple base stations in a network, such as an interface between a gNB and an ng-eNB, which can be used for communication and cooperation between nodes.

[0062] In some embodiments, the network device 102 can transmit a sensing reference signal, or perform sensing measurement based on the assistance information.

[0063] In some embodiments, the server 103 can be a core network. The core network can be an evolved packet system (EPC), a 5G core network (5GC), or the like. Taking the 5G core network as an example, the core network can include network elements such as an access and mobility management function (AMF), a session management function (SMF), a user plane function (UPF), a policy control function (PCF), a network repository function (NRF), a unified data management (UDM), a location management function (LMF), and the like. The AMF can be used to manage the access and mobility process of the user equipment. The SMF can be used to manage the session of the user equipment. The UPF can be used to process the transmission and forwarding of user data. The PCF can be used to specify and control the policy in the network. The NRF can be used to manage the network resources in the network. The UDM can be used to manage the user data in the network. The LMF can be used to complete the positioning function of the user equipment. In some embodiments, the LMF can be deployed independently, or can be deployed in combination with other network elements.

[0064] In some embodiments, the LMF can interact with other network elements in the core network, such as the AMF, the SMF, the UPF, and the like, to collect and process information related to the location of the user equipment. In some embodiments, the LMF can select a positioning method and trigger corresponding positioning measurement, and calculate the final result and accuracy of the positioning.

[0065] FIG. 1 is only an example and does not limit the type of communication system and the number and type of devices included in the communication system. The network architecture and business scenarios described in the embodiments of the disclosure are used to illustrate the technical solutions provided by the embodiments of the disclosure, and do not limit the technical solutions provided by the embodiments of the disclosure. It is known to those skilled in the art that as the network architecture evolves and new business scenarios appear, the technical solutions provided by the embodiments of the disclosure are also applicable to similar technical problems.

[0066] FIG. 2 is a flow chart illustrating a positioning method flow 200 according to some embodiments of the present disclosure. As shown in the figure, the positioning method mainly includes the following steps:

[0067] 201: The LMF obtains the network device configured downlink positioning reference signal (DL-PRS) information through the base station request message or response message (Information Request / Response) between the LMF and the network device (e.g., base station, which can include gNB or TRP (Transmit-Receive Point)).

[0068] 202: The user equipment and the LMF interact the information related to the user equipment positioning protocol capability (LTE Positioning Protocol capability).

[0069] 203: After obtaining the user equipment positioning protocol capability, the LMF sends a positioning information request to the network device to request the user equipment to report its positioning measurement results and / or position estimation results, etc.

[0070] 204: After receiving the positioning information request, the network device configures the user equipment sounding reference signal (UE SRS) resource, and sends the UE SRS resource to the user equipment in 204a.

[0071] 205: The network device sends a location information response to the LMF, which includes the configuration information used by the user equipment.

[0072] 206: The network device sends an instruction to the user equipment to activate its SRS transmission.

[0073] 207: The LMF sends a measurement request to the network device, which can be used to obtain the accurate position information of the user equipment. The measurement request instructs the network device to measure the SRS of the user equipment and collect relevant positioning data.

[0074] 208: During the measurement process, the LMF can send assistance information to the user equipment through LPP, which can be used to help the user equipment optimize its SRS transmission and reception. The assistance information can include time synchronization parameters, frequency offset correction, etc.

[0075] 209: The user equipment and the network device can perform downlink positioning reference signal (DL-PRS) measurement and uplink SRS measurement, respectively. Among them, the DL-PRS measurement and the uplink SRS measurement can be selected according to the network environment and the positioning demand.

[0076] 210: The user equipment can determine the final result and accuracy of the positioning based on the measurement results, and provide the positioning information to the LMF through LPP. The LMF can provide the final positioning information to other nodes in the network for various application scenarios such as navigation, map service, etc.

[0077] 211: The network equipment can send a measurement response to the LMF through NRPPa, confirming that all positioning measurements and data processing work have been completed.

[0078] FIG. 3 is a flow chart illustrating a method 300 of training a perception model according to some embodiments of the present disclosure. As shown, the method of training the perception model includes the following steps.

[0079] 301: The perception training server receives a first perception training request. Wherein the first perception training request can come from other network elements of the core network, such as from the perception server, can also come from the base station or the user equipment, and can also be triggered by itself. In response to the first perception training request, the perception training server starts to train the perception model.

[0080] In some embodiments, the first perception training request can include training data set requirements, perception model architecture, perception model type, model loss function index requirements, expected output of the perception model, perception model performance index, perception model update frequency, and / or any other suitable parameters that can be used to train the perception model. In some embodiments, the training data set requirements can be used by the perception training server to adopt training data sets for different target types and scenarios to train the perception model. In some embodiments, the training data set requirements can include one or more of expected data set types, expected data set time length, coverage frequency, number of sampling points, resolution, signal-to-noise ratio, dynamic range, normalization standard, and expected training data features. Wherein the expected data set types can include one or more of training data sets for various different types and scenarios (e.g., channel types), whether the training data is time series data, frequency domain data, or time-frequency joint data, etc. The expected training data features can include one or more of signal strength, time delay, frequency offset, target features of the perception signal, etc. The target features of the perception signal include one or more of Doppler shift, signal arrival time difference, angle information, channel impulse response, received power, perception signal transmission delay, perception signal reception delay, etc.

[0081] In some embodiments, the perception model architecture can include one or more of a number of layers, a number of neurons per layer, a type of activation function, etc. In some embodiments, the model type can include one or more of a convolutional neural network (CNN), a recurrent neural network (RNN), a support vector machine (SVM), etc. In some embodiments, the desired output of the perception model can include one or more of distinguishing a class or attribute of a perception target, a reliability of a prediction, a high-level feature representation of data, an anomaly or outlier in perception data, a perception target type, a perception target speed, a perception target height, a perception target geographic location, etc. In some embodiments, the performance metric of the perception model can include one or more of an accuracy, a recall, an F1 score, a ROC curve, etc.

[0082] 302: The perception training server can determine one or more base stations in the network device to participate in the perception training. When the perception training server determines multiple base stations to participate in the perception training, the perception training server can perform fusion processing on the training data sets generated by the multiple base stations, thereby improving the perception accuracy.

[0083] 303: The perception training server can send a second perception training request to the network device. The second perception training request includes a training data set requirement. The training data set requirement can refer to the training data set requirement in step 301, and the repeated parts will not be described here. In some embodiments, the training data set requirement can be used by the network device to generate training data sets for different target types and scenarios.

[0084] In some embodiments, the second perception training request can also be sent to the network device by other network elements of the core network, such as AMF, LMF, etc.

[0085] 304: The network device transmits a perception signal. In some embodiments, the network device can directly transmit the perception signal. In some embodiments, the network device can configure the perception signal based on the second perception training request and then transmit the perception signal. In some embodiments, the configuration information of the perception signal can include one or more of a sequence feature of the perception signal, frequency information, time domain information, etc. For example, the configuration information of the perception signal can include one or more of a number of slots, a number of symbols, a muting feature, a beam method, a beam, etc.

[0086] In some embodiments, the network device can send a sensing signal to a target area, the target area being a range of areas covered by the network device or a specific direction. The target area includes a sensing environment and / or a sensing target. The sensing signal reflected or scattered by the sensing environment and / or the sensing target is a reflected signal and / or a scattered signal of the sensing signal.

[0087] 305: The network device can send configuration information of the sensing signal to the sensing training server.

[0088] 306: The UE sends the labeling information to the sensing training server. The UE can send the labeling information to the sensing training server in response to the first request information of the sensing training server. The labeling information can include one or more of an identifier of the user equipment, a moving speed, a time corresponding to the moving speed, a position, a time corresponding to the position, a form of the user equipment, a volume of the user equipment, a maximum cross-sectional area of the user equipment, etc.

[0089] 307: The network device can be configured to receive the reflected signal and / or the scattered signal of the sensing signal, process the reflected signal and / or the scattered signal, generate a training data set, and send the training data set to the sensing training server. In some embodiments, the training data set can include one or more of a time-domain sampled signal of the reflected signal and / or the scattered signal, joint data of frequency-domain data and time-domain data converted from the reflected signal and / or the scattered signal, channel impulse response time-domain data of the reflected signal and / or the scattered signal, and channel impulse response frequency-domain data of the reflected signal and / or the scattered signal.

[0090] In some embodiments, the network device can perform feature extraction on the reflected signal and / or the scattered signal to obtain the training data set. The training data set can include one or more of signal strength, time delay, frequency offset, target features of the sensing signal, etc. The target features of the sensing signal can include one or more of Doppler shift, signal arrival time difference, angle information, channel impulse response, received power, sensing signal sending delay, sensing signal receiving delay, etc.

[0091] 308: The perception training server trains the perception model. The perception training server trains the perception model based on the training dataset and the labeling information, in combination with the first perception training request. In some embodiments, the perception training server completes the training of the perception model when the perception model satisfies the first perception training request. In some embodiments, the parameters of the perception model can include one or more of a model architecture, a model loss function, a model type, a model input, a model output, etc. In some embodiments, the model architecture can include one or more of a number of layers, a number of neurons per layer, a type of activation function, etc. In some embodiments, the model loss function can include a difference between a predicted result of the perception model and an actual label, such as a cross-entropy loss, a mean squared error, etc. In some embodiments, the model type can include one or more of a convolutional neural network (CNN), a recurrent neural network (RNN), a support vector machine (SVM), etc. In some embodiments, the model input can include one or more of a time domain signal, frequency domain data, time-frequency joint data, or a feature vector requirement, etc. The feature vector requirement can include one or more of a signal strength, a time delay, a frequency offset, a Doppler shift, a reference signal time difference (RSTD), angle information, a channel impulse response, a received power, a perception signal transmission delay, a perception signal reception delay, etc. In some embodiments, the model output can include one or more of a size, a shape, a position, a direction, a speed, a distance, a confidence, an error range, etc. of a perception target.

[0092] 309: The perception training server can send the trained perception model and / or the perception model parameters to a core network for perception. The core network can include a perception server.

[0093] The method for training a perception model provided in this embodiment can directly obtain reasonable labeling information from user equipment, reducing the workload of manual labeling and improving computational efficiency and accuracy. In addition, this method can directly process based on the training dataset and / or the configuration information of the perception signal sent by the network equipment, thereby obtaining the final perception model, without intermediate feature extraction, reducing intermediate steps and improving processing efficiency. Furthermore, the trained perception model can be suitable for different environmental conditions and channel changes in a communication perception network, and has higher generalization ability compared to traditional perception detection methods. During the training of the perception model, multiple base stations are used to share and cooperatively process perception data, thereby improving the final perception accuracy.

[0094] FIG. 4 is a flowchart illustrating a method that a perception training server can use to train a perception model, according to some embodiments of the present disclosure. As shown, the flowchart can include:

[0095] S401: The perception training server receives a training dataset from one or more base stations participating in perception.

[0096] In some embodiments, the perception training server receives a first perception training request. The first perception training request can be used to instruct the perception training server to start training a perception model. The first perception training request can come from other network elements of the core network, such as from a perception server, from a base station or a user equipment, or can be triggered by itself.

[0097] The first perception training request can include a training dataset requirement, a perception model architecture, a perception model type, a model loss function indicator requirement, an expected output of the perception model, a perception model performance indicator, a perception model update frequency, and / or any other suitable parameters that can be used to train the perception model. Among them, the various parameters of the first perception training request can refer to the parameter information of the first perception training request described in combination with FIG. 3, and the repeated parts will not be described here.

[0098] The training dataset can include one or more of time domain sampling signals of reflected signals and / or scattered signals, joint data of the reflected signals and / or scattered signals transformed into frequency domain data and time domain data, channel impulse response time domain data of the reflected signals and / or scattered signals, and channel impulse response frequency domain data of the reflected signals and / or scattered signals. The training dataset can also include features extracted from the reflected signals and / or scattered signals. The training dataset can include one or more of signal strength, time delay, frequency offset, target features of the perception signals, etc. The target features of the perception signals can include one or more of Doppler shift, signal arrival time difference, angle information, channel impulse response, received power, perception signal transmission delay, perception signal reception delay, etc. The reflected signals and / or scattered signals are reflected signals and / or scattered signals received after one or more base stations transmit perception signals.

[0099] In some embodiments, the perception training server can also receive configuration information of the perception signals. The configuration information of the perception signals is generated by one or more base stations based on a second perception training request. The configuration information of the perception signals can include one or more of sequence features of the perception signals, frequency information, time domain information, etc. For example, the configuration information of the perception signals can include one or more of the number of slots, the number of symbols, muting features, beam methods, beams, etc.

[0100] S402: The perception training server trains a perception model based on the training dataset.

[0101] In some embodiments, the perception training server receives the annotation information from the user device. The annotation information can include one or more of an identity of the user device, a moving speed, a time corresponding to the moving speed, a location, a time corresponding to the location, a form of the user device, a volume of the user device, a maximum cross-sectional area of the user device, and the like.

[0102] The annotation information can correspond to the training data set transmitted by the one or more base stations. The training data set transmitted by the one or more base stations can correspond to one or more same user devices.

[0103] In some embodiments, the perception training server can train the perception model based on the training data set and the annotation information, so that the perception model meets the first perception training request. When the perception model meets the first perception training request, the perception training server completes the training of the perception model. In some embodiments, the parameters of the perception model include one or more of a model architecture, a model loss function, a model type, a model input, a model output, and the like.

[0104] After the perception training server completes the training of the perception model, the perception training server can transmit the trained perception model and / or the perception model parameters to the perception server for subsequent perception.

[0105] The method flow of training the perception model provided in the embodiments can directly process the training data set and / or the configuration information of the perception signal transmitted by the network device, so as to obtain the final perception model, without intermediate feature extraction, reducing the intermediate steps and improving the processing efficiency. Moreover, the trained perception model can be suitable for different environmental conditions and channel changes in the communication perception network, and has higher generalization ability compared with the traditional perception detection method.

[0106] FIG. 5 is a flowchart illustrating a method that a base station can use to train a perception model, according to some embodiments of the present disclosure. As shown in the figure, the method flowchart includes:

[0107] S501: The base station transmits a perception signal to a target area, and receives a reflected signal and / or a scattered signal of the perception signal.

[0108] In some embodiments, the base station can receive a second sensing training request. The second sensing training request can be sent by the core network, e.g., a sensing training server, other network element in the core network; the second sensing training request can also be triggered by the base station itself. The second sensing training request includes training dataset requirements. The training dataset requirements can include one or more of expected dataset type, expected dataset duration, coverage frequency, number of sampling points, resolution, signal-to-noise ratio, dynamic range, normalization standard, expected training data features, etc. The expected dataset type can include one or more of training dataset for various different types and scenarios (e.g., channel type), whether the training data is time series data, frequency domain data, or time-frequency joint data, etc. The expected training data features can include one or more of signal strength, time delay, frequency offset, target features of the sensing signal, etc. The target features of the sensing signal can include one or more of Doppler shift, signal arrival time difference, angle information, channel impulse response, received power, sensing signal transmission delay, sensing signal reception delay, etc.

[0109] In some embodiments, the base station can configure the sensing signal to be available for different types and scenarios based on the training dataset requirements. The configuration information of the sensing signal is generated by the base station based on the second sensing training request. The configuration information of the sensing signal can include one or more of sequence features of the sensing signal, frequency information, time domain information, etc. For example, the configuration information of the sensing signal can include one or more of number of slots, number of symbols, muting features, beam method, beam, etc. In some embodiments, the base station can also not need to configure the sensing signal based on the second sensing training request. That is, after the base station determines that it needs to participate in training the sensing model, the base station can directly transmit the sensing signal without configuring the sensing signal.

[0110] In some embodiments, the base station can transmit the sensing signal to a target area, which is a range of areas covered by the base station or a specific direction. The target area includes the sensing environment and / or the sensing target. The sensing signal reflected or scattered by the sensing environment and / or the sensing target after being reflected or scattered by the sensing environment and / or the sensing target is the reflected signal and / or the scattered signal of the sensing signal. The base station can further process or perform feature extraction on the reflected signal and / or the scattered signal.

[0111] S502: The base station can generate a training dataset based on one or more of the sensing signal, the reflected signal of the sensing signal, and the scattered signal of the sensing signal.

[0112] The training data set can be obtained based on further processing of the reflection signal and / or scattering signal by the base station. The training data set can include one or more of time domain sampled signals of the reflection signal and / or scattering signal, joint data of the reflection signal and / or scattering signal transformed into frequency domain data and time domain data, channel impulse response time domain data of the reflection signal and / or scattering signal, and channel impulse response frequency domain data of the reflection signal and / or scattering signal.

[0113] The training data set can also be obtained based on feature extraction of the reflection signal and / or scattering signal by the base station. The training data set can also include one or more of signal strength, time delay, frequency offset, target features of the perception signal, etc. The target features of the perception signal can include one or more of Doppler shift, signal arrival time difference, angle information, channel impulse response, received power, perception signal sending delay, perception signal receiving delay, etc. The reflection signal and / or scattering signal is the reflection signal and / or scattering signal received after one or more base stations transmit the perception signal.

[0114] S503: The base station can send the training data set to the perception training server. In some embodiments, the base station can send the training data set to the core network. The core network can be and / or include the perception server or LMF.

[0115] The method flow for training the perception model provided by the embodiments has higher generalization capability compared to traditional perception detection methods, and can be applied to different environmental conditions and channel changes in a communication perception network. In addition, during the training of the perception model, multiple base stations can share and cooperatively process perception data at the same time, thereby improving the final perception accuracy.

[0116] FIG. 6 is a method flow diagram of a method that a user equipment can use to train a perception model according to some embodiments of the present disclosure. As shown in the figure, the method flow includes:

[0117] S601: The user equipment can determine the labeling information.

[0118] The user equipment can be configured to receive first request information. The first request information can be used by the perception training server to request the labeling information from the user equipment. In some embodiments, the first request information can come from the core network, for example, a network element in the core network or the perception training server, and can be sent to the user equipment by the AMF or the LMF. In some embodiments, after determining to participate in training the perception model, the user equipment can actively send the labeling information to the core network, for example, the perception training server, without responding to the first request information to send again.

[0119] The user equipment determines the labeling information in response to the first request information. The labeling information can include one or more of an identity of the user equipment, a moving speed, a time corresponding to the moving speed, a location, a time corresponding to the location, a form of the user equipment, a volume of the user equipment, a maximum cross-sectional area of the user equipment, and the like.

[0120] S602: The user equipment sends the labeling information to the perception training server.

[0121] The user equipment can send the determined labeling information to the perception training server for training the perception model.

[0122] The method for training the perception model provided in the embodiments can directly obtain reasonable labeling information from the user equipment, reducing the workload of manual labeling and improving the calculation efficiency and accuracy.

[0123] Based on the same technical concept, the embodiments of the disclosure further provide a perception architecture. The perception architecture can implement all the method steps implemented in the foregoing method embodiments.

[0124] FIG. 7 is an exemplary block diagram illustrating a perception architecture 700 according to some embodiments of the disclosure. As shown, the perception architecture 700 can include a user equipment 701, a network equipment 702, a server 703, and / or any other suitable component that can be used for perception in a mobile communication network. The user equipment 701 can be and / or include the user equipment 101 described in connection with FIG. 1. The network equipment 702 can be and / or include the network equipment 102 described in connection with FIG. 1.

[0125] In some embodiments, the server 703 can be a core network. In some embodiments, the server 703 can include various network elements included in the server 103 described in connection with FIG. 1. In some embodiments, the server 703 further includes a perception training server 7031 and a perception server 7032. The perception training server 7031 and the perception server 7032 can interact through an AMF network element. The perception training server 7031 can be used to train a perception model, and provide the trained perception model to the perception server 7032. The perception server 7032 can use the perception model to perceive and measure a perception target. The perception server 7032 can feed back the perception result to the perception training server 7031 to update the iterative perception model, and the perception training server 7031 can send the updated iterative perception model to the perception server 7032 again for subsequent perception.

[0126] In some embodiments, the user equipment 701 can receive a first request information. The first request information can be used to instruct the user equipment 701 to send the labeling information. In some embodiments, the first request information can be from the perception training server 7031. In other embodiments, the user equipment 701 can also not receive the first request information and actively send the labeling information to the server, e.g., the perception training server 7031. In some embodiments, the user equipment 701 sends the labeling information to the perception training server 7031 in response to the first request information. The labeling information can include one or more of the identification of the user equipment, the moving speed, the time corresponding to the moving speed, the location, the time corresponding to the location, the form of the user equipment, the volume of the user equipment, the maximum cross-sectional area of the user equipment, etc.

[0127] In some embodiments, the network equipment 702 can receive a second perception training request. The second perception training request can be from the core network, e.g., the perception training server, the perception server, or other network elements of the core network. In other embodiments, the second perception training request can also be from other equipment, e.g., the user equipment. In other embodiments, the network equipment 702 can also autonomously trigger the generation of the training dataset after determining to participate in training the perception model.

[0128] The second perception training request includes the training dataset requirement. In some embodiments, the training dataset requirement can be used by the network equipment 702 to generate the training dataset for different target types and scenarios. In some embodiments, the training dataset requirement can include one or more of the expected dataset type, the expected dataset duration, the coverage frequency, the number of sampling points, the resolution, the signal-to-noise ratio, the dynamic range, the normalization standard, the expected training data features, etc. The expected dataset type can include one or more of the training dataset for various different types and scenarios (e.g., channel type), the training data being time series data, frequency domain data, or time-frequency joint data, etc. The expected training data features can include one or more of the signal strength, the time delay, the frequency offset, the target features of the perception signal. The target features of the perception signal can include one or more of the Doppler shift, the signal arrival time difference, the angle information, the channel impulse response, the received power, the perception signal sending delay, the perception signal receiving delay, etc.

[0129] In some embodiments, after determining to participate in training the perception model, the network device 702 can transmit a perception signal to a target area, which is a range of areas covered by the network device or a specific direction. The target area includes a perception environment and / or a perception target. The perception signal reflected or scattered by the perception environment and / or the perception target after being reflected or scattered by the perception environment and / or the perception target is a reflected signal and / or a scattered signal of the perception signal. The base station can further process or perform feature extraction on the reflected signal and / or the scattered signal.

[0130] In some embodiments, the network device 702 can configure configuration information of the perception signal based on the training data set requirement, configure the perception signal according to the configuration information of the perception signal, transmit the perception signal, and receive the reflected signal and / or the scattered signal of the perception signal. In some embodiments, the configuration information of the perception signal can include one or more of sequence characteristics, frequency information, time domain information, etc. of the perception signal. For example, the configuration information of the perception signal can include one or more of a number of slots, a number of symbols, muting characteristics, beam methods, beams, etc.

[0131] In some embodiments, the network device 702 can generate a training data set based on the reflected signal and / or the scattered signal. For example, the network device 702 can further process the reflected signal and / or the scattered signal to obtain the training data set. In some embodiments, the training data set can include one or more of time domain sampling signals of the reflected signal and / or the scattered signal, joint data of the reflected signal and / or the scattered signal transformed into frequency domain data and time domain data, channel impulse response time domain data of the reflected signal and / or the scattered signal, and channel impulse response frequency domain data of the reflected signal and / or the scattered signal.

[0132] In some embodiments, the network device 702 can perform feature extraction on the reflected signal and / or the scattered signal to obtain a training data set. The training data set can include one or more of signal strength, time delay, frequency offset, target characteristics of the perception signal, etc. The target characteristics of the perception signal can include one or more of Doppler shift, signal arrival time difference, angle information, channel impulse response, received power, perception signal transmission delay, perception signal reception delay, etc.

[0133] In some embodiments, the network device 702 can send the training dataset to the core network. The core network can be and / or include the perception training server 7031, the perception server 7032. For example, the network device 702 can send the training dataset to the perception training server 7031. In some embodiments, the network device 702 can also send the configuration information of the perception signal to the perception training server 7031.

[0134] In some embodiments, the network device 702 can include multiple base stations. The multiple base stations can generate training datasets based on the above steps for the same perception target respectively. Then, the multiple base stations can send the generated training datasets to the perception training server 7031. The perception training server 7031 can perform fusion processing on the training datasets generated by the multiple base stations, thereby improving the perception accuracy.

[0135] In some embodiments, the perception training server 7031 can receive a first perception training request to perform perception model training. For example, the perception training server 7031 can receive a first perception training request from other network elements of the core network to perform perception model training.

[0136] In some embodiments, the first perception training request can include a training dataset requirement, a perception model architecture, a perception model type, a model loss function indicator requirement, an expected output of the perception model, a perception model performance indicator, a perception model update frequency, and / or any other suitable component that can be used to train the perception model. In some embodiments, the training dataset requirement can indicate that the perception training server 7031 uses training datasets for different target types and scenarios to train the perception model. The training dataset requirement can refer to the training dataset requirement in the second perception training request, which is not described herein again.

[0137] In some embodiments, the perception model architecture can include one or more of a number of layers, a number of neurons in each layer, a type of activation function, and / or the like. In some embodiments, the model type can include one or more of a convolutional neural network (CNN), a recurrent neural network (RNN), a support vector machine (SVM), and / or the like. In some embodiments, the expected output of the perception model can include one or more of distinguishing the classes or attributes of the perception target, the reliability of the prediction, a high-level feature representation of the data, an anomaly or outlier in the perception data, a perception target type, a perception target speed, a perception target height, a perception target geographical location, and / or the like. In some embodiments, the performance indicator of the perception model can include one or more of an accuracy, a recall rate, an F1 score, a ROC curve, and / or the like.

[0138] In some embodiments, the perception training server 7031 can receive a training dataset from the network device 702. In some embodiments, the perception training server 7031 can receive configuration information of the perception signal from the network device 702.

[0139] In some embodiments, the perception training server 7031 can also receive the labeling information from the user device 701.

[0140] In some embodiments, the perception training server 7031 can train a perception model based on the training dataset, the labeling information, and the first perception training request. In some embodiments, the perception model can include one or more of a model architecture, a model loss function, a model type, a model input, a model output, and the like. In some embodiments, the model architecture can include one or more of a number of layers, a number of neurons per layer, a type of activation function, and the like. In some embodiments, the model loss function can include a difference between a predicted result of the perception model and an actual label, such as a cross-entropy loss, a mean squared error, and the like. In some embodiments, the model type can include one or more of a convolutional neural network (CNN), a recurrent neural network (RNN), a support vector machine (SVM), and the like. In some embodiments, the model input can include one or more of a time-domain signal, frequency-domain data, time-frequency joint data, or a feature vector requirement, and the like. The feature vector requirement can include one or more of a signal strength, a time delay, a frequency offset, a Doppler shift, a reference signal time difference (RSTD), angle information, a channel impulse response, a received power, a perception signal transmission delay, a perception signal reception delay, and the like. In some embodiments, the model output can include one or more of a size, a shape, a position, a direction, a speed, a distance, a confidence, an error range, and the like of a perception target.

[0141] In some embodiments, the perception training server 7031 can send the trained perception model and / or the perception model parameters to other network elements in the core network for perception after completing the training of the perception model. For example, the perception training server 7031 can receive second request information from the perception server 7032, and thus send the perception model and / or the perception model parameters to the perception server 7032. For another example, the perception training server 7031 can send the perception model and / or the perception model parameters to other network elements in the core network, such as an LMF.

[0142] In some embodiments, the perception training server 7031 can also perform iterative updates of the perception model according to a model update frequency, and send the iteratively updated perception model and / or the perception model parameters to the perception server 7032.

[0143] In some embodiments, the perception server 7032 can send second request information to the perception training server 7031 to obtain the trained perception model and / or the perception model parameters available for perceiving the perception target. The perception server 7032 can also receive the updated perception model and / or the perception model parameters sent by the perception training server based on the model update frequency.

[0144] It should be noted that the above perception architecture provided by the embodiments of the present disclosure can implement all the method steps achieved by the above method embodiments and achieve the same technical effects. Therefore, the same parts and beneficial effects of the method embodiments will not be described in detail herein.

[0145] Based on the same technical concept, the embodiments of the present disclosure also provide a user equipment (e.g., UE), which can implement the functions of the user equipment (e.g., UE) in the above embodiments.

[0146] FIG. 8 is a structural schematic diagram of a user equipment 800 provided by the embodiments of the present disclosure. In some embodiments, the user equipment 800 can be and / or include the user equipment 101 and / or 701 described in combination with FIG. 1 or FIG. 7. As shown in the figure, the user equipment can include a receiving module 801, a processing module 802, and can also include a sending module 803.

[0147] The receiving module 801 can be configured to receive first request information. The first request information can be used for a perception training server to request annotation information from a user equipment. In some embodiments, the first request information can be from a network element of a core network, such as a perception training server, a perception server, or an LMF, etc. In other embodiments, the receiving module 801 can not receive the first request information, and the processing module 802 and the sending module 803 can send the annotation information to the core network.

[0148] The processing module 802 can be configured to determine the annotation information. The annotation information can include one or more of an identifier of the user equipment, a moving speed, a time corresponding to the moving speed, a location, a time corresponding to the location, a form of the user equipment, a volume of the user equipment, a maximum cross-sectional area of the user equipment, etc.

[0149] The sending module 803 can be configured to send the annotation information to a network element in the core network. In some embodiments, the sending module 803 can send the annotation information to a perception training server, a perception server, or an LMF.

[0150] It should be noted that the above user equipment provided by the embodiments of the present disclosure can implement all method steps achieved by the above method embodiments, and achieve the same technical effects. Therefore, the same parts and beneficial effects of the method embodiments will not be described in detail.

[0151] Based on the same technical concept, the embodiments of the present disclosure also provide a network device, which can implement the functions of the network device in the foregoing embodiments.

[0152] FIG. 9 is a structural schematic diagram of a network device 900 provided by the embodiments of the present disclosure. In some embodiments, the network device 900 can be and / or include the network device 102 and / or 702 described in combination with FIG. 1 or FIG. 7. As shown in the figure, the network device can include a receiving module 901, a configuration module 902, a processing module 903, and can further include a sending module 904.

[0153] The receiving module 901 can be configured to receive a second perception training request. The second perception training request can refer to the second perception training request described in FIG. 2, and repeated parts will not be described herein. In some embodiments, the second perception training request can come from a core network, for example, a perception training server, a perception server or other network element of the core network. In other embodiments, the second perception training request can also come from other devices, for example, a user equipment. In other embodiments, the network device can also autonomously trigger the generation of a training data set after determining to participate in training a perception model.

[0154] The configuration module 902 can be configured to configure a perception signal according to the second perception training request. In some embodiments, the configuration module 902 can be configured to directly configure the perception signal without being based on the second perception training request.

[0155] The sending module 904 can send the perception signal to a target area, which is a range of areas covered by the network device or a specific direction. The target area includes a perception environment and / or a perception target. The signal reflected or scattered by the perception environment and / or the perception target after the perception signal passes through the perception environment and / or the perception target is a reflected signal and / or a scattered signal of the perception signal.

[0156] The processing module 903 can obtain a training data set based on the reflected signal and / or the scattered signal. The processing module 903 can process the reflected signal and / or the scattered signal to obtain the training data set. The processing module 903 can also extract features from the reflected signal and / or the scattered signal to obtain the training data set. The training data set can refer to the training data set described in FIG. 2, and repeated parts will not be described herein.

[0157] The sending module 904 can also send the training data set and the configuration information of the sensing signal to the core network. In some embodiments, the core network can be a specific network element in the core network, such as a sensing training server, a sensing server, or other network elements of the core network. The configuration information of the sensing signal can include one or more of sequence characteristics, frequency information, time domain information, etc. of the sensing signal. For example, the configuration information of the sensing signal can include one or more of the number of slots, the number of symbols, muting characteristics, beam methods, beams, etc.

[0158] It should be noted that the network device provided by the embodiments of the present disclosure can implement all the method steps achieved by the method embodiments and achieve the same technical effects. Therefore, the same parts and beneficial effects of the method embodiments will not be described in detail.

[0159] Based on the same technical concept, the embodiments of the present disclosure also provide a sensing training server, which can implement the functions of the sensing training server in the foregoing embodiments.

[0160] Referring to FIG. 10, it is a structural schematic diagram of a sensing training server 1000 provided by the embodiments of the present disclosure. In some embodiments, the sensing training server 1000 can be and / or include the sensing training server 7031 described in FIG. 7. As shown, the sensing training server 1000 can include a receiving module 1001, a training module 1002, and can also include a sending module 1003.

[0161] The receiving module 1001 can be configured to receive a first sensing training request, so that the sensing training server 1000 performs training of a sensing model. In some embodiments, the sensing training server 7031 can receive the first sensing training request from other network elements of the core network to perform training of the sensing model.

[0162] The first sensing training request can refer to the first sensing training request described in combination with FIG. 2. The repeated parts will not be described here. The receiving module 1001 can also be configured to receive a training data set and / or configuration information of a sensing signal from a network device, which can be used to train the sensing model. The receiving module 1001 can also be configured to receive labeling information from a user equipment, which can be used to train the sensing model. The receiving module 1001 can also be configured to receive a sensing result from a sensing training server, which can be used to update an iterative sensing model.

[0163] The training module 1002 can include a perception model. The training module 1002 can be configured to train the perception model based on one or more of the first perception training request, the training dataset, the labeling information, and the like. The training module 1002 can train the perception model such that the perception model satisfies the first perception training request. Based on the perception model satisfying the first perception training request, the training module 1002 can complete training of the perception model. The training module 1002 can further perform an update iteration of the perception model based on the model update frequency.

[0164] The sending module 1003 can send the trained perception model to the core network. For example, the sending module 1003 can send the trained perception model to the perception training server or the LMF or other network element in the core network. The sending module 1003 can further send the updated iteration of the perception model based on the model update frequency to the core network. The sending module 1003 can further send the first request information to the user equipment to obtain the labeling information.

[0165] It should be noted that the perception training server provided by the embodiments of the present disclosure can implement all the method steps achieved by the method embodiments, and achieve the same technical effects. Therefore, the same parts and beneficial effects of the method embodiments will not be described in detail.

[0166] Based on the same technical concept, the embodiments of the present disclosure further provide a perception server, which can implement the functions of the perception server in the foregoing embodiments.

[0167] Referring to FIG. 11, FIG. 11 is a structural schematic diagram of a perception server 1100 provided by the embodiments of the present disclosure. In some embodiments, the perception server 1100 can be and / or include the perception server 7032 described in combination with FIG. 7. As shown in the figure, the perception server 1100 can include a receiving module 1101, a perception module 1102. The perception server 1100 can further include a sending module 1103.

[0168] The receiving module 1101 can be configured to receive the trained perception model and / or the perception model parameters sent by the perception training server. The receiving module 1101 can further be configured to receive the updated iteration of the perception model based on the model update frequency sent by the perception training server and / or the perception model parameters. The receiving module 1101 can further receive the perception dataset from the network device, which can be used for perception of the perception target. In some embodiments, the perception dataset can refer to the training dataset described in combination with FIG. 2, and the repeated parts will not be described herein. The receiving module 1101 can further receive the perception dataset from a plurality of network devices.

[0169] The perception module 1102 can be used to perceive the perception target based on the perception dataset and the perception model, and generate a perception result. When the receiving module 1101 receives a plurality of perception datasets, the perception module 1102 can perform fusion processing on the plurality of perception datasets to improve the perception accuracy.

[0170] The sending module 1103 can send second request information to the perception training server to obtain a trained perception model and / or an updated iterative perception model. The sending module 1103 can also send the perception result to the perception training server, which can be used for the perception training server to update the perception model parameters based on the perception result. The sending module 1103 can also send the perception result to a demand side, which refers to a device or apparatus that needs the perception result of the perception target for further processing.

[0171] It should be noted that the perception server provided by the embodiments of the present disclosure can implement all the method steps implemented by the method embodiments, and achieve the same technical effects. Therefore, the same parts and beneficial effects of the method embodiments will not be described in detail.

[0172] It should be noted that the division of units in the embodiments of the present disclosure is illustrative, and is only a logical functional division. In actual implementation, another division mode can be used. In addition, each functional unit in each embodiment of the present disclosure can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit.

[0173] When the integrated unit is realized in the form of a software functional unit and sold or used as an independent product, it can be stored in a processor-readable storage medium. Based on this understanding, the technical solutions of the present disclosure, essentially or the part that contributes to the related art, or all or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor to execute all or part of the steps of the methods described in the various embodiments of the present disclosure.

[0174] Based on the same technical concept, the embodiments of the present disclosure also provide a computer device, which can realize the functions of the user equipment, the network equipment, the perception training server, or the perception server in the foregoing embodiments.

[0175] FIG. 12 illustrates a structural diagram of a computer device in the embodiments of the present disclosure. As shown in the figure, the computer device can include a processor 1201, a memory 1202, a transceiver 1203, and a bus interface 1204.

[0176] The processor 1201 is responsible for managing the bus interface and general processing, and the memory 1202 can store data used by the processor 1201 in performing operations. The transceiver 1203 is used to receive and send data under the control of the processor 1201.

[0177] The bus interface 1204 can include any number of interconnected buses and bridges, which link together various circuits such as the processor 1201 represented by one or more processors and the memory 1202 represented by the memory. The bus interface can also link various other circuits such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art, and therefore, will not be further described herein. The bus interface provides an interface. The processor 1201 is responsible for managing the bus interface and general processing, and the memory 1202 can store data used by the processor 1201 in performing operations.

[0178] The flow disclosed in the embodiments of the present disclosure can be applied in the processor 1201 or implemented by the processor 1201. In the implementation process, each step of the signal processing flow can be completed by the integrated logic circuit of the hardware in the processor 1201 or the instructions in the form of software. The processor 1201 can be a general-purpose processor, a digital signal processor, an application-specific integrated circuit, a field programmable gate array, or other programmable logic device, a discrete gate or transistor logic device, a discrete hardware component, which can implement or execute the disclosed methods, steps, and logic block diagrams in the embodiments of the present disclosure. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in conjunction with the embodiments of the present disclosure can be directly embodied as hardware processor execution, or executed by a combination of hardware and software modules in the processor. The software module can be located in a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, or other mature storage media in the art. The storage medium is located in the memory 1202, and the processor 1201 reads the information in the memory 1202 and combines the hardware to complete the steps of the signal processing flow.

[0179] In some embodiments, the processor 1201 is configured to read computer instructions from the memory 1202 and execute the functions implemented by the related devices in the flow shown in any one of FIGS. 2 to 6.

[0180] It should be noted that the above device provided by the embodiments of the present disclosure can realize all the method steps achieved by the above method embodiments, and achieve the same technical effects. Therefore, the same parts and beneficial effects of the method embodiments will not be described in detail.

[0181] The embodiments of the present disclosure also provide a computer readable storage medium, which stores computer executable instructions. The computer executable instructions are used for causing a computer to execute the method executed by the above device in the above embodiments.

[0182] The embodiments of the present disclosure also provide a computer program product. When the computer program product is invoked by a computer, the computer is caused to execute the method executed by the above device in the above embodiments.

[0183] Those skilled in the art should understand that the embodiments of the present disclosure can be provided as a method, a system, or a computer program product. Therefore, the present disclosure can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present disclosure can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROMs, optical storage devices, etc.) containing computer-usable program code.

[0184] The present disclosure is described with reference to the flowcharts and / or block diagrams of the method, device (system), and computer program product according to the present disclosure. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of the flows and / or blocks in the flowcharts and / or block diagrams can be implemented by computer program instructions. These computer program instructions can be provided to a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to produce a machine, so that the instructions executed by the computer or other programmable data processing devices produce a device that implements the functions specified in one or more flows in the flowcharts and / or one or more blocks in the block diagrams.

[0185] These computer program instructions can also be stored in a computer readable storage medium that can direct the computer or other programmable data processing devices to work in a specific manner, so that the instructions stored in the computer readable storage medium produce a manufactured product including instruction devices that implement the functions specified in one or more flows in the flowcharts and / or one or more blocks in the block diagrams.

[0186] These computer program instructions can also be loaded into a computer or other programmable data processing devices, so that a series of operational steps are performed on the computer or other programmable devices to generate computer-implemented processes, so that the instructions executed on the computer or other programmable devices provide steps for implementing the functions specified in one or more flows in the flowchart and / or one or more blocks in the block diagram.

[0187] Obviously, various modifications and changes can be made to the present disclosure by those skilled in the art without departing from the spirit and scope of the present disclosure. Thus, it is intended that the present disclosure also cover such modifications and changes as fall within the scope of the claims of the present disclosure and their equivalents.

Claims

1. A method for training a perception model, applied to a perception training server, the method comprising: receiving a training data set sent by one or more base stations participating in perception; training a perception model based on the training data set.

2. The method of claim 1, wherein, The method further comprises: receiving configuration information of a perception signal sent by the one or more base stations participating in perception, wherein the configuration information of the perception signal comprises one or more of sequence characteristics, frequency information, time domain information of the perception signal; wherein the time domain information comprises one or more of the number of time slots, the number of symbols, silence characteristics, beam direction, beams.

3. The method of claim 1, wherein, The training data set comprises one or more of time domain sampling signals of reflection signals and / or scattering signals of the perception signal, joint data of the reflection signals and / or scattering signals of the perception signal transformed into frequency domain data or time domain data, joint data of the reflection signals and / or scattering signals of the perception signal transformed into frequency domain data and time domain data, channel impulse response time domain data of the reflection signals and / or scattering signals of the perception signal, channel impulse response frequency domain data of the reflection signals and / or scattering signals of the perception signal. The perception signal is a signal sent by the one or more base stations participating in perception to a target area.

4. The method according to claims 1-3, wherein, The training data set comprises one or more of signal strength, time delay, frequency offset, target characteristics of the perception signal of the reflection signals and / or scattering signals of the perception signal; the target characteristics of the perception signal comprise one or more of Doppler shift, signal arrival time difference, angle information, channel impulse response, received power, perception signal transmission delay, perception signal reception delay. The perception signal is a signal sent by the one or more base stations participating in perception to a target area.

5. The method of claim 1, wherein, The method further comprises: receiving annotation information sent by one or more user equipment participating in perception, wherein the annotation information comprises one or more of the identity, moving speed, time corresponding to the moving speed, position, time corresponding to the position, form, volume, maximum cross-sectional area of the one or more user equipment.

6. The method of claim 5, wherein, The training of the perception model based on the training data set comprises: training the perception model based on the training data set and the annotation information.

7. The method of claims 1-6, wherein, The method further comprises: sending the perception model to a perception server; wherein the perception model comprises one or more of model architecture, model loss function, model type, model input, model output.

8. A method for training a perception model, applied to a base station, the method comprising: sending a perception signal to a target area and receiving reflection signals and / or scattering signals of the perception signal; generating a training data set based on one or more of the perception signal, the reflection signals of the perception signal, and the scattering signals of the perception signal; sending the training data set to a perception training server.

9. The method of claim 8, wherein, The method further comprises: sending, to the perception training server, configuration information of the perception signal, wherein the configuration information of the perception signal comprises one or more of sequence characteristics, frequency information, time domain information of the perception signal, and the time domain information comprises one or more of time slot quantity, symbol quantity, muting characteristics, beam direction, and beam.

10. The method of claims 8-9, wherein, The training data set comprises one or more of time domain sampling signals of reflection signals and / or scattering signals of the perception signal, joint data of the reflection signals and / or scattering signals of the perception signal transformed into frequency domain data or time domain data, joint data of the reflection signals and / or scattering signals of the perception signal transformed into frequency domain data and time domain data, channel impulse response time domain data of the reflection signals and / or scattering signals of the perception signal, and channel impulse response frequency domain data of the reflection signals and / or scattering signals of the perception signal.

11. The method of claims 8-10, wherein, The training data set comprises one or more of signal strength, time delay, frequency offset, and target characteristics of the perception signal of the reflection signals and / or scattering signals of the perception signal, and the target characteristics of the perception signal comprises one or more of Doppler shift, signal arrival time difference, angle information, channel impulse response, received power, perception signal sending delay, and perception signal receiving delay.

12. A method for training a perception model, applied to a user equipment, the method comprising: sending, to a perception training server, labeling information for training a perception model; The labeling information comprises one or more of an identifier of the user equipment, a moving speed, a time corresponding to the moving speed, a position, a time corresponding to the position, a form of the user equipment, a volume of the user equipment, and a maximum cross-sectional area of the user equipment.

13. The method of claim 12, wherein, The perception model comprises one or more of a model architecture, a model loss function, a model type, a model input, and a model output.

14. A method for obtaining a perception model, applied to a perception server, the method comprising: receiving a perception model sent by a perception training server, wherein the perception model comprises one or more of a model architecture, a model loss function, a model type, a model input, and a model output.

15. The method of claim 14, wherein, The perception model is generated based on a training data set sent by one or more base stations participating in perception and labeling information sent by one or more user equipments participating in perception.

16. The method of claim 15, wherein, The training data set comprises one or more of time domain sampling signals of reflection signals and / or scattering signals of the perception signal, joint data of the reflection signals and / or scattering signals of the perception signal transformed into frequency domain data or time domain data, joint data of the reflection signals and / or scattering signals of the perception signal transformed into frequency domain data and time domain data, channel impulse response time domain data of the reflection signals and / or scattering signals of the perception signal, and channel impulse response frequency domain data of the reflection signals and / or scattering signals of the perception signal. The perception signal is a signal sent by the one or more base stations participating in perception to a target area.

17. The method of claim 15 or 16, wherein, The training data set comprises one or more of signal strength, time delay, frequency offset, target feature of the perception signal of the reflection signal and / or scattering signal of the perception signal; wherein the target feature of the perception signal comprises one or more of Doppler shift, signal arrival time difference, angle information, channel impulse response, received power, perception signal sending delay, perception signal receiving delay. The perception signal is a signal sent by the one or more base stations participating in perception to a target area.

18. The method of claim 15, wherein, The labeling information comprises one or more of the identity, moving speed, time corresponding to the moving speed, position, time corresponding to the position, shape, volume, maximum cross-sectional area of the user equipment.

19. A perception training server, comprising: a receiving module configured to receive a training data set sent by one or more base stations participating in perception; a training module configured to train a perception model based on the training data set.

20. The perception training server of claim 19, wherein, The receiving module is further configured to receive configuration information of the perception signal sent by the one or more base stations participating in perception, wherein the configuration information of the perception signal comprises one or more of sequence feature, frequency information, time domain information of the perception signal; wherein the time domain information comprises one or more of time slot number, symbol number, silence feature, beam direction, beam.

21. The perception training server of claim 19 or 20, wherein, The training data set comprises one or more of time domain sampling signal of the reflection signal and / or scattering signal of the perception signal, joint data of the reflection signal and / or scattering signal of the perception signal transformed into frequency domain data or time domain data, joint data of the reflection signal and / or scattering signal of the perception signal transformed into frequency domain data and time domain data, channel impulse response time domain data of the reflection signal and / or scattering signal of the perception signal, channel impulse response frequency domain data of the reflection signal and / or scattering signal of the perception signal. The perception signal is a signal sent by the one or more base stations participating in perception to a target area.

22. The perception training server of claims 19-20, wherein, The training data set comprises one or more of signal strength, time delay, frequency offset, target feature of the perception signal of the reflection signal and / or scattering signal of the perception signal; wherein the target feature of the perception signal comprises one or more of Doppler shift, signal arrival time difference, angle information, channel impulse response, received power, perception signal sending delay, perception signal receiving delay. The perception signal is a signal sent by the one or more base stations participating in perception to a target area.

23. The perception training server of claim 19, wherein, The receiving module is further configured to receive labeling information sent by one or more user equipment participating in perception, wherein the labeling information comprises one or more of the identity, moving speed, time corresponding to the moving speed, position, time corresponding to the position, shape, volume, maximum cross-sectional area of the one or more user equipment.

24. The perception training server of claim 23, wherein, The training module is further configured to train the perception model based on the training data set and the labeling information. The training data set comprises one or more of signal strength, time delay, frequency offset, target feature of the perception signal of the reflection signal and / or scattering signal of the perception signal; wherein the target feature of the perception signal comprises one or more of Doppler shift, signal arrival time difference, angle information, channel impulse response, received power, perception signal sending delay, perception signal receiving delay. The perception signal is a signal sent by the one or more base stations participating in perception to a target area. The receiving module is further configured to receive labeling information sent by one or more user equipment participating in perception, wherein the labeling information comprises one or more of the identity, moving speed, time corresponding to the moving speed, position, time corresponding to the position, shape, volume, maximum cross-sectional area of the one or more user equipment. The training module is further configured to train the perception model based on the training data set and the labeling information.

25. The perception training server of claims 19-24, wherein, The perception training server further includes a sending module configured to send the perception model to a perception server. The perception model includes one or more of a model architecture, a model loss function, a model type, a model input, and a model output.

26. A base station, comprising: a sending module configured to send a perception signal to a target area; a receiving module configured to receive a reflected signal and / or a scattered signal of the perception signal; a processing module configured to generate a training data set based on one or more of the perception signal, the reflected signal of the perception signal, and the scattered signal of the perception signal; The sending module is further configured to send the training data set to a perception training server.

27. The base station of claim 26, wherein, The sending module is further configured to send configuration information of the perception signal to the perception training server, wherein the configuration information of the perception signal includes one or more of sequence characteristics of the perception signal, frequency information, time domain information, and the time domain information includes one or more of a number of time slots, a number of symbols, silence characteristics, beam directions, and beams.

28. The base station of claim 26 or 27, wherein, The training data set includes one or more of time domain sampling signals of the reflected signal and / or the scattered signal of the perception signal, joint data of the reflected signal and / or the scattered signal of the perception signal transformed into frequency domain data or time domain data, joint data of the reflected signal and / or the scattered signal of the perception signal transformed into frequency domain data and time domain data, channel impulse response time domain data of the reflected signal and / or the scattered signal of the perception signal, and channel impulse response frequency domain data of the reflected signal and / or the scattered signal of the perception signal.

29. The base station of claim 26 or 28, wherein, The training data set includes one or more of signal strength, time delay, frequency offset, and target characteristics of the perception signal of the reflected signal and / or the scattered signal of the perception signal, and the target characteristics of the perception signal include one or more of Doppler shift, signal arrival time difference, angle information, channel impulse response, received power, perception signal sending delay, and perception signal receiving delay.

30. A user equipment, comprising: a sending module configured to send labeling information to a perception training server, the labeling information being used to train a perception model; The labeling information includes one or more of an identity of the user equipment, a moving speed, a time corresponding to the moving speed, a location, a time corresponding to the location, a form of the user equipment, a volume of the user equipment, and a maximum cross-sectional area of the user equipment.

31. The user equipment of claim 30, wherein, The perception model includes one or more of a model architecture, a model loss function, a model type, a model input, and a model output.

32. A perception server, comprising: a receiving module configured to receive a perception model sent by a perception training server, wherein the perception model includes one or more of a model architecture, a model loss function, a model type, a model input, and a model output.

33. The perception server of claim 32, wherein, The perception model is generated based on one or more training data sets sent by one or more base stations participating in perception and one or more labeling information sent by one or more user equipments participating in perception.

34. The perception server of claim 33, wherein, The training data set comprises one or more of time domain sampling signals of reflection signals and / or scattering signals of the perception signal, joint data of the reflection signals and / or scattering signals of the perception signal transformed into frequency domain data or time domain data, joint data of the reflection signals and / or scattering signals of the perception signal transformed into frequency domain data and time domain data, channel impulse response time domain data of the reflection signals and / or scattering signals of the perception signal, channel impulse response frequency domain data of the reflection signals and / or scattering signals of the perception signal; The perception signal is a signal sent by the one or more base stations participating in perception to a target area.

35. The perception server of claim 33 or 34, wherein, The training data set comprises one or more of signal strength, time delay, frequency offset, target features of the perception signal of the reflection signals and / or scattering signals of the perception signal; wherein the target features of the perception signal comprise one or more of Doppler shift, signal arrival time difference, angle information, channel impulse response, received power, perception signal sending delay, perception signal receiving delay. The perception signal is a signal sent by the one or more base stations participating in perception to a target area.

36. The perception server of claim 33, wherein, The labeling information comprises one or more of an identifier of the user equipment, a moving speed, a time corresponding to the moving speed, a position, a time corresponding to the position, a form of the user equipment, a volume of the user equipment, a maximum cross-sectional area of the user equipment.

37. A server comprising: A memory and a processor; The memory is configured to store program instructions; The processor is configured to invoke the program instructions stored in the memory, and execute the method according to any one of claims 1-7 or 14-18 based on the program instructions.

38. A base station comprising: A memory and a processor; The memory is configured to store program instructions; The processor is configured to invoke the program instructions stored in the memory, and execute the method according to any one of claims 8-11 based on the program instructions.

39. A user equipment comprising: A memory and a processor; The memory is configured to store program instructions; The processor is configured to invoke the program instructions stored in the memory, and execute the method according to any one of claims 12-13 based on the program instructions.

40. A computer readable storage medium storing computer instructions, when the computer instructions are run on a computer, causing the computer to execute the method according to any one of claims 1-18.

Images