Methods and apparatus used for wireless communication

Abstract

This invention discloses a method and apparatus for wireless communication. A first node receives a first MAC frame, wherein the MAC header of the first MAC frame includes a first BSSID, and the first MAC frame indicates the deactivation time of a first AI model for a second BSSID; transmits a second MAC frame, wherein the second MAC frame requests the download of the first AI model; and receives at least one MAC frame, wherein the at least one MAC frame includes the first AI model; wherein the request of the second MAC frame depends on the indication of the first MAC frame. This application is beneficial for improving the efficiency of AI model download.

Description

Technical Field

[0001] This application relates to methods and apparatus in wireless communication systems, and more particularly to methods and apparatus for AI (artificial intelligence) in wireless local area networks (WLANs). Background Technology

[0002] Wi-Fi (Wireless Fidelity) has become a ubiquitous technology in the world today, providing connectivity for billions of devices and becoming the preferred method for more and more users to access the internet, with a growing trend of gradually replacing wired access. Due to its relatively simple implementation, reliable communication, high flexibility, and relatively low implementation cost, Wi-Fi technology has become the mainstream technical standard for WLAN (Wireless Local Area Network). To adapt to the ever-emerging new business applications and narrow the performance gap with wired networks, each generation of the Wi-Fi standard has been dedicated to improving wireless transmission speeds.

[0003] With the widespread application of AI (Artificial Intelligence) or ML (Machine Learning) technologies, the IEEE (Institute of Electrical and Electronics Engineers) 802.11 Working Group established an AI / ML research interest group to study the impact of related technologies on WLANs. The application scenarios studied include CSI (Channel State Information) feedback compression, distributed channel access, roaming, and so on.

[0004] An AI model, sometimes called an AI function or AI agent, is typically described by hyperparameters and model parameters. Hyperparameters are parameters that need to be pre-set during machine learning model training and cannot be automatically learned through conventional training processes; examples include AI model structure, learning rate, and batch size. Model parameters, sometimes called parameter sets, are internal variables that the machine learning model automatically learns from data during training. AI models typically need to be trained and meet certain performance requirements before they can be used for inference. A portion of the training data can be used to test the AI ​​model's performance. Generally, the complexity and overhead of training are far greater than those of inference.

[0005] Typical model structures include classic models such as the Transformer structure, RNN (Recurrent Neural Network), and CNN (Conventional Neural Network), or hybrid models composed of multiple models. Summary of the Invention

[0006] For the application of AI / ML in WLAN or 802.11 series protocols, one problem that needs to be solved is how the STA (site) determines whether it supports or downloads an AI model.

[0007] This application discloses a solution to the aforementioned problems. Unless otherwise specified, the embodiments and features described in this application can be arbitrarily combined. It should be noted that although this application is initially intended for WLAN, it can also be used for wireless networks such as WMAN (Wireless Metropolitan Area Network), WRAN (Wireless Regional Area Network), and WSN (Wireless Specialty Network). Furthermore, adopting a unified solution for different scenarios (including but not limited to WLAN and WMAN scenarios) helps reduce hardware complexity and cost. Unless otherwise specified, the embodiments and features described in the first node device of this application can be applied to the second node device, and vice versa. In particular, the interpretation of terms, nouns, functions, and variables in this application (unless otherwise specified) can refer to the definitions in IEEE 802.11, IEEE 802.15, IEEE 802.22, and other relevant standards and protocols.

[0008] This application discloses a method used in a first node for wireless communication, characterized by comprising:

[0009] Receive a first MAC (Medium Access Control) frame, wherein the MAC header of the first MAC frame includes the first BSSID (Basic Service Set Identifier), and the first MAC frame indicates the deactivation time of the first AI model for the second BSSID;

[0010] Send a second MAC frame, wherein the second MAC frame requests the download of the first AI model;

[0011] Receive at least one MAC frame, wherein the at least one MAC frame includes the first AI model;

[0012] The request in the second MAC frame depends on the indication in the first MAC frame.

[0013] The above method helps the first node determine whether to download the first AI model based on the availability of the first AI model in the second BSSID, avoiding insufficient utilization efficiency caused by the BSSID being too close to deactivation time after downloading the AI ​​model.

[0014] Furthermore, the above method provides the possibility of flexibly activating or deactivating the first AI model.

[0015] Specifically, according to one aspect of this application, the above method is characterized by comprising:

[0016] Send a third MAC frame; send a fourth MAC frame before the deactivation time of the first AI model for the second BSSID;

[0017] The third MAC frame includes the second BSSID in its MAC header; the third MAC frame indicates support for the first AI model; the third MAC frame is sent after the reception of at least one MAC frame; the fourth MAC frame includes the second BSSID in its MAC header.

[0018] The above method avoids the waste of air interface resources caused by continuing to send related data in the second BSSID after the first AI model has been deactivated.

[0019] Specifically, according to one aspect of this application, the above method is characterized by comprising:

[0020] In response to the deactivation time of the first AI model reaching the second BSSID, inference based on the first AI model is stopped.

[0021] The above method helps to terminate inference based on the first AI model in a timely manner, thus saving power.

[0022] Specifically, according to one aspect of this application, the above method is characterized in that the first MAC frame indicates at least one of the input format or output format of the first AI model.

[0023] The above aspects help the first node to assist in determining whether to download the first AI model.

[0024] Specifically, according to one aspect of this application, the above method is characterized in that the site supporting the first AI model must have the capability to support wireless measurement.

[0025] The above methods provide the possibility for optimizing capability reporting, which helps to reduce signaling overhead or the probability of false alarms.

[0026] This application discloses a method used in a second node for wireless communication, characterized by comprising:

[0027] Receive a fifth MAC frame; wherein the MAC header of the fifth MAC frame includes a second BSSID; the fifth MAC frame indicates the deactivation time of the first AI model for the second BSSID;

[0028] Send a first MAC frame, wherein the MAC header of the first MAC frame includes the first BSSID, and the first MAC frame indicates the deactivation time of the first AI model for the second BSSID;

[0029] Receive a second MAC frame, wherein the second MAC frame requests the download of the first AI model;

[0030] Send at least one MAC frame, wherein the at least one MAC frame includes the first AI model;

[0031] The request in the second MAC frame depends on the indication in the first MAC frame.

[0032] This application discloses a method used in a third node for wireless communication, characterized by comprising:

[0033] Send a fifth MAC frame, wherein the MAC header of the fifth MAC frame includes a second BSSID, and the fifth MAC frame indicates the deactivation time of the first AI model for the second BSSID;

[0034] Receive a third MAC frame; receive a fourth MAC frame before the deactivation time of the first AI model for the second BSSID; in response to the arrival of the deactivation time of the first AI model for the second BSSID, stop processing MAC frames for inference based on the first AI model;

[0035] The third MAC frame includes the second BSSID in its MAC header; the third MAC frame indicates support for the first AI model; the third MAC frame is received after the fifth MAC frame is sent; the fourth MAC frame includes the second BSSID in its MAC header.

[0036] This application discloses a first node used for wireless communication, characterized in that it includes:

[0037] A first receiver receives a first MAC frame, wherein the MAC header of the first MAC frame includes the first BSSID, and the first MAC frame indicates the deactivation time of a first AI model for a second BSSID.

[0038] The first transmitter sends a second MAC frame, wherein the second MAC frame requests the download of the first AI model;

[0039] The first receiver receives at least one MAC frame, wherein the at least one MAC frame includes the first AI model;

[0040] The request in the second MAC frame depends on the indication in the first MAC frame.

[0041] This application discloses a second node used for wireless communication, characterized by comprising:

[0042] The second receiver receives the fifth MAC frame; wherein the MAC header of the fifth MAC frame includes the second BSSID; the fifth MAC frame indicates the deactivation time of the first AI model for the second BSSID;

[0043] The second transmitter sends a first MAC frame, wherein the MAC header of the first MAC frame includes the first BSSID, and the first MAC frame indicates the deactivation time of the first AI model for the second BSSID;

[0044] The second receiver receives the second MAC frame, wherein the second MAC frame requests the download of the first AI model;

[0045] The second transmitter sends at least one MAC frame, wherein the at least one MAC frame includes the first AI model;

[0046] The request in the second MAC frame depends on the indication in the first MAC frame. Attached Figure Description

[0047] Other features, objects, and advantages of this application will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0048] Figure 1 A communication flowchart of a first node according to an embodiment of this application is illustrated;

[0049] Figure 2 A schematic diagram of a network architecture according to an embodiment of this application is illustrated;

[0050] Figure 3 A schematic diagram of a wireless protocol stack according to an embodiment of this application is illustrated;

[0051] Figure 4 A schematic diagram of the hardware module of a communication device according to an embodiment of this application is illustrated;

[0052] Figure 5 A communication flowchart of a first node, a second node, and a third node according to an embodiment of this application is illustrated;

[0053] Figure 6 A flowchart illustrating the transmission of a first AI model according to an embodiment of this application is provided;

[0054] Figure 7 A schematic diagram illustrating a first AI model according to an embodiment of this application is provided;

[0055] Figure 8 A schematic diagram illustrating reconnection to a second BSSID according to one embodiment of this application is shown;

[0056] Figure 9 A structural block diagram of a processing apparatus in a first node according to an embodiment of this application is illustrated;

[0057] Figure 10 A structural block diagram of a processing apparatus in a second node according to an embodiment of this application is illustrated;

[0058] Figure 11 A structural block diagram of a processing apparatus in a third node according to an embodiment of this application is illustrated. Detailed Implementation

[0059] The technical solution of this application will be further described in detail below with reference to the accompanying drawings. It should be noted that, unless otherwise specified, the embodiments and features in the embodiments of this application can be arbitrarily combined with each other.

[0060] Example 1

[0061] Example 1 illustrates a signal processing flowchart in a first node according to an embodiment of this application, as shown in the attached diagram. Figure 1 As shown.

[0062] In step 101, the first node 100 receives a first MAC frame, wherein the MAC header of the first MAC frame includes the first BSSID, and the first MAC frame indicates the deactivation time of the first AI model for the second BSSID; in step 102, it sends a second MAC frame, wherein the second MAC frame requests the download of the first AI model; and in step 103, it receives at least one MAC frame, wherein the at least one MAC frame includes the first AI model.

[0063] In Example 1, the request of the second MAC frame depends on the indication of the first MAC frame.

[0064] Typically, the deactivation time of the first AI model for the second BSSID is different from the deactivation time of the first AI model for the second BSSID.

[0065] Typically, the first node 100 is a non-AP (access point) site, and the first node 100 is associated with the AP indicated by the first BSSID.

[0066] Typically, the first node roams from the first BSSID to the second BSSID before sending the second MAC.

[0067] As an example, the first MAC frame explicitly indicates a plurality of BSSIDs and indicates the deactivation time of the first AI model for each of the plurality of BSSIDs; the second BSSID is one of the plurality of BSSIDs.

[0068] As a sub-implementation of the above embodiment, the plurality of BSSIDs includes the first BSSID.

[0069] As an example, the first MAC frame is a beacon frame or a probe response frame.

[0070] The above embodiments enable the first node to know the second BSSID as early as possible, for example, to obtain the configuration of the first AI model of the second BSSID before associating with the AP (access point) indicated by the second BSSID.

[0071] As an example, the AP (access point) indicated by the first BSSID and the AP indicated by the second BSSID belong to the same ESS (Extended Basic Service Set).

[0072] As an example, the AP indicated by the second BSSID no longer supports the distribution of the first AI model after the deactivation time of the first AI model for the second BSSID.

[0073] As an example, the AP indicated by the second BSSID no longer supports inference using the first AI model after the deactivation time of the first AI model for the second BSSID.

[0074] As an example, the AP indicated by the second BSSID no longer supports receiving the inference output of the first AI model after the deactivation time of the first AI model for the second BSSID.

[0075] As an example, the AP indicated by the second BSSID supports the distribution of the first AI model, or supports inference using the first AI model, or supports receiving the inference output of the first AI model before the deactivation time of the first AI model for the second BSSID.

[0076] As an example, the at least one MAC frame includes multiple MAC frames, and each of the multiple MAC frames includes only a portion of the first AI model.

[0077] The above embodiments make it possible to download / distribute large-sized AI models.

[0078] As a sub-example of the above embodiment, the plurality of MAC frames are triggered by the same request frame sent by the first node.

[0079] As a specific implementation of the above sub-example, each of the at least one MAC frames is a management frame.

[0080] As an example, each of the at least one MAC frames is a data frame, and the MSDU (MAC service data unit) in each of the at least one MAC frames does not include an LLC (Logical Link Control) header.

[0081] The above aspects utilize existing frame types and avoid delegating the processing of the first information to higher layers, thereby improving processing efficiency.

[0082] As an example, the AP indicated by the first BSSID and the AP indicated by the second BSSID are managed by the same AC (Access Controller).

[0083] As an example, the first MAC frame indicates the computing resources required by the first AI model, or the storage resources required, or both.

[0084] The above embodiments are beneficial for the first node to determine whether it can support the first AI model before downloading the first AI model.

[0085] As an example, the unit of the required computing resources is FLOPS (Floating-point Operations Per Second).

[0086] As an example, the unit of the required computing resources is megabytes.

[0087] As one embodiment, the first MAC frame indicates the identifier of the first AI model, or the activation time of the first AI model, or the identifier of the first AI model and the activation time of the first AI model.

[0088] As an example, the identifier of the first AI model is unique within the ESS to which the first node belongs.

[0089] As an example, the identifier of the first AI model is unique among at least the BSSs identified by the first BSSID and the second BSSID.

[0090] Typically, the activation time of the first AI model for the second BSSID is earlier than the transmission time of the first MAC frame, and the deactivation time of the first AI model for the second BSSID is later than the transmission time of the first MAC frame.

[0091] As an example, the deactivation time of the first AI model for the second BSSID is the value of the TSF (timing synchronization function) timer.

[0092] As an example, the first MAC frame indicates the time interval between the deactivation time of the first AI model for the second BSSID and the start transmission time of the PPDU (Physical Layer Protocol Data Unit) of the first MAC frame.

[0093] As an example, the first MAC frame indicates at least one of the input format or output format of the first AI model.

[0094] As an example, the input format of the first AI model depends on one or more types of wireless measurements, such as beacons, frames, channel load, noise histogram, STA statistics, location configuration information, neighbor report, link measurement, and transmit stream / category measurement.

[0095] As an example, the site supporting the first AI model must have the capability to support wireless measurement.

[0096] As a sub-implementation of the above embodiment, the wireless measurement bit in the Capability Information field sent by the station supporting the first AI model is 1.

[0097] Typically, how to determine the request for the first AI model (i.e., sending the second MAC frame) based on the indication of the first MAC frame is implementation-dependent or determined by the vendor of the first node N1; some non-limiting implementations are given below.

[0098] For example, the first node 100 predicts the roaming path and the association / stay time corresponding to each AP on the roaming path; if the association / stay time in the AP indicated by the second BSSID exceeds a certain threshold before the deactivation time of the first AI model for the second BSSID, the first node 100 determines to send the second MAC frame; typically, the prediction of the roaming path is AI-based.

[0099] For example, the first node 100 maintains a list of BSSIDs, each of which supports the first AI model before a corresponding deactivation time, and the list of BSSIDs includes the second BSSID; the first node 100 predicts the sum of the activation times of the first AI model associated with the APs indicated by the list of BSSIDs on the roaming path, and if the sum exceeds a certain threshold, the first node 100 determines to send the second MAC frame.

[0100] Compared to relying solely on the first AI model applicable to the first BSSID, applying the first AI model to multiple AI models, including the second BSSID, provides the first node 100 with more accurate information to help determine whether to download the first AI model. For example, even if the first node 100 is about to leave the BSS indicated by the first BSSID, if the reconnecting AP corresponds to the second BSSID and the activation time is long enough, the first node 100 can still choose to download the first AI model.

[0101] Example 2

[0102] Example 2 illustrates a network architecture diagram according to an embodiment of this application, as shown in the attached diagram. Figure 2 As shown. Figure 2This section explains the architecture of an IEEE 802.11 local area network (LAN). A Basic Service Set (BSS) is a fundamental building block of an IEEE 802.11 LAN, consisting of a group of terminals that can communicate with each other. BSSs include Independent Basic Service Sets (IBSSs) and Basic Basic Service Sets. Terminals in an IBSS can communicate directly with each other, while terminals in a Basic Basic Service Set communicate with each other through a central terminal, also known as an access point. Basic Basic Service Sets can interconnect with other Basic Basic Service Sets via the access point through a Distribution System (DS). Multiple Basic Basic Service Sets interconnected via DS can form an Extended Basic Service Set (ESS) to extend the network coverage area. As shown in the figure, BSS231 is an IBSS, which includes terminals 201 and 202 that can communicate directly. As shown in the figure, BSS232 and BSS233 are basic service sets. Communication between terminals within a basic service set requires forwarding through an access point. For example, if terminal 203 in BSS232 needs to send data to terminal 204 in BSS232, it first needs to send the data to access point 211, and then access point 211 forwards the data to terminal 204; the reverse is also true. As shown in the figure, BSS232 and BSS233 together form ESS234. Access points 211 and 212, located within the same ESS234, use the same Service Set Identifier (SSID). BSS232 and BSS233 typically overlap. Terminals belonging to the same ESS234 can communicate with each other. As shown in the figure, terminal 203 located in BSS232 can communicate with terminal 205 located in BSS233. At this time, access point 211 and access point 212 have bridging functions and transmit data between different access points through the distribution system medium (DSM).

[0103] The BSS can identify user terminals via its SSID and other devices via its Basic Service Set Identifier (BSSID). The BSSID can be the MAC (Medium Access Control) address of access point 211. Access point 211 periodically broadcasts beacon frames including the BSSID, enabling any terminal within its wireless coverage area to associate with or re-associate with access point 211 to establish corresponding downlink 223 and uplink 224 (uplink and downlink can be collectively referred to as Wi-Fi links). The beacon may include: an identifier of the primary channel used by the corresponding access point 211 and a timing synchronization function for establishing or maintaining timing synchronization with access point 211.

[0104] A terminal is a single addressable instance of a MAC and a logical entity of a physical layer (PHY) interface for wireless medium (WM). In IEEE 802.11, a terminal is an addressable unit; in IBSS, a terminal can send beacon frames to announce the presence of a WLAN. In the Basic Services Set, to establish a Wi-Fi link with access point 211, terminal 203 is configured to perform passive or active scanning on frequency channels in one or more frequency bands (e.g., 2.4 GHz, 5 GHz, 6 GHz, or 60 GHz bands). To perform passive scanning, terminal 203 listens for beacons periodically sent by the corresponding access point 211. To perform active scanning, terminal 203 generates probe requests and sequentially transmits these probe requests on each channel to be scanned, and listens for probe responses from access point 211. Terminal 203 identifies or selects an access point 211 to associate with using scanning information obtained through passive or active scanning, and performs authentication and association operations to establish a communication link with the selected access point 211. Upon completion of the association operation, access point 211 assigns an association identifier (AID) to terminal 203, which is used by access point 211 to track terminal 203. Terminals include, but are not limited to, mobile phones, laptops, personal digital assistants (PDAs), media devices, video devices, digital audio players (e.g., MP3 players), cameras, game consoles, drones, aircraft, narrowband IoT devices, machine-type communication devices, land vehicles, automobiles, in-vehicle equipment, in-vehicle communication units, wearable devices, or any other similar functional devices. Those skilled in the art may also refer to the terminal as a non-AP (access point) workstation, a non-AP multi-link device (non-AP MLD), a mobile station, a subscriber station, a mobile unit, a subscriber unit, a wireless unit, a remote unit, a mobile device, a wireless device, a wireless communication device, a remote device, a mobile subscriber station, an access terminal, a mobile terminal, a wireless terminal, a remote terminal, a handheld device, a user agent, a mobile client, a client, or any other suitable term.

[0105] An access point is an entity that includes terminals and provides access to a distribution system (DS) for associated terminals via a wireless medium (WM). An access point includes both a terminal and a distribution system access function (DSAF). Access points may be referred to as AP multi-link devices (AP MLDs), base stations, base transceivers, radio base stations, radio transceivers, transceiver functions, Basic Service Sets (BSS), Extended Service Sets (ESS), Transmission Reception Points (TRPs), or other suitable terms. Access points can provide access to external networks to terminals in a WLAN via a corresponding WiFi link.

[0106] The Distribution System (DS) is the backbone network for data transmission between access points, often referred to as the backbone network, which is typically Ethernet. Access points use the Logical Link Control (LLC) layer as the upper layer, such as the IP (Internet Protocol) layer, to provide a unified interface. All user IP packets are transmitted through a gateway (GW), which provides IP address allocation and other functions for terminals within its corresponding network segment. The gateway is also called an IP router and connects to Internet Service 230240. Internet Service 230240 includes carrier-specific Internet Protocol services, specifically including the Internet, intranets, IMS (IP Multimedia Subsystem), and PS (Packet Switching) streaming services.

[0107] As an example, terminal 201 corresponds to the first node in this application, and terminal 202 corresponds to the second node in this application.

[0108] As one embodiment, the terminal 203 corresponds to the first node in this application, and the access point 211 corresponds to the second node in this application.

[0109] As an example, the terminal 201 supports Wi-Fi, such as the IEEE 802.11 protocol.

[0110] As an example, the terminal 201 is a DMG (directional multi-gigabit) terminal, a QoS (Quality of Service) terminal, a HT (High Throughput) terminal, an EHT (Extreme High Throughput) terminal, or an MLD (multi-link device) terminal, etc.

[0111] As an example, the terminal 203 is a DMG (directional multi-gigabit) terminal, a QoS (Quality of Service) terminal, a HT (High Throughput) terminal, an EHT (Extreme High Throughput) terminal, or an MLD (multi-link device) terminal, etc.

[0112] As an example, the access point 211 is a base station, or a femtocell, or an AP multi-link device, or a test device (such as a transceiver device simulating part of the base station's functions, a signaling tester), or a vehicle-mounted device, or an MLD (multi-link device).

[0113] As an example, the wireless link from terminal 202 to terminal 201 is downlink 221, where terminal 201 is the source terminal and terminal 202 is the destination terminal, and downlink 221 is used to perform downlink transmission.

[0114] As an example, the wireless link from terminal 201 to terminal 202 is uplink 222, where terminal 201 is the source terminal and terminal 202 is the destination terminal, and uplink 222 is used to perform uplink transmission.

[0115] As an example, the wireless link from the access point 211 to the terminal 203 is a downlink 223, which is used to perform downlink transmission.

[0116] As an example, the wireless link from the terminal 203 to the access point 211 is an uplink 224, which is used to perform uplink transmission.

[0117] Example 3

[0118] Example 3 illustrates a schematic diagram of a wireless protocol stack according to an embodiment of this application, as shown in the attached diagram. Figure 3 As shown, IEEE 802.11 uses CSMA (carrier sense multiple access) / CA (collision avoidance) protocols to control access to the transmission medium. Figure 3 The wireless protocol architecture is illustrated using two layers: the data link layer and the physical layer. The physical layer, the lowest layer, implements various physical layer signal processing functions and will be referred to as PHY 301 in this paper. The data link layer, above the physical layer, is responsible for the link between terminals, or between a terminal and an access point. The data link layer includes MAC 302. MAC 302 works in conjunction with PHY 301 to complete data transmission and various management services.

[0119] MAC302 is responsible for encapsulating MSDU (MAC Service Data Unit) into MPDU (MAC Protocol Data Unit) frames, defining the frames, synchronizing frames, processing the destination and source MAC addresses, and handling frame errors. For data packets from higher layers or larger management packets, it performs packet fragmentation and defragmentation, and performs integrity protection and cryptographic encapsulation on data requiring protection. PHY301 is divided into two sub-layers: the Physical Layer Convergence Procedure (PLCP) sub-layer and the Physical Medium Dependent (PMD) sub-layer. The frames received from MAC302 have a PHY header added to the PLCP sublayer to generate a PPDU (PHY protocol data unit). Typically, the PPDU contains a pilot signal (preamble) and a PHY header to assist in the synchronization of received data, as well as demodulation of the MPDU. Then, the PMD is responsible for encoding and modulating the MPDU and transmitting it into the air using an antenna.

[0120] The MAC302 implements access control. For CSMA / CA-based access, it is managed by the distributed coordination function (DCF). If contention-free service is required, it can be managed by the point coordination function (PCF), which is built on top of the DCF. Alternatively, a hybrid coordination function (HCF) can be used, somewhere between DCF and PCF. In the DCF, carrier sensing determines the availability of the medium. Physical carrier sensing is provided by the PHY301, and virtual carrier sensing is provided by the network allocation vector (NAV). To ensure uninterrupted medium usage and data transmission, terminals can use RTS (request to send) / CTS (clear to send) switching. IEEE 802.11 frames typically include a duration field to reserve the medium usage time. Before attempting to transmit any data, the medium must be checked for idleness. If the network is busy, access must be delayed, and a backoff algorithm is used to avoid collisions. If the medium idle time is longer than DIFS (distributed interframe space), transmission can proceed immediately. In PCF, after the access point takes over the wireless medium, it polls the connected terminals for data to be transmitted according to the polling list. During contention-free periods, terminals must not transmit data unless the access point requests it with a polling frame. MAC302 can include error recovery functionality, meaning the sender is responsible for retransmitting each frame if an acknowledgment (ACK) is not received. MAC302 can prioritize latency-sensitive services. When there is only one transmission queue, latency-sensitive service data frames are placed at the front of the queue. When multiple transmission queues are used, one queue is dedicated to transmitting high-priority sensitive services. To achieve energy saving for terminals, they can periodically enter a sleep state. In the sleep state, the access point temporarily stores frames for each terminal in the sleep state. If there are temporary frames, the access point will inform the terminal in subsequent Beacon frames. Terminals woken up from power-saving mode can retrieve these temporary frames using PS (power save)-Poll frames. The MAC302's management functions also include channel management, connection management, quality of service, power control, security management, and time synchronization.Channel management includes channel scanning, channel measurement, and channel handover; connection management includes user authentication, association, reassociation, and disassociation, and point-to-point connections; quality of service includes QoS service scheduling and traffic management; power management includes transmit power management and adaptive transmit power control; security management includes key generation and distribution; and time synchronization includes higher-layer synchronization support. Although not illustrated, the terminal may also have several upper layers above the MAC302, including an LLC layer.

[0121] As an example, Appendix Figure 3 The wireless protocol architecture described herein is applicable to the first node in this application.

[0122] As an example, Appendix Figure 3 The wireless protocol architecture described herein is applicable to the second node in this application.

[0123] As an example, Appendix Figure 3 The wireless protocol architecture described herein is applicable to the third node in this application.

[0124] Example 4

[0125] Example 4 illustrates a hardware module schematic diagram of a communication device according to an embodiment of this application, as shown in the attached diagram. Figure 4 As shown. Figure 4 This is a block diagram of a first communication device 450 and a second communication device 410 communicating with each other in the access network.

[0126] The first communication device 450 includes a controller / processor 459, a memory 460, a data source 467, a transmitting processor 468, a receiving processor 456, a multi-antenna transmitting processor 457, a multi-antenna receiving processor 458, a transmitter / receiver 454, and an antenna 452.

[0127] The second communication device 410 includes a controller / processor 475, a memory 476, a data source 477, a receiver processor 470, a transmitter processor 416, a multi-antenna receiver processor 472, a multi-antenna transmitter processor 471, a transmitter / receiver 418, and an antenna 420.

[0128] In the transmission from the second communication device 410 to the first communication device 450, at the second communication device 410, upper-layer data packets from the Internet or from data source 477 are provided to the controller / processor 475. The Internet and data source 477 represent all protocol layers above the MAC layer. The controller / processor 475 implements the functionality of the MAC layer. In the transmission from the second communication device 410 to the first communication device 450, the controller / processor 475 provides header compression, encryption, packet segmentation and reordering, service-to-link mapping, and radio resource allocation to the first communication device 450. The controller / processor 475 is also responsible for retransmitting lost packets and signaling to the first communication device 450. The transmission processor 416 and the multi-antenna transmission processor 471 implement various signal processing functions for the PHY layer (i.e., the physical layer). Transmit processor 416 performs encoding and interleaving to facilitate forward error correction (FEC) at the second communication device 410, and mapping of signal clusters based on various modulation schemes (e.g., Binary Phase Shift Keying (BPSK), Quadrature Phase Shift Keying (QPSK), M-Phase Shift Keying (M-PSK), M-QAM). Multi-antenna transmit processor 471 performs digital spatial precoding and beamforming processing on the encoded and modulated symbols to generate one or more spatial streams. Transmit processor 416 then maps each spatial stream to subcarriers, multiplexes it with a reference signal (e.g., a pilot) in the time and / or frequency domains, and subsequently uses inverse fast Fourier transform (IFFT) to generate a physical channel carrying the time-domain multicarrier symbol stream. Multi-antenna transmit processor 471 then performs transmit analog precoding / beamforming operations on the time-domain multicarrier symbol stream. Each transmitter 418 converts the baseband multicarrier symbol stream provided by the multi-antenna transmitter processor 471 into an radio frequency stream, which is then provided to different antennas 420.

[0129] In the transmission from the second communication device 410 to the first communication device 450, at the first communication device 450, each receiver 454 receives a signal through its corresponding antenna 452. Each receiver 454 recovers the information modulated onto the radio frequency carrier and converts the radio frequency stream into a baseband multicarrier symbol stream, which is then provided to the receiver processor 456. The receiver processor 456 and the multi-antenna receiver processor 458 implement various signal processing functions of the PHY layer. The multi-antenna receiver processor 458 performs receive analog precoding / beamforming operations on the baseband multicarrier symbol stream from the receiver 454. The receiver processor 456 uses a Fast Fourier Transform (FFT) to convert the baseband multicarrier symbol stream after the receive analog precoding / beamforming operations from the time domain to the frequency domain. In the frequency domain, the physical layer data signal and the reference signal are demultiplexed by the receiver processor 456, where the reference signal is used for channel estimation, and the data signal is recovered in the multi-antenna receiver processor 458 after multi-antenna detection to recover any spatial stream destined for or received by the first communication device 450. Symbols on each spatial stream are demodulated and recovered in the receive processor 456, generating soft decisions. The receive processor 456 then decodes and deinterleaves the soft decisions to recover the upper-layer data and control signals transmitted by the second communication device 410 over the physical channel. The upper-layer data and control signals are then provided to the controller / processor 459. The controller / processor 459 implements the functions of the MAC layer. The controller / processor 459 may be associated with a memory 460 storing program code and data. The memory 460 may be referred to as computer-readable media. During transmission from the second communication device 410 to the first communication device 450, the controller / processor 459 provides packet segmentation, decryption, header decompression, and control signal processing to recover the upper-layer data packets from the second communication device 410. The upper-layer data packets are then provided to all protocol layers above the MAC layer.

[0130] In the transmission from the first communication device 450 to the second communication device 410, at the first communication device 450, upper-layer data packets are provided to the controller / processor 459 using a data source 467. The data source 467 represents all protocol layers above the MAC layer. Similar to the transmission function at the second communication device 410 described in the transmission from the second communication device 410 to the first communication device 450, the controller / processor 459 performs header compression, encryption, packet segmentation and reordering, and mapping between services and links. The controller / processor 459 is also responsible for retransmitting lost packets and signaling to the second communication device 410. The transmit processor 468 performs modulation mapping and channel coding processing, and the multi-antenna transmit processor 457 performs digital multi-antenna spatial precoding and beamforming processing. Subsequently, the transmit processor 468 modulates the resulting spatial stream into a multi-carrier symbol stream, which undergoes analog precoding / beamforming operations in the multi-antenna transmit processor 457 before being provided to different antennas 452 via the transmitter 454. Each transmitter 454 first converts the baseband symbol stream provided by the multi-antenna transmitter processor 457 into a radio frequency symbol stream, and then provides it to the antenna 452.

[0131] In the transmission from the first communication device 450 to the second communication device 410, the function at the second communication device 410 is similar to the receiving function at the first communication device 450 described in the transmission from the second communication device 410 to the first communication device 450. Each receiver 418 receives radio frequency signals through its corresponding antenna 420, converts the received radio frequency signals into baseband signals, and provides the baseband signals to the multi-antenna receiving processor 472 and the receiving processor 470. The receiving processor 470 and the multi-antenna receiving processor 472 jointly implement the functions of the PHY layer. The controller / processor 475 implements the functions of the MAC layer. The controller / processor 475 may be associated with a memory 476 that stores program code and data. The memory 476 may be referred to as computer-readable media. In the transmission from the first communication device 450 to the second communication device 410, the controller / processor 475 provides packet segmentation, decryption, header decompression, and control signal processing to recover the upper-layer data packets from the first communication device 450. Upper-layer packets from the controller / processor 475 can be provided to all protocol layers above the Internet or MAC layer.

[0132] As one embodiment, the first communication device 450 includes: at least one processor and at least one memory, the at least one memory including computer program code; the at least one memory and the computer program code are configured to be used with the at least one processor, and the first communication device 450 at least: receives a first MAC frame, wherein the MAC header of the first MAC frame includes the first BSSID, and the first MAC frame indicates the deactivation time of a first AI model for a second BSSID; sends a second MAC frame, wherein the second MAC frame requests the download of the first AI model; receives at least one MAC frame, wherein the at least one MAC frame includes the first AI model; wherein the request of the second MAC frame depends on the indication of the first MAC frame.

[0133] As one embodiment, the first communication device 450 includes: a memory storing a computer-readable instruction program that, when executed by at least one processor, produces actions including: receiving a first MAC frame, wherein the MAC header of the first MAC frame includes a first BSSID, and the first MAC frame indicates a deactivation time for a first AI model for a second BSSID; sending a second MAC frame, wherein the second MAC frame requests the download of the first AI model; and receiving at least one MAC frame, wherein the at least one MAC frame includes the first AI model; wherein the request of the second MAC frame depends on the indication of the first MAC frame.

[0134] As one embodiment, the second communication device 410 includes: at least one processor and at least one memory, the at least one memory including computer program code; the at least one memory and the computer program code are configured to be used with the at least one processor. The second communication device 410 at least: receives a fifth MAC frame; wherein the MAC header of the fifth MAC frame includes a second BSSID; the fifth MAC frame indicates the deactivation time of a first AI model for the second BSSID; transmits a first MAC frame, wherein the MAC header of the first MAC frame includes the first BSSID, and the first MAC frame indicates the deactivation time of the first AI model for the second BSSID; receives a second MAC frame, wherein the second MAC frame requests the download of the first AI model; transmits at least one MAC frame, wherein the at least one MAC frame includes the first AI model; wherein the request of the second MAC frame depends on the indication of the first MAC frame; or, the second communication device 410... The apparatus includes at least: transmitting a fifth MAC frame, wherein the MAC header of the fifth MAC frame includes a second BSSID, and the fifth MAC frame indicates a deactivation time for a first AI model for the second BSSID; receiving a third MAC frame; receiving a fourth MAC frame before the deactivation time for the first AI model for the second BSSID; and, in response to reaching the deactivation time for the first AI model for the second BSSID, ceasing processing of MAC frames for inference based on the first AI model; wherein the MAC header of the third MAC frame includes the second BSSID; the third MAC frame indicates support for the first AI model, and the reception of the third MAC frame follows the transmission of the fifth MAC frame; and the MAC header of the fourth MAC frame includes the second BSSID.

[0135] As one embodiment, the second communication device 410 includes: a memory storing a computer-readable instruction program that, when executed by at least one processor, produces actions including: receiving a fifth MAC frame; wherein the MAC header of the fifth MAC frame includes a second BSSID; the fifth MAC frame indicates a deactivation time of a first AI model for the second BSSID; sending a first MAC frame, wherein the MAC header of the first MAC frame includes the first BSSID, and the first MAC frame indicates the deactivation time of the first AI model for the second BSSID; receiving a second MAC frame, wherein the second MAC frame requests the download of the first AI model; sending at least one MAC frame, wherein the at least one MAC frame includes the first AI model; wherein the request of the second MAC frame depends on the first MAC frame. The C-frame indicates; or, the action includes: sending a fifth MAC frame, wherein the MAC header of the fifth MAC frame includes a second BSSID, the fifth MAC frame indicating a deactivation time for a first AI model for the second BSSID; receiving a third MAC frame; receiving a fourth MAC frame before the deactivation time for the first AI model for the second BSSID; and, in response to reaching the deactivation time for the first AI model for the second BSSID, stopping processing MAC frames for inference based on the first AI model; wherein the MAC header of the third MAC frame includes the second BSSID; the third MAC frame indicates support for the first AI model, the reception of the third MAC frame follows the transmission of the fifth MAC frame; and the MAC header of the fourth MAC frame includes the second BSSID.

[0136] As one embodiment, the first communication device 450 corresponds to the first node in this application, and the second communication device 410 corresponds to the second node or the third node in this application.

[0137] As an example, the first communication device 450 is a non-AP STA, and the second communication device 410 is an access point.

[0138] As an example, the first communication device 450 is an MLD.

[0139] As one embodiment, the second communication device 410 is an access point.

[0140] As one embodiment, the second communication device 410 is an MLD.

[0141] As one embodiment, the second communication device 410 is a base station.

[0142] As one embodiment, the second communication device 410 is a base station distribution unit.

[0143] As one embodiment, the second communication device 410 is a piece of code in the distribution unit of a base station.

[0144] As one embodiment, at least one of the antenna 452, the transmitter 454, the multi-antenna transmitter processor 457, the transmitter processor 468, or the controller / processor 459 is used to transmit the second, third, or fourth MAC frame in this application.

[0145] As an example, at least one of the antenna 452, the receiver 454, the multi-antenna receiving processor 458, the receiving processor 456, or the controller / processor 459 is used to receive the first MAC frame and the at least one MAC frame in this application.

[0146] As an example, at least one of the antenna 420, the receiver 418, the multi-antenna receiving processor 472, the receiving processor 470, or the controller / processor 475 is used to receive the second, third, or fourth MAC frame in this application.

[0147] As an example, at least one of the antenna 420, the transmitter 418, the multi-antenna transmitter processor 471, the transmitter processor 416, or the controller / processor 475 is used to transmit the first MAC frame and the at least one MAC frame in this application.

[0148] Example 5

[0149] Example 5 illustrates a communication flowchart of a first node, a second node, and a third node according to an embodiment of this application, as shown in the attached diagram. Figure 5 As shown. (Attached) Figure 5 In the diagram, the steps in box F1 and box F2 are optional.

[0150] For the first node N1, in step S5101, a first MAC frame is received, wherein the MAC header of the first MAC frame includes the first BSSID, and the first MAC frame indicates the deactivation time of the first AI model for the second BSSID; in step S5102, a second MAC frame is sent, wherein the second MAC frame requests the download of the first AI model; in step S5103, at least one MAC frame is received, wherein the at least one MAC frame includes the first AI model; in step S5104, a third MAC frame is sent; in step S5105, a fourth MAC frame is sent before the deactivation time of the first AI model for the second BSSID; in step S5106, in response to the arrival of the deactivation time of the first AI model for the second BSSID, inference based on the first AI model is stopped;

[0151] For the second node N2, in step S5200, a fifth MAC frame is received; wherein the MAC header of the fifth MAC frame includes a second BSSID; the fifth MAC frame indicates the deactivation time of the first AI model for the second BSSID; in step S5201, a first MAC frame is sent, wherein the MAC header of the first MAC frame includes the first BSSID, and the first MAC frame indicates the deactivation time of the first AI model for the second BSSID; in step S5202, a second MAC frame is received, wherein the second MAC frame requests the download of the first AI model; in step S5203, at least one MAC frame is sent, wherein the at least one MAC frame includes the first AI model;

[0152] For the third node N3, in step S5300, a fifth MAC frame is sent, wherein the MAC header of the fifth MAC frame includes a second BSSID, and the fifth MAC frame indicates the deactivation time of the first AI model for the second BSSID; in step S5301, a third MAC frame is received; in step S5302, a fourth MAC frame is received before the deactivation time of the first AI model for the second BSSID; in step S5303, in response to the arrival of the deactivation time of the first AI model for the second BSSID, processing of MAC frames for inference based on the first AI model is stopped.

[0153] In Example 5, the request of the second MAC frame depends on the indication of the first MAC frame; the MAC header of the third MAC frame includes the second BSSID; the third MAC frame indicates support for the first AI model; the third MAC frame is sent after the reception of at least one MAC frame; the MAC header of the fourth MAC frame includes the second BSSID; the MAC header of the third MAC frame includes the second BSSID; the third MAC frame indicates support for the first AI model; the third MAC frame is received after the transmission of the fifth MAC frame; the MAC header of the fourth MAC frame includes the second BSSID.

[0154] Typically, the Address 3 field of the first MAC frame and the second MAC frame both indicate the first BSSID, and the Address 3 field of the third MAC frame, the fourth MAC frame and the fifth MAC frame indicate the second BSSID; the first node N1 is a non-AP site, and the second node N2 and the third node N3 are the APs indicated by the first BSSID and the second BSSID, respectively.

[0155] Typically, the first node is reconnected from the second node N2 to the third node N3 after step S5103.

[0156] As an example, the second node N2 downloads the first AI model from the storage node of the AI ​​model (e.g., a remote server or access controller) and then distributes the first AI model through the at least one MAC frame.

[0157] As an example, the third node N3 is a neighbor AP of the second node N2.

[0158] As an example, the second MAC frame triggers the transmission of the at least one MAC frame, where the second MAC frame is a management frame.

[0159] As an example, the second MAC frame requests the first AI model, and the at least one MAC frame is a response to the second MAC frame.

[0160] As an example, the first MAC frame includes an identifier of the first AI model; the identifier of the first AI model includes a first part and a second part, the first part being a MAC address, an SSID (Service Set Identifier), or an IP address, and the second part identifying the first AI model in the AI ​​model corresponding to the first part; the second MAC frame includes the second part.

[0161] As an example, the identifier of the first AI model includes a first part and a second part. The first part is a MAC address, an SSID (Service Set Identifier), or an IP address. The second part identifies the first AI model in the AI ​​model corresponding to the first part.

[0162] As an example, the MAC address is the address of the server used to train the first AI model, or the address of the server used to store the first AI model, or the address of the AC corresponding to the AP associated with the first node.

[0163] As an example, the service set indicated by the SSID includes the first node.

[0164] As one embodiment, the IP address is either the address of a server used to train the first AI model or the address of a server used to store the first AI model.

[0165] Typically, the second part comprises fewer bits than the first part; for example, the size of the second part is one byte.

[0166] The above method can save the load overhead of the second MAC frame.

[0167] As an example, both the first MAC frame and the second MAC frame include the identifier of the first AI model.

[0168] As an example, the first MAC frame is a beacon frame, and the second MAC frame is an association request frame.

[0169] The above embodiments allow the first node N1 to obtain the first AI model as early as possible, so as to start AI inference as early as possible.

[0170] As an example, the fifth MAC frame is a management frame or an action frame.

[0171] As an example, the fifth MAC frame is a beacon frame or a probe response frame.

[0172] The above embodiments have good compatibility and enable the recipient to know the deactivation time of the first AI model for the second BSSID as early as possible.

[0173] As an example, the third MAC frame indicating support for the first AI model includes: the third MAC frame indicating that the first node N1 has downloaded the first AI model.

[0174] As an example, the third MAC frame is a reassociation request frame, which includes the identifier of the first AI model.

[0175] The above embodiments enable the third node N3 to provide services based on the first AI model to the first node N1 as early as possible.

[0176] As a sub-implementation of the above embodiment, in response to receiving the third MAC frame, the third node N3 sends a third response MAC frame to the first node N1, wherein the third response MAC frame is a reconnection response frame. (Appendix) Figure 5 The third response MAC frame is omitted from the text.

[0177] The above embodiments allow the first node N1 to continue using the first AI model after reconnection.

[0178] As one example, the first node N1 and the second node N2 are a non-AP site and an AP, respectively, and the third node N3 is a neighboring AP of the second node N2.

[0179] Depending on the application scenario or the supplier, the first AI model may be used to perform different functions. Some non-limiting implementation methods are given below.

[0180] As an example, the first AI model is a beamforming matrix encoder. The input of the first AI model includes a beamforming matrix measured by, for example, listening to a PPDU (Physical Layer Protocol Data Unit). The output of the first AI model is compressed feedback information. Correspondingly, the third node N3 maintains a decoder to restore the compressed feedback information into the beamforming matrix.

[0181] As a sub-implementation of the above embodiments, stopping inference based on the first AI model includes stopping the transmission of MAC frames including the inference output of the first AI model, or includes deleting the first AI model; stopping the processing of MAC frames based on the inference of the first AI model includes discarding the received MAC frames based on the inference of the first AI model, or includes deleting the stored first AI model.

[0182] In the above embodiments, when the deactivation time of the first AI model for the second BSSID is reached, the first node N1 stops inference based on the first AI model (i.e. stops sending the compressed feedback information), and the third node N3 synchronously stops processing the MAC frames based on the inference of the first AI model (i.e. stops decoding the compressed feedback information using the decoder).

[0183] As an example, the first AI model is an AI agent. The input of the first AI model includes wireless measurement results, such as beacon measurements, frame measurements, channel load measurements, site statistics measurements, noise histogram measurements, etc. The output of the first AI model includes an intent. The third node N3 maintains a second AI model, which can interpret the intent output by the first AI model and perform the required operations according to the intent.

[0184] As a sub-implementation of the above embodiments, stopping inference based on the first AI model includes stopping the transmission of MAC frames including the inference output of the first AI model, or includes deleting the first AI model; stopping the processing of MAC frames based on the inference of the first AI model includes discarding the received MAC frames based on the inference of the first AI model, or includes deleting the stored second AI model.

[0185] Example 6

[0186] Example 6 illustrates a flowchart of the transmission of a first AI model according to an embodiment of this application, as shown in the attached diagram. Figure 6 As shown.

[0187] In step S6401, the fourth node N4 sends the first AI model;

[0188] In step S6201, the second node N2 receives the first AI model; in step S6202, it sends at least one MAC frame.

[0189] The first node N1 receives the at least one MAC frame in step S6101;

[0190] In Example 6, the at least one MAC frame includes the first AI model; the first node N1 is a non-AP site, and the second node N2 is an AP.

[0191] In a third implementation, the first AI model is trained on the fourth node N4, or the first AI model is stored on the fourth node N4, or the fourth node N4 is an access controller for managing the second node N2.

[0192] As an example of the third implementation, the identifier of the first AI model includes the MAC address of the fourth node N4 or the IP address of the fourth node N4.

[0193] As a sub-implementation of the above embodiment, the identifier of the first AI model includes a first part and a second part. The first part is the MAC address or IP address of the fourth node N4, and the second part uniquely identifies an AI model associated with the fourth node N4. For example, the second part consists of 8 bits, and the fourth node N4 is associated with no more than 256 AI models.

[0194] As a sub-example of the above embodiment, the APs indicated by the first BSSID and the second BSSID are managed by the fourth node N4.

[0195] In the fourth implementation, the fourth node N4 is a neighboring AP of the second node N2, and the first AI model in step S6401 is sent through one or more MAC frames.

[0196] Example 7

[0197] Example 7 illustrates a schematic diagram of a reference encoder and a reference decoder according to an embodiment of this application, as shown in the attached diagram. Figure 7 As shown. (Attached) Figure 7 In the table, parameters or functions identified by S800, S801, S802, ..., S808 belong to the reference encoder, and parameters or functions identified by S900, S901, S902, ..., S908 belong to the reference decoder.

[0198] The reference encoder and the reference decoder are inverse operations of each other, both employing a Transformer structure; the parameters and functions identified in block F800 represent a Transformer block, and the reference encoder typically includes multiple cascaded Transformer blocks. Figure 7 Only one example is shown; similarly, the parameters and functions identified in box F900 constitute a Transformer block, and the reference decoder typically comprises multiple cascaded Transformer blocks, as shown below. Figure 7 Only one example is shown in the text.

[0199] In Example 7, the first AI model is a beamforming matrix encoder, which is typically used in beamforming matrix compression scenarios; it is a typical deployment of two AI models, that is, the first AI model is deployed on the beamformer side, and the corresponding decoder is deployed on the beamformer side.

[0200] The first AI model can adopt an appendix Figure 7 The structure of the reference encoder described in the document, or to implement a similar function; the second node deploys attached... Figure 7 The reference decoder described above, or one that implements similar functionality, typically involves deploying a first AI model on the first node side and a first decoder on the second node side; the first AI model and the first decoder are two-sided AI models. (See appendix.) Figure 7 The reference encoder and the reference decoder can be used to train or calibrate the first AI model and the first decoder. However, the first AI model and the first decoder used for actual inference do not need to completely reuse the structure / parameters / hyperparameters of the reference encoder and the reference decoder.

[0201] As an example, the first node trains the parameter set of the first AI model based on the parameter set of the reference decoder. The trained first AI model and the reference decoder are not necessarily completely inverse operations of each other, as long as the output of the inference data used for performance testing after passing through the first AI model and the reference decoder in sequence meets the error requirements of the inference data.

[0202] The KPIs (Key Performance Indicators) used to calculate errors may vary depending on the training algorithm or AI model structure. Common KPIs include the performance requirements of GCS (Generalized Cosine Similarity); SGCS (Squared Generalized Cosine Similarity); and NMSE (Normalized Mean Squared Error).

[0203] Since the reference decoder can be considered the inverse operation of the reference encoder, the specific implementation of the reference encoder is given below, while the reference decoder will not be described in detail.

[0204] The beamforming vector of the first measurement in S800, such as a column vector of a beamforming matrix, is input to the embedding layer S801; the feature dimension of each input is T, where T is the number of antenna ports in the first RS resource set; each input is a T-dimensional complex vector, which can be represented by two real vectors representing the real part and the imaginary part respectively; it is assumed below that each token is described as a T-dimensional real vector;

[0205] After the S801 embedding layer, the embedded representation of each token is a A dimensional real vector;

[0206] The dimension of the S802 location information is the same as the dimension of the embedded representation of the token;

[0207] The dimension of the Transformer block in F800 is as described The S803 normalization, S804 multi-head attention, S805 normalization, and S806 multi-head perception, which are included, adopt the classic Transformer structure.

[0208] The S807 linear layer, also known as the output linear layer, functions to output P1* tokens representing P1 tokens. Transform the matrix into For 3D vectors, typical transformation methods include linear operations using matrices; generally speaking, Less than That is, the S807 linear layer has completed the dimensionality reduction function;

[0209] The function of S808 vector quantization is to... A real vector of dimension is discretized into entries or codewords in a finite, learnable quantized codebook, i.e., appended... Figure 7 The first inference output in the process.

[0210] It should be noted that although the S808 is called Vector Quantization (VQ), it also supports SQ (Scalar Quantization).

[0211] As an example, the The real reference (SQ) quantization codebook for each dimension of the dimensional real vector is quantized into bits, where L is the The at least one codeword is L codewords, and the payload size of the first inference output is [missing value]. 1 bit.

[0212] As an example, the A 3D real vector is divided into Each paragraph includes a real number; the aforementioned Each segment in the segment is mapped to a codeword in the (VQ) quantization codebook according to criteria such as minimum SGSC or minimum NMSE; the mapped codeword corresponds to a... A dimensional vector, where each dimension corresponds to bits; that is, the output of S808 is _ bits; that is, the output of S808 is indivual dimensional vector, each of the The load size of the dimensional vector is 1 bit.

[0213] As a sub-implementation of the above embodiments, the at least one codeword is a codeword, and L is the... The load size of the first inference output is 1 bit.

[0214] As a sub-implementation of the above embodiments, the The value is 1, meaning that L is the... The load size of the first inference output is the same as the load size of the output of S808.

[0215] Provided that the inference performance meets the requirements, for example, if the error between the first recovered beamforming vector in S908 and the first measured beamforming vector in S800 meets the KPI requirements (e.g., does not exceed a specific threshold), the first AI model can adopt the same structure as the reference encoder or a different structure; this is determined by the supplier of the first node N1. The reference encoder and the reference decoder are used to train the first AI model and are not necessarily used for actual inference.

[0216] As an example, when the number of column vectors in the beamforming matrix to be compressed is greater than 1, the operation in Example 7 is performed once for each column vector.

[0217] Example 8

[0218] Example 8 illustrates a schematic diagram of reconnection to a second BSSID according to one embodiment of this application, as shown in the attached diagram. Figure 8 As shown.

[0219] In Example 8, when the first node receives the first MAC frame, it connects to or resides in the first BSSID (the indicated AP); then it reconnects to the second BSSID (the indicated AP, i.e., the second node), and the deactivation time of the first AI model for the second BSSID indicated by the first MAC frame occurs after reconnecting to the AP indicated by the second BSSID (the third node).

[0220] In Example 8, after the first node downloads the first AI model from the second node, it can continue to use the first AI model after reconnecting to the third node and before the deactivation time of the first AI model for the second BSSID.

[0221] The first node determines to send a request to download the first AI model based on the deactivation time of the first AI model for the second BSSID.

[0222] As a sub-implementation of the above embodiment, the first MAC frame indicates the time interval between the start transmission time of the first MAC frame and the deactivation time of the first AI model for the second BSSID.

[0223] Example 9

[0224] Example 9 illustrates a structural block diagram of a processing apparatus in a first node according to an embodiment of this application, as shown in the attached diagram. Figure 9 As shown. In the appendix Figure 9 In this context, the first node 1600 includes a first receiver 1601 and a first transmitter 1602; the first node 1600 is a station or a non-access point station.

[0225] The first receiver 1601 receives a first MAC frame, wherein the MAC header of the first MAC frame includes the first BSSID, and the first MAC frame indicates the deactivation time of the first AI model for the second BSSID.

[0226] The first transmitter 1602 sends a second MAC frame, wherein the second MAC frame requests the download of the first AI model;

[0227] The first receiver 1601 receives at least one MAC frame, wherein the at least one MAC frame includes the first AI model;

[0228] In Example 9, the request of the second MAC frame depends on the indication of the first MAC frame.

[0229] As one embodiment, the first transmitter 1602 sends a third MAC frame; and sends a fourth MAC frame before the deactivation time of the first AI model for the second BSSID;

[0230] The third MAC frame includes the second BSSID in its MAC header; the third MAC frame indicates support for the first AI model; the third MAC frame is sent after the reception of at least one MAC frame; the fourth MAC frame includes the second BSSID in its MAC header.

[0231] As an example, the first transmitter 1602 stops inference based on the first AI model in response to the deactivation time of the first AI model for the second BSSID.

[0232] As an example, the first MAC frame indicates at least one of the input format or output format of the first AI model.

[0233] As an example, the site supporting the first AI model must have the capability to support wireless measurement.

[0234] As one embodiment, the first receiver 1601 includes the appendix to this application. Figure 4 The receiver 454 (including antenna 452), receiver processor 456, multi-antenna receiver processor 458, and controller / processor 459 are included.

[0235] As one embodiment, the first receiver 1601 includes the appendix to this application. Figure 4 The receiver 454 (including antenna 452), the receiver processor 456, the multi-antenna receiver processor 458, or the controller / processor 459 are at least one of them.

[0236] As one embodiment, the first transmitter 1602 includes the appendix to this application. Figure 4 The transmitter 454 (including antenna 452), the transmitter processor 468, the multi-antenna transmitter processor 457, and the controller / processor 459 are included.

[0237] As one embodiment, the first transmitter 1602 includes the appendix to this application. Figure 4 The receiver 454 (including antenna 452), the transmitter processor 468, the multi-antenna transmitter processor 457, or the controller / processor 459 are at least one of them.

[0238] Example 10

[0239] Example 10 illustrates a structural block diagram of a processing apparatus in a second node according to an embodiment of this application, as shown in the attached diagram. Figure 10 As shown. In the appendix Figure 10 In the second node processing device 1700, there are a second transmitter 1701 and a second receiver 1702; the second node 1700 is a station or an access point.

[0240] The second receiver 1702 receives a fifth MAC frame; wherein, the MAC header of the fifth MAC frame includes a second BSSID; the fifth MAC frame indicates the deactivation time of the first AI model for the second BSSID;

[0241] The second transmitter 1701 sends a first MAC frame, wherein the MAC header of the first MAC frame includes the first BSSID, and the first MAC frame indicates the deactivation time of the first AI model for the second BSSID;

[0242] The second receiver 1702 receives a second MAC frame, wherein the second MAC frame requests the download of the first AI model;

[0243] The second transmitter 1701 transmits at least one MAC frame, wherein the at least one MAC frame includes the first AI model;

[0244] In Example 10, the request of the second MAC frame depends on the indication of the first MAC frame.

[0245] As an example, the first MAC frame indicates at least one of the input format or output format of the first AI model.

[0246] As one embodiment, the second receiver 1702 includes the appendix to this application. Figure 4 The receiver 418 (including antenna 420), receiver processor 470, multi-antenna receiver processor 472, and controller / processor 475 are included.

[0247] As one embodiment, the second receiver 1702 includes the appendix to this application. Figure 4 The receiver 418 (including antenna 420), the receiver processor 470, the multi-antenna receiver processor 472, and the controller / processor 475 are at least one of them.

[0248] As one embodiment, the second transmitter 1701 includes the appendix to this application. Figure 4 The transmitter 418 (including antenna 420), the transmitter processor 416, the multi-antenna transmitter processor 471, and the controller / processor 475 are included.

[0249] As one embodiment, the second transmitter 1701 includes the appendix to this application. Figure 4The transmitter 418 (including antenna 420), the transmission processor 416, the multi-antenna transmission processor 471, or the controller / processor 475 are at least one of them.

[0250] Example 11

[0251] Example 11 illustrates a structural block diagram of a processing apparatus in a third node according to an embodiment of this application, as shown in the attached diagram. Figure 11 As shown. In the appendix Figure 11 In this context, the third node processing device 1800 includes a third transmitter 1801 and a third receiver 1802; the third node 1800 is a station or an access point.

[0252] The third transmitter 1801 sends a fifth MAC frame, wherein the MAC header of the fifth MAC frame includes a second BSSID, and the fifth MAC frame indicates the deactivation time of the first AI model for the second BSSID;

[0253] The third receiver 1802 receives a third MAC frame; receives a fourth MAC frame before the deactivation time of the first AI model for the second BSSID; and stops processing MAC frames based on the inference of the first AI model in response to the arrival of the deactivation time of the first AI model for the second BSSID.

[0254] In Example 11, the MAC header of the third MAC frame includes the second BSSID; the third MAC frame indicates support for the first AI model, and the reception of the third MAC frame follows the transmission of the fifth MAC frame; the MAC header of the fourth MAC frame includes the second BSSID.

[0255] As one embodiment, the second receiver 1702 includes the appendix to this application. Figure 4 The receiver 418 (including antenna 420), receiver processor 470, multi-antenna receiver processor 472, and controller / processor 475 are included.

[0256] As one embodiment, the second receiver 1702 includes the appendix to this application. Figure 4 The receiver 418 (including antenna 420), the receiver processor 470, the multi-antenna receiver processor 472, and the controller / processor 475 are at least one of them.

[0257] As one embodiment, the second transmitter 1701 includes the appendix to this application. Figure 4 The transmitter 418 (including antenna 420), the transmitter processor 416, the multi-antenna transmitter processor 471, and the controller / processor 475 are included.

[0258] As one embodiment, the second transmitter 1701 includes the appendix to this application. Figure 4 The transmitter 418 (including antenna 420), the transmission processor 416, the multi-antenna transmission processor 471, or the controller / processor 475 are at least one of them.

[0259] Those skilled in the art will understand that all or part of the steps in the above methods can be implemented by a program instructing related hardware, and the program can be stored in a computer-readable storage medium, such as a read-only memory, hard disk, or optical disk. Optionally, all or part of the steps in the above embodiments can also be implemented using one or more integrated circuits. Correspondingly, each module unit in the above embodiments can be implemented in hardware or in the form of software functional modules. This application is not limited to any specific combination of software and hardware. The first node in this application includes, but is not limited to, wireless communication devices such as mobile phones, tablets, laptops, network cards, low-power devices, vehicle communication devices, aircraft, airplanes, drones, and remote-controlled airplanes. The second node in this application includes, but is not limited to, wireless communication devices such as access points, wireless routers, sites or handheld terminals with access point functions, and transmission and reception points (TRPs). The above descriptions are merely preferred embodiments of this application and are not intended to limit the scope of protection of this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A first node used for wireless communication, characterized in that, include: A first receiver receives a first MAC frame, wherein the MAC header of the first MAC frame includes a first BSSID, and the first MAC frame indicates the deactivation time of a first AI model for a second BSSID. The first transmitter sends a second MAC frame, wherein the second MAC frame requests the download of the first AI model; The first receiver receives at least one MAC frame, wherein the at least one MAC frame includes the first AI model; The request in the second MAC frame depends on the indication in the first MAC frame.

2. The first node according to claim 1, characterized in that, include: The first transmitter sends a third MAC frame; and sends a fourth MAC frame before the deactivation time of the first AI model for the second BSSID; The MAC header of the third MAC frame includes the second BSSID; The third MAC frame indicates support for the first AI model, and the third MAC frame is sent after the reception of at least one MAC frame; the MAC header of the fourth MAC frame includes the second BSSID.

3. The first node according to claim 2, characterized in that, include: The first transmitter, in response to the deactivation time of the first AI model arriving at the second BSSID, stops inference based on the first AI model.

4. The first node according to any one of claims 1 to 3, characterized in that, The first MAC frame indicates at least one of the input format or output format of the first AI model.

5. The first node according to claim 4, characterized in that, Sites supporting the first AI model must have the capability to perform wireless measurements.

6. A second node used for wireless communication, characterized in that, include: The second receiver receives the fifth MAC frame; wherein the MAC header of the fifth MAC frame includes the second BSSID; the fifth MAC frame indicates the deactivation time of the first AI model for the second BSSID; The second transmitter sends a first MAC frame, wherein the MAC header of the first MAC frame includes a first BSSID, and the first MAC frame indicates the deactivation time of the first AI model for the second BSSID; The second receiver receives the second MAC frame, wherein the second MAC frame requests the download of the first AI model; The second transmitter sends at least one MAC frame, wherein the at least one MAC frame includes the first AI model; The request in the second MAC frame depends on the indication in the first MAC frame.

7. A third node used for wireless communication, characterized in that, include: The third transmitter sends a fifth MAC frame, wherein the MAC header of the fifth MAC frame includes a second BSSID, and the fifth MAC frame indicates the deactivation time of the first AI model for the second BSSID; A third receiver receives a third MAC frame; receives a fourth MAC frame before the deactivation time of the first AI model for the second BSSID; and, in response to the arrival of the deactivation time of the first AI model for the second BSSID, stops processing MAC frames based on inference from the first AI model. The third MAC frame includes the second BSSID in its MAC header; the third MAC frame indicates support for the first AI model; the third MAC frame is received after the fifth MAC frame is sent; the fourth MAC frame includes the second BSSID in its MAC header.

8. A method used in a first node of wireless communication, characterized in that, include: Receive a first MAC frame, wherein the MAC header of the first MAC frame includes a first BSSID, and the first MAC frame indicates the deactivation time of a first AI model for a second BSSID; Send a second MAC frame, wherein the second MAC frame requests the download of the first AI model; Receive at least one MAC frame, wherein the at least one MAC frame includes the first AI model; The request in the second MAC frame depends on the indication in the first MAC frame.

9. A method used in a second node of wireless communication, characterized in that, include: Receive a fifth MAC frame; wherein the MAC header of the fifth MAC frame includes a second BSSID; the fifth MAC frame indicates the deactivation time of the first AI model for the second BSSID; Send a first MAC frame, wherein the MAC header of the first MAC frame includes a first BSSID, and the first MAC frame indicates the deactivation time of the first AI model for the second BSSID; Receive a second MAC frame, wherein the second MAC frame requests the download of the first AI model; Send at least one MAC frame, wherein the at least one MAC frame includes the first AI model; The request in the second MAC frame depends on the indication in the first MAC frame.

10. A method used in a third node of wireless communication, characterized in that, include: Send a fifth MAC frame, wherein the MAC header of the fifth MAC frame includes a second BSSID, and the fifth MAC frame indicates the deactivation time of the first AI model for the second BSSID; Receive a third MAC frame; receive a fourth MAC frame before the deactivation time of the first AI model for the second BSSID; in response to the arrival of the deactivation time of the first AI model for the second BSSID, stop processing MAC frames for inference based on the first AI model; The third MAC frame includes the second BSSID in its MAC header; the third MAC frame indicates support for the first AI model; the third MAC frame is received after the fifth MAC frame is sent; the fourth MAC frame includes the second BSSID in its MAC header.

Images