Random access event processing method, ai model training method, apparatus, and device

Abstract

The application discloses a random access event processing method, an AI model training method, a device and equipment, and belongs to the technical field of communication. The random access event processing method of the application embodiment comprises the following steps: a first device acquires first output information of a first artificial intelligence (AI) model, wherein the first output information is related to a random access event; and the first device performs a first operation based on the first output information.

Description

Technical Field

[0001] This application belongs to the field of communication technology, specifically relating to a random access event processing method, an AI model training method, an apparatus, and a device. Background Technology

[0002] Currently, many events trigger random access by terminals to maintain synchronization with the network or to allow the terminal to access a suitable cell and obtain system messages. In connected mode, events that trigger random access typically include handover, beam failure recovery, and uplink synchronization loss. These random access events often require a large number of continuous measurements, which consume significant amounts of the terminal's power. Furthermore, these events can also cause data transmission interruptions, leading to decreased throughput and increased data transmission latency for either the terminal or the network. Summary of the Invention

[0003] This application provides a random access event processing method, an AI model training method, an apparatus, and a device that can solve the problem that random access events require a large number of continuous measurements, resulting in high power consumption of communication equipment.

[0004] In a first aspect, a random access event handling method is provided, executed by a first device, the method comprising:

[0005] The first device acquires the first output information of the first artificial intelligence (AI) model, and the first output information is related to the random access event;

[0006] The first device performs a first operation based on the first output information.

[0007] Secondly, an AI model training method is provided, executed by a second device, the method comprising:

[0008] The second device trains the first AI model based on the second input information and sends the trained first AI model to the first device. The trained first AI model is used by the first device to obtain first output information, which is related to the random access event.

[0009] The second input information includes at least one of the following:

[0010] First input information, the first AI model obtains the first output information based on the first input information;

[0011] Labels used for training the first AI model;

[0012] Algorithm or algorithm index used for training the first AI model;

[0013] The loss function used for training the first AI model;

[0014] Adjustment information used for training the first AI model;

[0015] The triggering conditions for training the first AI model.

[0016] Thirdly, a random access event processing apparatus is provided, the apparatus being applied to a first device, the apparatus comprising:

[0017] The acquisition module is used to acquire the first output information of the first AI model, which is related to the random access event;

[0018] An execution module is used to perform a first operation based on the first output information.

[0019] Fourthly, an AI model training device is provided, which is applied to a second device, and the device includes:

[0020] The training module is used to train the first AI model based on the second input information and send the trained first AI model to the first device. The trained first AI model is used by the first device to obtain first output information, which is related to the random access event.

[0021] The second input information includes at least one of the following:

[0022] First input information, the first AI model obtains the first output information based on the first input information;

[0023] Labels used for training the first AI model;

[0024] Algorithm or algorithm index used for training the first AI model;

[0025] The loss function used for training the first AI model;

[0026] Adjustment information used for training the first AI model;

[0027] The triggering conditions for training the first AI model.

[0028] Fifthly, a random access event processing apparatus is provided, the apparatus being configured to perform the steps of the method described in the first aspect.

[0029] In a sixth aspect, an AI model training apparatus is provided, the apparatus being configured to perform the steps of the method described in the second aspect.

[0030] In a seventh aspect, a communication device is provided, the communication device including a processor and a memory, the memory storing a program or instructions executable on the processor, the program or instructions, when executed by the processor, implementing the steps of the method as described in the first aspect, or implementing the steps of the method as described in the second aspect.

[0031] Eighthly, a communication device is provided, including a processor and a communication interface, wherein the processor is configured to acquire first output information of a first AI model, the first output information being related to a random access event; and to perform a first operation based on the first output information; or,

[0032] The processor is used to train a first AI model based on second input information and send the trained first AI model to a first device. The trained first AI model is used by the first device to obtain first output information, which is related to a random access event.

[0033] The second input information includes at least one of the following:

[0034] First input information, the first AI model obtains the first output information based on the first input information;

[0035] Labels used for training the first AI model;

[0036] Algorithm or algorithm index used for training the first AI model;

[0037] The loss function used for training the first AI model;

[0038] Adjustment information used for training the first AI model;

[0039] The triggering conditions for training the first AI model.

[0040] A ninth aspect provides a readable storage medium on which a program or instructions are stored, which, when executed by a processor, implement the steps of the method described in the first aspect, or implement the steps of the method described in the second aspect.

[0041] In a tenth aspect, a wireless communication system is provided, comprising: a first device and a second device, wherein the first device is configured to perform the steps of the method as described in the first aspect, and the second device is configured to perform the steps of the method as described in the second aspect.

[0042] Eleventhly, a chip is provided, the chip including a processor and a communication interface, the communication interface being coupled to the processor, the processor being used to run programs or instructions to implement the method as described in the first aspect, or to implement the method as described in the second aspect.

[0043] In a twelfth aspect, a computer program / program product is provided, which is stored in a storage medium and is executed by at least one processor to implement the steps of the method as described in the first aspect, or to implement the steps of the method as described in the second aspect.

[0044] In this embodiment, the first device can perform the first operation based on the first output information output by the first AI model. The first operation is an operation related to random access events. Thus, the first device can predict the possible time period or non-occurrence period of future random access events (such as handover, beam failure recovery, uplink synchronization failure, wireless link failure, LBT failure, small data transmission, SR failure, on-demand signal activation or deactivation, uplink data arrival or downlink data arrival, etc.) or the RACH process itself based on the first AI model's inference result (i.e., the first output information). It can also perform the first operation based on the inference result of the first AI model (i.e., the first output information), thereby avoiding unnecessary measurements, reducing the number of RACH occurrences, reducing the energy consumption of the first device, and reducing the probability of RACH collisions by predicting and planning RACH transmission resources in advance, reducing the data transmission interruption time, ensuring the data transmission rate, which is beneficial to improving throughput, reducing latency, and ensuring user experience. Attached Figure Description

[0045] Figure 1a This is a block diagram of a wireless communication system applicable to embodiments of this application;

[0046] Figure 1b This is a block diagram of another wireless communication system to which the embodiments of this application can be applied;

[0047] Figure 1c This is a schematic diagram of a neuron that can be applied to the embodiments of this application;

[0048] Figure 1d This is a diagram illustrating the lifecycle management of an AI model;

[0049] Figure 2 This is a flowchart of a random access event handling method provided in an embodiment of this application;

[0050] Figure 3 This is a flowchart of an AI model training method provided in an embodiment of this application;

[0051] Figure 4 This is a structural diagram of a random access event processing device provided in an embodiment of this application;

[0052] Figure 5This is a structural diagram of an AI model training device provided in an embodiment of this application;

[0053] Figure 6 This is a structural diagram of a communication device provided in an embodiment of this application;

[0054] Figure 7 This is a structural diagram of a terminal provided in an embodiment of this application;

[0055] Figure 8 This is a structural diagram of a network-side device provided in an embodiment of this application;

[0056] Figure 9 This is a structural diagram of another network-side device provided in an embodiment of this application. Detailed Implementation

[0057] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0058] The terms "first," "second," etc., used in this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such terms can be used interchangeably where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first" and "second" are generally of the same class, not limited in number; for example, the first object can be one or more. Furthermore, "or" in this application indicates at least one of the connected objects. For example, the scope of protection for "A or B" covers at least three scenarios: Scenario 1: including A but not B; Scenario 2: including B but not A; Scenario 3: including both A and B. In addition, the terms "A and / or B," "at least one of A and B," and "at least one of A or B" also cover at least the above three scenarios. The character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0059] The term "instruction" in this application can be either a direct instruction (or explicit instruction) or an indirect instruction (or implicit instruction). A direct instruction can be understood as one in which the sender explicitly informs the receiver of specific information, the operation to be performed, or the requested result, etc., in the instruction sent. An indirect instruction can be understood as one in which the receiver determines the corresponding information based on the instruction sent by the sender, or makes a judgment and determines the operation to be performed or the requested result, etc., based on the judgment result.

[0060] It is worth noting that the technologies described in this application are not limited to Long Term Evolution (LTE) / LTE-Advanced (LTE-A) systems, but can also be used in other wireless communication systems, such as Code Division Multiple Access (CDMA), Time Division Multiple Access (TDMA), Frequency Division Multiple Access (FDMA), Orthogonal Frequency Division Multiple Access (OFDMA), Single-carrier Frequency-Division Multiple Access (SC-FDMA), or other systems. The terms "system" and "network" in this application are often used interchangeably, and the described technologies can be used with the systems and radio technologies mentioned above, as well as with other systems and radio technologies. The following description describes New Radio (NR) systems for illustrative purposes, and the term NR is used in most of the following description; however, these technologies can also be applied to systems other than NR systems, such as 6th generation (6G) radio systems. th Generation 6G communication system.

[0061] Figure 1a and Figure 1bThis diagram illustrates two wireless communication systems applicable to embodiments of this application. The wireless communication system includes a terminal 11 and a network-side device 12. The terminal 11 can be a mobile phone, tablet computer, laptop computer, notebook computer, personal digital assistant (PDA), handheld computer, netbook, ultra-mobile personal computer (UMPC), mobile internet device (MID), augmented reality (AR), virtual reality (VR) device, robot, wearable device, flight vehicle, vehicle user equipment (VUE), shipboard equipment, pedestrian user equipment (PUE), smart home devices (home appliances with wireless communication capabilities, such as refrigerators, televisions, washing machines, or furniture), game consoles, personal computers (PCs), ATMs, or self-service machines, etc. Wearable devices include: smartwatches, smart bracelets, smart earphones, smart glasses, smart jewelry (smart bracelets, smart chains, smart rings, smart necklaces, smart anklets, smart anklets, etc.), smart wristbands, smart clothing, etc. Among these, in-vehicle devices can also be referred to as in-vehicle terminals, in-vehicle controllers, in-vehicle modules, in-vehicle components, in-vehicle chips, or in-vehicle units, etc. It should be noted that the specific type of terminal 11 is not limited in the embodiments of this application.

[0062] like Figure 1bAs shown, network-side device 12 may include access network equipment or core network equipment. Access network equipment may also be referred to as Radio Access Network (RAN) equipment, radio access network function, or radio access network unit. Access network equipment may include base stations, Wireless Local Area Network (WLAN) access points (APs), or Wireless Fidelity (WiFi) nodes, etc. The term "base station" can be referred to as Node B (NB), Evolved Node B (eNB), Next Generation Node B (gNB), New Radio Node B (NR Node B), Access Point, Relay Base Station (RBS), Serving Base Station (SBS), Base Transceiver Station (BTS), Radio Base Station, Radio Transceiver, Basic Service Set (BSS), Extended Service Set (ESS), Home Node B (HNB), Home Evolved Node B, Transmit / Receive Point (TRP), or any other suitable term in the relevant field, as long as the same technical effect is achieved. The term "base station" is not limited to any specific technical terminology. It should be noted that this application embodiment only uses a base station in an NR system as an example for description and does not limit the specific type of base station.

[0063] Core network equipment, also known as core network nodes, core network functions, or core network elements, includes, but is not limited to, at least one of the following: Mobility Management Entity (MME), Access and Mobility Management Function (AMF), Session Management Function (SMF), User Plane Function (UPF), Policy Control Function (PCF), Policy and Charging Rules Function (PCRF), Edge Application Server Discovery Function (EASDF), Unified Data Management (UDM), Unified Data Repository (UDR), Home Subscriber Server (HSS), Centralized network configuration (CNC), Network Repository Function (NRF), Network Exposure Function (NEF), Local NEF (or L-NEF), and Binding Support Function. Support Functions (BSF), Application Functions (AF), Location Management Functions (LMF), Gateway Mobile Location Centres (GMLC), and Network Data Analytics Functions (NWDAF), etc. It should be noted that this application embodiment only uses core network equipment in the NR system as an example and does not limit the specific type of core network equipment. If the name of the core network equipment mentioned in this application embodiment changes in subsequent protocol versions (e.g., 6G), it will still be within the scope of protection of this application.

[0064] Optionally, the core network equipment can be implemented by one or more functional modules in a single device, or by multiple devices working together; this application does not specifically limit this. It is understood that the aforementioned functional modules can be network elements in hardware devices, software functional modules running on dedicated hardware, or virtualized functional modules instantiated on a platform (e.g., a cloud platform).

[0065] To better understand, the relevant concepts that may be involved in the embodiments of this application are explained below.

[0066] The purpose and events of random access:

[0067] Random Access Channel (RACH) serves many purposes. For example, random access triggered by a Physical downlink control channel (PDCCH) order is primarily for the UE to obtain uplink time synchronization. Another example is when the UE establishes the initial radio link, the random access procedure can be used to obtain the user identifier—the Cell-Radio Network Temporary Identifier (C-RNTI) information.

[0068] The random access process is typically triggered by one of the events listed in Table 1 below: (Refer to relevant protocols)

[0069] Table 1

[0070]

[0071] The English abbreviations in Table 1 above are explained as follows:

[0072] RRC: Radio Resource Control;

[0073] CBRA: Contention Based Random Access;

[0074] CFRA: Contention-Free Random Access;

[0075] SI: System Information;

[0076] SR: Scheduling Request;

[0077] PUCCH: Physical Uplink Control Channel;

[0078] SCell: Secondary Cell;

[0079] TAG: Timing Advance Group;

[0080] MAC: Medium Access Control.

[0081] Random access procedure:

[0082] In the prior art, random access procedures can be either contention-based random access (CBRA) or contention-free random access (CFRA).

[0083] The CBRA process is as follows: The UE randomly selects a preamble from the contention-based preamble pool shared with other UEs in the cell and transmits it on the Physical Random Access Channel (PRACH). After the network detects the preamble, it sends a Msg2 (Random Access Response (RAR)) message. The RAR contains the number of the preamble detected by the network and the uplink radio resources allocated to the UE to send Msg3. After receiving Msg2, the UE confirms that at least one of the preamble numbers carried in Msg2 matches the number of its own preamble. Then, according to the resources indicated by the RAR, it sends Msg3 containing contention resolution information. After receiving Msg3, the network sends Msg4 containing contention resolution information. After receiving Msg4, the UE confirms that the contention resolution information is consistent with the one it sent in Msg3, thus completing the 4-step random access process. The network includes UL grant information in the RAR to indicate Msg3 Physical Uplink Shared Channel (PUSCH) scheduling information, and also includes information such as the Random Access Channel preamble ID (RAPID), Temporary Cell RNTI (TC-RNTI), and Timing Advance (TA). If the network does not receive the Msg3 PUSCH, it can schedule a retransmission of the Msg3 PUSCH in the PDCCH scrambled by the TC-RNTI. Since different UEs randomly select preambles for transmission in CBRA, different UEs may select the same preamble for transmission on the same time-frequency radio resources (Random Access Hour (RO) resources), which can be understood as a UE preamble conflict.

[0084] In CFRA, the preamble is allocated by the base station. This preamble is called a dedicated random access preamble and is provided to the UE via RRC signaling or PDCCH order signaling, thus eliminating preamble conflicts. When dedicated random access preamble resources are insufficient, the base station notifies the terminal to initiate CBRA-based access.

[0085] Switch:

[0086] Handover Type Classification: 5G NR handover types are divided into inter-system handover and intra-system handover. Inter-system handover is further divided into handover between 5G NR and 4G LTE systems and handover between 5G NR and 3G Wideband Code Division Multiple Access (WCDMA) systems. Intra-system handover is divided into intra-base station handover and inter-base station handover. See Table 2 below for details:

[0087] Table 2

[0088]

[0089]

[0090] Switching steps:

[0091] The switchover is mainly divided into three stages:

[0092] 1. Switching to the preparation phase

[0093] The UE performs measurements on the target cell according to the measurement configuration issued by the gNB and reports the measurement report of the target cell.

[0094] 2. Switching to the judgment phase

[0095] The gNB evaluates the measurement report reported by the UE and determines whether to perform a handover. The handover trigger condition is generally based on the Reference Signal Received Power (RSRP) of the Synchronization Signal Block (SSB), such as ss-RSRP.

[0096] 3. Switch execution phase

[0097] The gNB controls the UE to complete the handover behavior to the target cell.

[0098] The switching measurement modes are divided into measurement mode and blind cut mode:

[0099] (1) Measurement mode: gNB performs handover behavior based on the measurement report reported by UE.

[0100] (2) Blind handover mode: No measurement is performed on the target cell, and the handover is performed directly. This mode is only used when a handover must be initiated as soon as possible, because the possibility of UE accessing a neighboring cell is relatively high.

[0101] Measurement event classification:

[0102] Measurement reports are typically submitted based on coverage (RSRP) and quality (Reference Signal Received Quality (RSRQ) and signal-to-noise and interference ratio (SINR)). The submission methods can be periodic or event-triggered. Periodic submissions mainly report the cell with the strongest coverage, which is the network coverage data. Event-triggered submissions are usually due to handover (or redirection, etc.).

[0103] Ideally, the base station should allow the UE to report the signal quality of the serving cell and neighboring cells, triggering a handover through a single measurement. However, in practice, unnecessary ping-pong handovers can cause overload. To avoid this, the 3GPP specification proposes a set of predefined measurement reporting mechanisms executed by the UE. These predefined measurement reporting types are called "events." The types of "events" that the UE must report are specified by the RRC signaling messages sent by the base station.

[0104] There are typically eight types of measurement events: A1 to A6 (intra-system switching) and B1 to B2 (inter-system switching). A description of each event is shown in Table 3 below. Different measurement times correspond to different functions; A3 is used for intra-frequency switching, while B1 and B2 are used for inter-system switching.

[0105] Table 3

[0106]

[0107] The interpretation of each parameter in the judgment criteria in Table 3 above can be found in the relevant agreements.

[0108] Beam Failure Recovery (BFR):

[0109] When a user terminal (UE) is indoors or conducting mobile communication, the wireless link between the UE and the base station (gNB) is susceptible to radio frequency signal blockage and degradation, leading to beam failure and communication link interruption. Therefore, to detect beam failures as quickly as possible, the UE has implemented mechanisms to measure these sudden and rapid changes in the communication link and restore service as soon as possible. The UE completes this process with the help of beam failure recovery. The beam failure recovery (BFR) process is handled in combination at the UE's physical (PHY) and MAC layers, without involving any higher-layer signaling.

[0110] Primary cell beam failure recovery mechanism:

[0111] In high-frequency communication systems, due to the short wavelength of wireless signals, signal propagation is more easily blocked, leading to signal interruption. Re-establishing the wireless link takes a long time. Therefore, the Primary Cell (PCell) Baseline Frame Reset (BFR) mechanism was introduced in 3GPP Release 15.

[0112] When the downlink beam of a UE's PCell fails, the UE triggers a beam failure recovery request procedure. Specifically, the UE sends a preamble (random access preamble) on the PCell and waits on the PCell for feedback from the network side (i.e., a PDCCH scrambled with C-RNTI (Cell Radio Network Temporary Identifier). Upon receiving this PDCCH (which is equivalent to the base station notifying the user that it has received and allowed the user's request), the UE determines that the (serving) beam for that cell has been successfully restored.

[0113] This mechanism mainly consists of the following four parts:

[0114] Beam failure detection (BFD): The terminal measures the beam failure detection reference signal (BFD RS) at the physical layer and determines whether a beam failure event has occurred based on the measurement results. The determination condition is: if the metric of all serving beams (i.e., assuming the hypothetical PDCCH block error rate (BLER)) meets a preset condition (exceeding a preset threshold), it is identified as a beam failure instance (BFI). The UE physical layer reports an indication to the UE MAC layer. This reporting process is periodic. The UE MAC layer increments the counter (BFI_COUNTER) by 1 and starts or restarts the beam failure detection timer (beamFailureDetectionTimer). When BFI_COUNTER is greater than or equal to the maximum count value, i.e., BFI_COUNTER >= beamFailureInstanceMaxCount, the UE determines that a beam failure event has occurred.

[0115] New candidate beam identification: The terminal physical layer measures the candidate beam reference signal (RS) to find new candidate beams. This step is not mandatory after a beam failure event; it can also be performed beforehand. When the UE physical layer receives a request, instruction, or notification from the UE higher layer (MAC layer), it reports the measurement results that meet preset conditions (the measurement quality of the candidate beam RS exceeds a preset L1-RSRP threshold) to the UE higher layer. The reported content is {candidate beam RS index, L1-RSRP}. The UE higher layer selects the candidate beam based on the physical layer's report.

[0116] Beam Failure Recovery Request Transmission (BFRQ): The UE MAC layer determines the PRACH resource based on the selected candidate beam. If the UE determines that the triggering conditions for a Beam Failure Recovery Request (BFRQ) are met, the UE sends the BFRQ to the base station on a contention-free PRACH. The terminal needs to send the BFRQ according to the number of BFRQ transmissions and / or timers configured in the network. If, after sending the BFRQ, the UE does not detect a PDCCH in the search space set indicated by the higher-layer parameter recoverySearchSpaceId within a time window, and the maximum number of BFRQ transmissions has not been reached, the UE retransmits the BFRQ.

[0117] Beam failure recovery response (BFRR): After receiving the BFRQ, the base station sends a response in the PDCCH on the CORESET-BFR, which may include switching to a new candidate beam, restarting beam search, or other indications. The CRC of the downlink control information (DCI) format of this PDCCH is scrambled by C-RNTI or MCS-C-RNTI. If the BFR fails, the physical layer sends an indication to the higher layers of the UE for the higher layers to determine the subsequent radio link failure process.

[0118] It should be noted that in this embodiment, the beam can also be referred to as a spatial filter, spatial domain transmission filter, etc. Beam information can also be represented by other terms, such as Transmission Configuration Indicator (TCI) state information, Quasi co-location (QCL) information, spatial relation information, etc. The aforementioned CORESET-BFR and BFR search space are dedicated to BFR and can only be used to transmit the aforementioned C-RNTIPDCCH used for response; they are not used to transmit PDCCH for other purposes.

[0119] SCell BFR mechanism:

[0120] The BFR procedure for Scell ​​was introduced in R16.

[0121] The base station does not configure BFR RACH, BFR search space, and BFR CORESET for the Scell. When a BF occurs in the Scell, if uplink resources are available, an Scell ​​BFR MAC CE will be included in the uplink resources. This MAC CE carries the ID of the Scell ​​that experienced the BF. If no uplink resources are available when the Scell ​​experiences a BF, the user first sends an SR request to the base station to allocate uplink resources, and then sends the Scell ​​BFR MAC CE in the uplink resources allocated by the base station. This SR is sent on a dedicated SR resource configured for Scell ​​BFR.

[0122] By reading the SCell BFR MAC CE, the base station can determine which Scell ​​experienced a Beam Failure Recovery (BFR). Similar to the Pcell BFR procedure, the base station also needs to send a response to the user. This response is a C-RNTI-scrambled scheduling newtransmission ULgrant (also a PDCCH), and the Hybrid Automatic Repeat Request (HARQ) process ID carried in this PDCCH is the same as the HARQ process ID of the SCell BFR MAC CE. For example, if a user transmits an SCell BFR MAC CE on uplink resources allocated by a DCI carrying HARQ process ID = 1, and then receives another scheduling uplink DCI carrying HARQ process ID = 1 with a toggled New Data Indicator (NDI), the user can consider the Scell ​​BFR successful and cancel all triggered BFRs for this Serving Cell.

[0123] SR-BFR and RACH-BFR:

[0124] User movement, angle rotation, and congestion can cause signal degradation in data or control beams. Beam management monitors and switches beams to ensure reliable transmission and reception between the gNB and UE. However, in some cases, if a sufficient window is not provided for beam switching, signal quality can degrade rapidly, leading to beam misalignment. In such situations, control channel performance on the UL or DL ​​may be affected, potentially resulting in radio link failure and connection re-establishment. These processes cause additional latency that impacts data throughput.

[0125] To overcome these issues, a beam recovery procedure is considered when beam misalignment is detected at the UE. Through beam recovery, the gNB and UE can reconstruct data and control channels using alternative beams. Beam recovery needs to consider two scenarios: UL synchronization and UL asynchrony.

[0126] UL Synchronization:

[0127] When the gNB and UE are UL synchronized, beam restoration is performed using SR resources. The gNB monitors scheduling requests, and upon receiving a beam restoration message, the gNB and UE re-establish data and control channels.

[0128] Beam recovery using scheduling requests has the following advantages: (1) SR regions may contain more resources (cyclic shift); beam recovery using SR is faster than the competition-based RACH method.

[0129] UL not synchronized:

[0130] When the gNB and UE are out of UL synchronization, the UE sends a random access preamble for a contention-based RACH procedure. After the RACH procedure is successful, the gNB and UE re-establish data and control channels.

[0131] Radio link failure (RLF):

[0132] In 2G / 3G / 4G single-beam scenarios, beam failure can be considered equivalent to radio link failure (RLF). However, in multi-beam scenarios, a radio link failure is determined to occur between the base station and the terminal (UE) only when the wireless problem within the cell cannot be resolved through the beam recovery process, or when the terminal (UE) cannot find a suitable beam and the connection between the terminal (UE) and the base station (gNB) cannot be successfully restored through the random access process of the beam.

[0133] In related technologies, under the RRC_CONNECTED state, the reported "out-of-sync" status is used to determine whether an RLF (radio link failure) has occurred, and the reported "in-sync" status is used to determine whether the RLF has been recovered. The specific process is as follows:

[0134] 1. When N310 consecutive "out-of-sync" calls are reported, start the T310 timer.

[0135] 2. If the T310 timer reports N311 consecutive "in-sync" messages while it is running, then stop the T310 timer.

[0136] 3. When the T310 timer times out, it is considered that an RLF has occurred, and the UE side triggers the RRC connection re-establishment procedure.

[0137] In this application, N310, N311 and T310 are referred to as RLF detection configuration parameters.

[0138] Reporting of "out-of-sync" and "in-sync":

[0139] Terminal determines Q out and Q inThe threshold value is used to compare the results obtained from measuring the Radio Link Monitoring-Reference Signal (RLM-RS) with this threshold value. Measurement results worse than Q are considered acceptable. out The system reports "out-of-sync" events in a timely manner, and the measurement results are better than Q. in When this happens, an "in-sync" event is reported.

[0140] Q in It is the threshold value at which the downlink channel quality of the terminal is good enough to enable reliable transmission, which is actually converted into the BLER of PDCCH detection to achieve BLER. in Channel quality at that time, Q out This is the threshold value when the downlink channel quality of the terminal is no longer sufficient for reliable transmission, and it is actually converted into the BLER of the PDCCH detection reaching BLER. out The RLM-RS is the channel quality at that time. It can be an SSB, a Channel State Information Reference Signal (CSI-RS), or a combination of SSB and CSI-RS.

[0141] Artificial Intelligence (AI) / Machine Learning (ML):

[0142] AI has been widely applied in various fields. Integrating artificial intelligence into wireless communication networks to significantly improve technical indicators such as throughput, latency, and user capacity is an important task for future wireless communication networks. AI modules can be implemented in various ways, such as neural networks, decision trees, support vector machines, and Bayesian classifiers.

[0143] Neural networks are composed of neurons, and a diagram of a neuron is shown below. Figure 1c As shown in the diagram. Here, a1, a2, ..., aK are the inputs, w is the weight (multiplicative coefficient), b is the bias (additive coefficient), and σ(.) is the activation function. Common activation functions include Sigmoid, tanh, and the Rectified Linear Unit (ReLU), etc.

[0144] The parameters of a neural network are optimized using gradient optimization algorithms. Gradient optimization algorithms are a class of algorithms that minimize or maximize an objective function (sometimes called a loss function), which is often a mathematical combination of model parameters and data. For example, given data X and its corresponding label Y, we construct a neural network model f(.). With the model, we can obtain the predicted output f(x) based on the input x, and calculate the difference between the predicted value and the true value (f(x) - Y), which is the loss function. Our goal is to find suitable W and b that minimize the value of the above loss function. The smaller the loss value, the closer our model is to the reality.

[0145] Most common optimization algorithms are based on the error back propagation (BP) algorithm. The basic idea of ​​the BP algorithm is that the learning process consists of two parts: forward propagation of the signal and backward propagation of the error. During forward propagation, the input sample is introduced from the input layer, processed layer by layer by the hidden layers, and then propagated to the output layer. If the actual output of the output layer does not match the expected output, the process transitions to the error back propagation stage. Error back propagation involves propagating the output error back to the input layer layer by layer through the hidden layers, distributing the error to all units in each layer, thus obtaining the error signal of each unit. This error signal serves as the basis for adjusting the weights of each unit. This process of adjusting the weights through forward and backward propagation is repeated continuously. This continuous adjustment of weights is the learning and training process of the network. This process continues until the error of the network output is reduced to an acceptable level, or until the predetermined number of learning iterations is reached.

[0146] Common optimization algorithms include gradient descent, stochastic gradient descent (SGD), mini-batch gradient descent, momentum method, Nesterov (named after the inventor, specifically stochastic gradient descent with momentum), adaptive gradient descent (Adagrad), Adadelta, root mean square propagation (RMSprop), and adaptive momentum estimation (Adam).

[0147] During error backpropagation, these optimization algorithms calculate the gradient based on the error / loss obtained from the loss function with respect to the current neuron, add the learning rate, previous gradients / derivatives / partial derivatives, etc., and then pass the gradient to the previous layer.

[0148] Generally, the AI ​​algorithms and models selected vary depending on the type of problem being solved. Among related technologies, the main method for improving 5G network performance using AI is to enhance or replace existing algorithms or processing modules through neural network-based algorithms and AI models. In specific scenarios, neural network-based algorithms and AI models can achieve better performance than deterministic algorithms. Commonly used neural networks include deep neural networks, convolutional neural networks, and recurrent neural networks. Existing AI tools can be used to build, train, and validate neural networks.

[0149] Background information on fine-tuning:

[0150] In practice, due to the insufficient size of real-time datasets, directly training a neural network often fails to achieve convergence. A common approach is to pre-train the network using a large amount of offline collected data until it converges. Then, the parameters of the pre-trained neural network are fine-tuned using real-time data to adapt the network to the real-world environment. Fine-tuning can be considered a training process that uses the parameters of the pre-trained neural network as initialization. During the fine-tuning phase, parameters of some layers can be frozen; generally, layers closer to the input are frozen, while layers closer to the output are activated. This ensures that the network can still converge. The smaller the amount of data during the fine-tuning phase, the more layers should be frozen, with only a small number of layers near the output being fine-tuned.

[0151] The generalization problem of neural networks:

[0152] Generalization refers to the ability of a neural network to produce reasonable outputs on data not encountered during its training (learning) process. To address the generalization problem caused by the variable wireless transmission environment, neural network-based wireless communication systems offer two solutions. The first is to train different neural networks under different transmission conditions, obtaining multiple sets of network parameters, and then switching these parameters as the actual environment changes. The second is to train a single, common neural network based on mixed data, where the network parameters do not change with the environment. Each approach has its advantages and disadvantages: the first approach performs excellently under different transmission conditions, but requires storing multiple network parameters and switching them as needed (incurring signaling overhead and frequent switching issues); the second approach only requires storing one set of neural network parameters without switching, but cannot achieve optimal performance under every transmission condition. The construction method of the mixed dataset affects the performance of the second approach.

[0153] Label:

[0154] In machine learning and deep learning, a label typically refers to an identifier or annotation of the true category or target value of a data sample. Labels are used to represent the information that the model should learn and predict, for example:

[0155] Labels in classification tasks: In classification tasks, labels indicate which category a data sample belongs to. For example, in image classification, each image sample has a label that indicates the category of the object or scene contained in the image, such as "dog" or "cat".

[0156] Labels in object detection: In object detection tasks, labels typically include the object's location information (bounding box) and category information. Each label identifies a target object in an image, including its location and category.

[0157] Labels in regression tasks: In regression tasks, labels typically represent the continuous or real-valued objective to be predicted. For example, in a house price prediction task, the label could be the actual selling price of a house.

[0158] Labels in sequence labeling: In natural language processing, labels in sequence labeling tasks are often used for tasks such as part-of-speech tagging and named entity recognition. Labels are used to represent the attributes or categories of each word or character in a text sequence.

[0159] Labels are a crucial component in supervised learning tasks, used to train machine learning models. Models learn patterns and regularities by comparing themselves to true labels in order to make predictions or classifications on unseen data. The quality and accuracy of the labels are critical to the model's performance.

[0160] AI lifecycle management:

[0161] Please refer to Figure 1d The lifecycle management of AI / ML models includes multiple AI functional modules: model training, model deployment, model inference, model monitoring, and model updates.

[0162] Model training:

[0163] This function performs AI model training, validation, and testing, and can generate model performance metrics that can be used as part of the model testing process. If needed, it also handles data preparation based on the training data provided by the data collection function (e.g., data preprocessing and cleaning, formatting, and transformation).

[0164] Training / Update Model: If a model storage function is available, it is used to transfer trained, validated, and tested AI models to the model storage function, or to transfer updated versions of the model to the model storage function.

[0165] Model Management:

[0166] This module monitors the operation of AI models or the deployment of AI functions (e.g., model selection / deactivation / switching / rollback) and provides feedback on model monitoring performance. It is also responsible for making decisions based on data received from the data collection and inference modules to ensure correct inference operations.

[0167] Management instructions: Information input from the model management function to the model inference function. This information may include selecting / (deactivating) the AI ​​model or AI / ML-based functions to activate / switch the model, reverting to non-AI / ML operations (i.e., operations independent of the inference process), etc.

[0168] Model transfer request: Used to request a model from the model storage function.

[0169] Performance Feedback / Retraining Request: Information required for model training function input, such as for model (re)training or updating purposes.

[0170] Model inference:

[0171] The data (i.e., inference data) provided by the data collection function is used as input to provide the output of the applied AI model. If necessary, the inference function is also responsible for data preparation based on the inference data provided by the data collection function (e.g., data preprocessing and cleaning, formatting and transformation).

[0172] Inference output: Data used by management functions to monitor the performance of AI models or AI / ML functions.

[0173] It should be noted that the cell in the embodiments of this application can also be replaced by TRP, network node, carrier, bandwidth, bandwidth part (BWP), etc.

[0174] In the embodiments of this application, RACH, PRACH, or preamble can also be any module that includes at least one of the following: synchronization signal, random access signal / channel, uplink control signal / channel, wake-up signal (WUS), configured licensed uplink channel (e.g., CG PUSCH), and other control channels for access (e.g., PUCCH). Random access channel timing (RACH Occasion, RO) and physical random access channel timing (PRACH Occasion) both refer to the time-frequency resources required to transmit the RACH or preamble.

[0175] The SSB or synchronization signal appearing in the embodiments of this application can also be any module that includes at least one of the following: synchronization signal, broadcast signal, broadcast channel (PBCH), other system message downlink broadcast channel or its control channel or its search space or its control resource set.

[0176] The SSB to RO mapping described in this application embodiment can also refer to the generalized association between downlink signals and uplink signals / resources, such as the relationship between CSI-RS and RO.

[0177] The AI ​​described in this application embodiment can also be represented as machine learning (ML), which has various implementation methods, such as neural networks, decision trees, support vector machines, Bayesian classifiers, etc. This application does not make specific limitations on it.

[0178] The AI ​​model described in this application embodiment may also be referred to as an AI unit, AI structure, etc., or the AI ​​model may refer to a processing unit that can implement specific algorithms, formulas, processing flows, capabilities, etc. related to AI, or the AI ​​model may be a processing method, algorithm, function, module or unit for a specific dataset, or the AI ​​model may be a processing method, algorithm, function, module or unit running on AI-related hardware such as a graphics processing unit (GPU), a neural network processing unit (NPU), a tensor processing unit (TPU), or an application specific integrated circuit (ASIC). This application does not make any specific limitations in this regard.

[0179] The AI ​​model identifier (i.e., ID) described in this application embodiment may be an AI structure identifier, an AI algorithm identifier, or an identifier of a specific dataset associated with the AI ​​model, or an identifier of a specific scenario, environment, channel characteristic, or device related to the AI, or an identifier of a function, characteristic, capability, or module related to the AI. This application does not specifically limit this. The AI ​​model identifier can be described in various ways, such as a function ID and / or a model ID, a physical model ID, a logical model ID, a global model ID, a local model ID, etc.

[0180] The server described in this application embodiment may specifically refer to an entity used for training, predicting, or providing AI-related information, or it may refer to services that bypass operators (e.g., over-the-top, OTT).

[0181] The following description, in conjunction with the accompanying drawings, details the random access event processing method, AI model training method, apparatus, and related equipment provided in this application through some embodiments and application scenarios.

[0182] Please refer to Figure 2 , Figure 2 This is a flowchart of a random access event handling method provided in an embodiment of this application, wherein the method is executed by a first device. Figure 2 As shown, the method includes the following steps:

[0183] Step 201: The first device obtains the first output information of the first AI model, which is related to the random access event.

[0184] The first device can be a terminal, a network-side device, or a server. The first output information is the output information output by the first AI model, for example, the first AI model performs model inference based on the first input information and outputs the first output information.

[0185] In some embodiments, the first AI model may be located on the first device side, and the first device may use the first AI model to perform inference and prediction to obtain the first output information output by the first AI model.

[0186] In some embodiments, the first AI model may be located on the second device side. The second device uses the first AI model to perform inference and prediction, obtains the first output information output by the first AI model, and sends the first output information to the first device. That is, the first device obtains the first output information sent by the second device.

[0187] In this embodiment, the first output information is related to a random access event. For example, the first output information is related to the random access event predicted by the first AI model. The predicted random access event can also be understood as an event or scenario that may trigger a random access procedure in the future, or a potential event or scenario that may trigger a random access procedure, such as whether beam failure or uplink synchronization loss will occur within a future period. Alternatively, the first output information may be a random access procedure that may occur in the future or a potential random access procedure.

[0188] Step 202: The first device performs a first operation based on the first output information.

[0189] Understandably, the first output information is related to a random access event, and the first operation can be an operation related to a random access event. For example, if the first output information includes information that beam failure will not occur within a certain period of time as predicted by the first AI model inference, the first operation can be that the terminal reduces measurements on a specific beam. Alternatively, if the first output information includes information that beam failure may occur within a certain period of time as predicted by the first AI model inference, the first operation can be that the terminal switches to a new beam in advance for data transmission and reception. Another example is that if the first output information includes information related to wireless link failure predicted by the first AI model inference, the first operation can be that the terminal switches or reselects to a new cell in advance. Yet another example is that if the first output information includes information about potential beams / beam sets that the terminal will switch to within a certain period of time as predicted by the first AI model inference, the first operation can be that the terminal only measures measurement information on these beams, and so on, without further specific examples. In this embodiment, the first device can obtain the first output information of the first AI model and perform the first operation based on the first output information. The first output information of the first AI model is related to the random access event. That is, the first output information is the information related to the random access event that the first AI model infers and predicts. Thus, the first AI model can infer and predict the random access event, thereby avoiding unnecessary measurements, reducing the number of RACH events, helping to reduce the power consumption of the first device, and also helping to reduce the probability of data transmission interruption events caused by triggering random access events, thereby improving throughput and reducing data transmission latency.

[0190] In some embodiments, the first operation includes at least one of the following:

[0191] The first information is reported to the training node of the first AI model, and the first information includes at least one of the following: the input information of the first AI model, the output information, and the current state information of the first device (e.g., location information); the first information can be used for fine-tuning of the first AI model or for updating the first AI model.

[0192] Revert to the non-AI random access event triggering process;

[0193] Trigger a first action of the first AI model, the first action including at least one of switching, updating, changing input information, training, fine-tuning and supervision;

[0194] Wherein, if the first device is a terminal, the first operation further includes at least one of the following:

[0195] Measurements are performed on at least one specific beam or reference signal, which may be an SSB, a Channel State Information Reference Signal (CSI-RS), a Tracking Reference Signal (TRS), or a Sounding Reference Signal (SRS).

[0196] Stop the measurement on at least one beam or reference signal;

[0197] Continuous measurement or cessation of measurement over a period of time;

[0198] The measurement is stopped on at least one beam or reference signal for a period of time;

[0199] Increase the measurement cycle;

[0200] Switch to a new neighborhood;

[0201] Switch to a new first object for data transmission and reception, the first object including at least one of beam, band, carrier, subband, TRP, and BWP;

[0202] Add a secondary cell (SCell);

[0203] Random access is performed based on the prediction of possible consecutive Listen Before Talk (LBT) failures;

[0204] Small data transmissions are performed in an idle or inactive state based on the prediction of possible small data transmissions, such as RACH based random access for small data transmissions;

[0205] Activate or deactivate on-demand signals, such as activating on-demand SSB by sending a specific PRACH signal, or activating on-demand SSB for the corresponding RA;

[0206] Initiate a scheduling request (SR);

[0207] Initiate Beam Failure Recovery (BFR);

[0208] Initiate uplink synchronization;

[0209] Initiate the random access procedure corresponding to Radio Link Failure (RLF);

[0210] Initiate a non-RACH-based access procedure.

[0211] The on-demand signals include at least one of the following: SSB, System Information Block (SIB, e.g., SIB1), PRACH, SRS, other system information, and common PDCCH resource set.

[0212] Optionally, when the first device is a terminal and the first operation performed by the terminal is to switch to a new cell, the terminal predicts based on the first AI model that the cell signal quality is deteriorating and a cell handover is imminent. The terminal can then switch to a new cell before the cell signal quality deteriorates. This effectively ensures the signal quality of the terminal.

[0213] Optionally, when the first device is a terminal, and the first operation performed by the terminal is to switch to a new beam for data transmission and reception, the terminal infers based on the first AI model that the currently accessed beam will fail. That is, the first output information includes information that the currently accessed beam will fail. The terminal can switch to the new beam in advance for data transmission, thus effectively ensuring the terminal's transmission quality and reducing data latency. The terminal can switch to a new beam without initiating a random access procedure; it can perform beam switching based on direct instructions from the network-side device.

[0214] In this embodiment, the first device can perform the first operation based on the first output information output by the first AI model. The first operation is an operation related to random access events. Thus, the first device can predict the possible time period or non-occurrence of future random access events (such as handover, beam failure recovery, uplink synchronization failure, wireless link failure, LBT failure, small data transmission, SR failure, on-demand signal activation or deactivation, uplink data arrival or downlink data arrival, etc.) or the RACH process itself based on the first AI model's inference result (i.e., the first output information). It can also perform the first operation based on the inference result of the first AI model (i.e., the first output information), thereby avoiding unnecessary measurements, reducing the number of RACH occurrences, reducing the energy consumption of the first device, and reducing the probability of RACH collisions by predicting and planning RACH transmission resources in advance, reducing the data transmission interruption time, ensuring the data transmission rate, which is beneficial to improving throughput, reducing latency, and ensuring user experience.

[0215] Optionally, the first device performs a first operation based on the first output information, including:

[0216] If the first condition is met, the first device performs the first operation based on the first output information;

[0217] The first condition includes at least one of the following:

[0218] No first output information meeting the preset performance requirements was obtained after the first preset time period had elapsed;

[0219] The first AI model either failed to complete AI inference, or it successfully completed AI inference using the first AI model.

[0220] The first AI model was used for N AI inferences, where N is a positive integer.

[0221] For example, if a RACH transmission occurs within a time period determined by the first device based on the first AI model inference that a RACH would not occur, this can be considered a failure of the first AI model inference, or it can be understood as the first output information of the first AI model not meeting the preset performance requirements, i.e., no first output information meeting the preset performance requirements was obtained. If the number of times the first AI model inference fails exceeds a preset number, or if no first output information meeting the preset performance requirements is obtained after a first preset time period, then the first device performs the first operation based on the first output information.

[0222] For example, if the first device successfully completes AI inference using the first AI model, or if it fails to successfully complete AI inference using the first AI model, then the first device executes the first operation based on the first output information. Here, "successfully completing AI inference using the first AI model" can be understood as the first output information of the first AI model meeting preset performance requirements, such as a confidence level greater than or equal to a preset threshold; "failing to successfully complete AI inference using the first AI model" can be understood as the first output information of the first AI model not meeting preset performance requirements, such as a confidence level less than a preset threshold.

[0223] For example, if the first device performs one (N=1) AI inference using the first AI model, then the first device executes the first operation based on the first output information of the first AI model. Understandably, N can also be other values, and this application does not specifically limit this.

[0224] In this embodiment, when the aforementioned first condition is met, the first device executes the first operation based on the first output information of the first AI model. This limits the prerequisites for the first device to execute the first operation, making the operation of the first device more standardized.

[0225] Optionally, in this embodiment of the application, the first output information includes at least one of the following:

[0226] Information related to upward movement failure;

[0227] RACH related information;

[0228] SR-related information;

[0229] The time period during which BF or RLF will not occur;

[0230] The probability that BF or RLF will not occur;

[0231] BFR methods or procedures;

[0232] At least one candidate beam or a reference signal corresponding to the candidate beam for beam failure detection (BFD) or radio link failure detection (RLFD);

[0233] The pre-compensation information for the first parameter, wherein the first parameter includes at least one of frequency offset, time delay offset, phase offset, Doppler offset, and timing offset;

[0234] The first device may use at least one beam in the future (e.g., a future period of time or multiple periods of time or a future moment) and the probability of using each beam;

[0235] The first device may use at least one timing information at a future second time (e.g., a future period of time or multiple periods of time or a future moment) and the probability of using each timing information. The timing information may be a timing advance (TA) or a TA offset.

[0236] Whether the first device will enter the first state and the probability of entering the first state, the first state includes any of the following: idle state, inactive state, disconnected state, connected state and power saving state;

[0237] Information related to the auxiliary community;

[0238] Information related to LBT failure;

[0239] Information related to data transmission in idle or inactive states;

[0240] Information related to triggering on-demand signals;

[0241] Uplink or downlink data arrives.

[0242] Optionally, the uplink synchronization failure related information includes at least one of the following:

[0243] Will the upward trend falter in the near future, or at some point in the future?

[0244] The probability of an upward breakout occurring in the near future or at some point in the future;

[0245] The time period or moment during which upward movement will not be interrupted;

[0246] The probability of an upward step failure will not occur;

[0247] The probability of a time period or moment during which an upward step loss will not occur;

[0248] The potential time point or period during which an upward step loss may occur;

[0249] The probability of a potential time point or time period during which an upward step loss occurs.

[0250] In this embodiment, the time point can be an absolute time point, such as a specific hour and minute, a time slot or subframe numbered N, etc. Alternatively, the time point can also be a relative time, that is, a time point x hours after a reference time point. The reference time point can be the time when the first AI model outputs the inference result (e.g., the first output information), the time when the terminal last performed a measurement, the time when uplink synchronization failure, beam failure, or wireless link failure most recently occurred, the time when AI configuration information was last received, the time when system message updates occurred most recently, etc. The unit of the x duration can be s, ms, transmission time interval (TTI), slot, subframe, frame, or symbol. The size of x can be inferred from the AI ​​model (e.g., the first AI model).

[0251] In this embodiment of the application, the time period may include at least one of the following:

[0252] A period of time following the start of the specified time point;

[0253] A period of time after the first output information is obtained;

[0254] The duration from obtaining the first output information to the time point mentioned above;

[0255] The duration between time point A and time point B; time point A and time point B are inferred through an AI model (e.g., the first AI model), and the definitions of time point A and time point B are the same as those of the time points mentioned above.

[0256] The unit of the duration of the segment can be s, ms, TTI, slot, subframe, frame, or symbol. The duration of the segment is inferred through an AI model (e.g., the first AI model).

[0257] Optionally, the RACH-related information includes at least one of the following:

[0258] Will RACH be triggered in the future, within a certain period of time or at some point in the future?

[0259] The probability of triggering RACH in the future, within a certain period of time or at some point in the future;

[0260] The potential time point, time period, or moment that triggers RACH;

[0261] The probability of a potential time point, time period, or moment that triggers RACH;

[0262] The time period during which RACH will not occur;

[0263] The probability that RACH will not occur;

[0264] The probability of a RACH not occurring during a specific time period;

[0265] The type of RACH that was triggered;

[0266] Will Msg1 be sent?

[0267] The RACH may include a RACH-based BFR process.

[0268] In some embodiments, when the input information of the first AI model includes service-related information, if there is no service to be transmitted within a certain period of time, there is no need to trigger RACH, and the first output information may include information that RACH will not be triggered within a certain period of time.

[0269] In this embodiment of the application, the types of RACH include: CBRA RACH, CFRA RACH, 2-step RACH, 4-step RACH, RACH-less access, and handover triggered by the physical layer or MAC layer (such as LTM, L1 / L2-triggered mobility), etc.

[0270] Optionally, the SR-related information includes at least one of the following:

[0271] Will SR be triggered in the future, within a certain period of time or at some point in the future?

[0272] The probability of triggering SR in the future, within a certain period of time or at some point in the future;

[0273] The potential time point or time period that triggers SR;

[0274] The probability of triggering SR at potential time points or time periods;

[0275] The period during which SR transmission will not occur;

[0276] The probability that SR transmission will not occur;

[0277] The probability of a time period during which SR transmission will not occur;

[0278] Channel types used for SR transmission, such as PUSCH or PUCCH;

[0279] PUCCH format used for SR transmission;

[0280] The time-frequency domain configuration information that triggers SR transmission, such as the location of time-domain resources and the location of frequency-domain resources;

[0281] Transmission power-related information that triggers SR transmission;

[0282] The transmission cycle that triggers SR transmission.

[0283] It should be noted that the SRR includes the SR-based BFR process.

[0284] In some embodiments, when the input information of the first AI model includes business-related information, if there is no business to be transmitted within a certain period of time, there is no need to trigger SR. Therefore, the first output information may include information that SR will not be triggered within a certain period of time.

[0285] Optionally, the secondary cell (SCell) related information includes at least one of the following:

[0286] Will SCell be added in the near future, or at some point in the future?

[0287] The probability of adding SCell in the future, or at some point in the future;

[0288] The type of SCell to be added in the future, or at some point in the future, such as whether it belongs to the same Timing Advance Group (TAG);

[0289] The TA or TA offset of the SCell to be added in the future, or at some point in the future. For example, the TA offset may be the offset relative to the TA of the primary cell (PCell).

[0290] Optionally, the LBT failure-related information includes at least one of the following:

[0291] Will LBT failure occur in the future, or at some point in the future?

[0292] The probability of LBT failure occurring in the future, within a certain period of time or at some point in the future;

[0293] Will the preset number of consecutive LBT failures occur in the future, or at some point in the future?

[0294] The probability of a predetermined number of consecutive LBT failures occurring within a future period or multiple periods, or at some point in the future;

[0295] The time period during which LBT failure will not occur;

[0296] The probability of LBT failure occurring during the time period;

[0297] The time period during which a preset number of consecutive LBT failures will not occur;

[0298] The probability of a time period during which a preset number of consecutive LBT failures will not occur;

[0299] Potential time points or time periods when LBT failure may occur;

[0300] The probability of LBT failure occurring within a given time period;

[0301] The potential time point or time period during which a preset number of consecutive LBT failures occur;

[0302] The probability of a time period during which a preset number of consecutive LBT failures occur.

[0303] Optionally, the relevant information for data transmission (e.g., small data transmission) in the idle or inactive state includes at least one of the following:

[0304] Will data transmission occur in an idle or inactive state in the future, or at some point in the future?

[0305] The probability of data transmission in an idle or inactive state occurring within a future period or multiple periods, or at some future moment;

[0306] The time period during which data transmission will not occur in the idle or inactive state;

[0307] The probability that data transmission will not occur in the idle or inactive state;

[0308] The probability of data transmission occurring during periods of idle or inactive states;

[0309] The potential time point or time period during which data transmission occurs in an idle or inactive state;

[0310] The probability of a potential time point or time period for data transmission in an idle or inactive state;

[0311] The probability of data transmission occurring in an idle or inactive state.

[0312] Optionally, the relevant information for triggering the on-demand signal includes at least one of the following:

[0313] Whether on-demand signals will be triggered to activate or deactivate will occur in the future, within a certain period of time or at some point in the future;

[0314] The probability of triggering or deactivating on-demand signals in the future, or at some point in the future;

[0315] There will be no time period during which on-demand signals are triggered or deactivated.

[0316] There is no probability of triggering or deactivating the on-demand signal;

[0317] The probability of a time period during which the on-demand signal is triggered or deactivated will not occur;

[0318] The potential time point or time period during which an on-demand signal is triggered to activate or deactivate;

[0319] The probability of triggering activation or deactivating on-demand signals;

[0320] The probability of a potential time point or time period corresponding to the occurrence of a trigger activation or deactivation on-demand signal.

[0321] Optionally, the information related to the arrival of uplink or downlink data includes at least one of the following:

[0322] Whether upstream or downstream data will arrive in the near future, or at some point in the future;

[0323] The probability of uplink or downlink data arriving in the future, or at some point in the future;

[0324] The time period during which neither uplink nor downlink data will arrive;

[0325] The probability that neither uplink nor downlink data will arrive within a given time period;

[0326] The probability of no upstream or downstream data arrivals;

[0327] The potential time point or time period when upstream or downstream data arrives;

[0328] The probability of the potential time point or time period when upstream or downstream data arrives;

[0329] The probability of upstream or downstream data arriving.

[0330] In this embodiment of the application, the first output information obtained by the first device based on the first AI model inference is as described above. The first output information is related to the random access event, thereby enabling the first device to perform a first operation related to the random access event based on the first output information, which helps the first device reduce the number of times it initiates random access and effectively saves the energy consumption of the first device.

[0331] Optionally, the first AI model infers the first output information based on the first input information, wherein the first input information includes at least one of the following:

[0332] The measurement value of the second object, the second object including at least one of beam, reference signal, cell, frequency band, carrier, subband, TRP, BWP;

[0333] The changes in the measured values ​​of the second object;

[0334] Distance information;

[0335] Path loss information;

[0336] Configuration information of the reference signal;

[0337] Beam configuration and deployment information;

[0338] Identification information for a cell or cell group or TRP or TRP group or TAG or tracking area or radio access network notification area;

[0339] Location-related information for base stations, TRPs, network nodes, terminals, or satellite equipment;

[0340] Mobile-related information of network nodes, terminals, or satellite equipment;

[0341] Business information or load information;

[0342] Transmission power information of base stations, TRPs, network nodes, terminals, or satellite equipment;

[0343] Information regarding the orientation of the antenna panel of a base station, TRP, network node, terminal, or satellite equipment;

[0344] The terminal's battery status and / or energy consumption status;

[0345] The type of terminal, such as power class;

[0346] Network information supported by the terminal, such as the carrier information and network type information supported by the terminal;

[0347] The information obtained by the first device through sensing;

[0348] Weather information;

[0349] Information related to Non-Terrestrial Networks (NTN);

[0350] Multipath information of the channel, such as the first path or strongest path information of the terminal and different base stations / TRPs and its changes;

[0351] Channel fading information, such as frequency-selective fading;

[0352] Time information;

[0353] Frequency band information, frequency zone information, frequency point information, carrier frequency information, frequency layer information, or BWP information, such as the actual frequency band / frequency zone / frequency point / carrier frequency / frequency layer / BWP where the cell is operating;

[0354] The number or percentage of successful or failed LBTs over a period of time;

[0355] The number of times and / or the time for sending small data in the idle or inactive state;

[0356] Information related to on-demand signals;

[0357] Information related to SR;

[0358] The number of times and / or the duration of upward step loss;

[0359] The number of beam failures and / or the duration of the failures;

[0360] The number of cell handovers and / or the duration of handovers;

[0361] The number of times and / or the time when upstream data arrives;

[0362] The number of times and / or the time when downlink data arrives.

[0363] Optionally, the measured value is related to at least one of the following: signal strength information, signal quality information, interference signal strength information, and timing advance information.

[0364] For example, the signal strength information may be RSRP, RSSI, etc. The RSRP and RSSI may be layer 1 measurements or layer 3 measurements.

[0365] For example, the signal strength information may include RSRP, RSSI, and other measurements on one or a group of downlink broadcast signals, synchronization signals, and reference signals (such as SSB, CSI-RS, TRS, etc.). If the first AI model is used for inference on the UE side, these measurements are obtained by the UE side; if the first AI model is used for inference on the network side, these measurements are obtained by the UE side and reported to the network side.

[0366] Alternatively, the signal strength information may include measurements such as RSRP and RSSI corresponding to one or a set of uplink reference signals (e.g., SRS). If the first AI model is used for inference by network-side devices, these measurements are obtained by the network-side devices.

[0367] Alternatively, the signal strength information may include measured values ​​such as RSRP and RSSI on a specific beam-related SSB, CSI-RS, TRS, or SRS. The specific beam direction may be the currently accessed beam, two beams adjacent to the current beam, or multiple beams separated from the current beam by a certain angle.

[0368] Alternatively, the signal strength information may include RSRP, RSSI, and other measured values ​​on the SSB / CSI-RS / TRS / SRS of the currently accessed or camped cell (or TRP).

[0369] Alternatively, the signal strength information may include measured values ​​such as RSRP, RSRQ, and RSSI on the SSB / CSI-RS / TRS / SRS of neighboring cells (or TRPs).

[0370] Optionally, the signal quality information may include RSRQ, signal-to-noise ratio (SNR), signal-to-noise and interference ratio (SINR), latency, channel quality indicator (CQI), block error rate (BLER), etc.

[0371] For example, the signal quality information may include received RSRQ, SINR, SNR, latency, CQI, BLER, etc. on uplink signals (such as SRS) or downlink signals (such as SSB, CSI-RS, TRS) between the terminal and the current access / camped cell (or TRP).

[0372] Alternatively, the signal quality information may include received RSRQ, SINR, SNR, latency, CQI, BLER, etc., on uplink signals (such as SRS) or downlink signals (such as SSB, CSI-RS, TRS) between the terminal and neighboring cells (or TRPs).

[0373] Alternatively, the signal quality information may include received RSRQ, SINR, SNR, latency, CQI, BLER, etc., on at least one uplink signal (such as SRS) or downlink signal (such as SSB, CSI-RS, TRS) related to a specific beam direction. The specific beam direction may be the currently accessed beam, two beams adjacent to the current beam, or multiple beams separated from the current beam by a certain angle, etc.

[0374] Optionally, in this embodiment, the interference includes at least one of self-interference, mutual interference, inter-cell interference, and intra-cell interference. The interference signal strength information may include interference generated by the terminal itself, or mutual interference between the terminal and terminal equipment in the first cell, or mutual interference between the source cell and the target cell, or mutual interference between network nodes of the Pcell and the Scell, or the magnitude of interference between the terminal and network nodes of the first cell. The first cell includes at least one of the following: source cell, target cell, currently camped cell, currently accessed cell, Pcell, and Scell.

[0375] Optionally, the Timing Advance (TA) related information can be one or more absolute TA values ​​or relative TA values ​​within the current period or a certain time period, or one or more recently used absolute TA values ​​or relative TA values. For example, the TA related information can be the absolute TA value or relative TA value between the terminal and the network node of the first cell. The first cell includes at least one of the following: source cell, target cell, currently camped cell, currently accessed cell, Pcell, and Scell.

[0376] Optionally, in this embodiment of the application, the change in the measured value includes at least one of the following:

[0377] The amount of increase or decrease in the measured value over a period of time;

[0378] The percentage increase or decrease of the measured value over a period of time;

[0379] Whether the increase or decrease in the measured value over a period of time exceeds a first preset threshold;

[0380] The number of times or the duration during which the increase or decrease of the measured value exceeds a first preset threshold within a certain period of time.

[0381] It should be noted that the "time period" can refer to a unit of time, which can be s, ms, TTI, slot, subframe, frame, or symbol. Alternatively, the time period can be from time T to time T+n; for example, the increase or decrease in the measured value within the time period can be the difference between the measured value at time T+n and the measured value at time T. Or, the increase or decrease in the measured value within the time period can be the difference between the maximum and minimum values ​​of the measured value within the time period. Understandably, the increase or decrease in the measured value within the time period can characterize whether the measured value increases or decreases within that time period.

[0382] The percentage increase or decrease of the measured value over a given period of time can be the percentage change of the measured value relative to the initial value over that period of time. The initial value can be the first measured value obtained when the measurement started, or it can be the measured value corresponding to the start time of the period of time.

[0383] Optionally, in this embodiment of the application, the distance information may include the distance between the terminal and the currently accessed or camped cell (or TRP); or, the distance information may include the distance between the terminal and other terminals in the currently accessed or camped cell (or TRP).

[0384] Optionally, in this embodiment of the application, the path loss information includes the path loss between the terminal and the currently accessed or camped cell (or TRP); or the path loss information includes the path loss between the terminal and the neighboring cell (or TRP).

[0385] Optionally, the reference signal configuration information includes: the reference signal pattern, the time-frequency domain position or index of the reference signal, and the transmission power of the reference signal. The reference signal includes at least one of the following: Primary Synchronization Signal (PSS) / Secondary Synchronization Signal (SSS), CSI-RS, Demodulation Reference Signal (DMRS), SSB, TRS, and SRS.

[0386] Optionally, the beam configuration and deployment information includes at least one of the following:

[0387] The number of beams transmitted or received by network-side devices / terminals, beam direction, and the relationship between beams and corresponding reference signals;

[0388] One or a set of reference signal indices, beam indices, beam directions, or Transmission Configuration Indicator (TCI) statuses. For example, this could include the reference signal indices, beam indices, or beam directions actually transmitted between the terminal and the network nodes of the cell.

[0389] Optionally, in implementation of this application, the TRP identification information may include the identification of at least one of the following: source TRP, target TRP, currently residing TRP, currently accessing TRP, primary TRP, and secondary TRP.

[0390] The TRP group identification information may include the group identifier of at least one of the following: source TRP, target TRP, currently residing TRP, currently accessing TRP, primary TRP, and secondary TRP.

[0391] The cell identification information may include the physical layer cell identifier of the source cell / target cell / currently camped cell / currently accessed cell / Pcell / Scell.

[0392] The cell group identifier may include the group identifier of the cell group to which the source cell / target cell / currently camped cell / currently accessed cell / Pcell / Scell ​​belongs.

[0393] The TAG group identifier information may include the group identifier of the TAG containing the source cell / target cell / currently camped cell / currently accessed cell / Pcell / Scell.

[0394] The tracking area identification information may include the TA identifier of the source cell / target cell / currently camped cell / currently accessed cell / Pcell / Scell.

[0395] The Radio Access Network Notification Area (RNA) identification information may include the RNA identifier of the source cell / target cell / currently camped cell / currently accessed cell / Pcell / Scell.

[0396] Optionally, in this embodiment of the application, the location-related information of the base station or TRP may include the geographical location information of the base station or TRP, such as the geographical location of the base station of the first cell. The first cell includes the source cell / target cell / currently camped cell / currently accessed cell / Pcell / Scell.

[0397] Optionally, the location-related information of the terminal may be the terminal's specific geographic coordinates (such as GPS coordinates), or the terminal's approximate location range information (such as the street it is located on), or the terminal's location information relative to the host cell, access cell, a certain TRP, or a group of TRPs (such as being due east of the host cell). Alternatively, the terminal's location information may also include the terminal's distribution information, for example, it may include the number of terminals in different areas (such as the host cell, access cell, a certain TRP, or a group of TRPs).

[0398] It should be noted that the location-related information (e.g., location-related information of the terminal) described in the embodiments of this application may include the rate of change or the change status of the location information. For example, whether the location information changes within a period of time, whether the change of the location information within a period of time exceeds a preset threshold, the amount of change of the location information at time T+n relative to the location information at time T, the number of times or the duration of the change of the location information within a period of time that is greater than or less than a preset threshold, etc.

[0399] Optionally, in this embodiment of the application, the mobility-related information of the network node, terminal, or satellite equipment includes at least one of the following:

[0400] The direction and / or speed of movement of network nodes, terminals, or satellite equipment. For example, the direction of movement can be an absolute direction, such as 40 degrees east of south; or it can be a relative direction, such as the direction relative to a certain base station or reference node.

[0401] The change in the speed / direction of movement of a network node, terminal, or satellite device per unit time, such as an increase in speed of 3 m / s or a change in direction of movement of 5 degrees; the unit time can be s, ms, TTI, slot, subframe, frame, or symbol.

[0402] The difference between the speed / direction of movement of a network node, terminal, or satellite device over a period of time, from the initial value to the final value, such as the difference between the speed at time T+n and the speed at time T.

[0403] The percentage change in the final value of the moving speed / direction of a network node, terminal, or satellite device over a period of time relative to its initial value;

[0404] The difference between the maximum and minimum values ​​of the moving speed / direction of a network node, terminal, or satellite device over a period of time;

[0405] Whether the speed / direction of movement of a network node, terminal, or satellite device increases or decreases over a period of time;

[0406] Whether the degree to which the speed / direction of movement of a network node, terminal, or satellite device increases or decreases over a period of time exceeds a preset threshold;

[0407] The number of times or the duration during which the speed / direction of movement of a network node, terminal, or satellite device increases or decreases beyond a preset threshold within a certain period of time;

[0408] The number of times or duration during which the speed / direction of movement of a network node, terminal, or satellite device changes less than or equal to a preset threshold within a certain period of time.

[0409] Optionally, in this embodiment of the application, the service information or load information includes at least one of the following:

[0410] The service information or historical service information of the cell currently accessed / resided within a specific time period, including at least one of the following: service type, service cycle, service pattern, service packet size, service packet delay budget (PDB), and distribution of service packet size;

[0411] The current and / or specific time period's load status of the access / staying cell;

[0412] The load status of one or more SSBs / beams / TCIs / TRPs that are currently accessing / residing in the cell and / or within a specific time period;

[0413] The load status of one or more carriers or frequency points in the cell currently accessed / stayed for a specific period of time, including uplink or downlink traffic volume, number of users, maximum uplink or downlink traffic volume, maximum number of users, etc.

[0414] It should be noted that the load information includes changes in load information, and the changes in load information include at least one of the following:

[0415] The amount of increase or decrease in load information per unit time, where the unit time can be s, ms, TTI, slot, subframe, frame, or symbol;

[0416] The difference between the load information from the initial value to the final value over a period of time, for example, the difference between the load information at time T+n and the load information at time T;

[0417] The percentage change in final load information relative to initial load information over a period of time;

[0418] The difference between the maximum and minimum values ​​of load information over a period of time.

[0419] Optionally, in this embodiment of the application, the information obtained by the first device through sensing includes communication environment information, scene information, channel state information (e.g., line-of-sight (LOS) or non-line-of-sight (NLOS), obstructions, etc.), and the number of terminals. For example, whether there are obstacles and / or the number of obstacles under a specific beam sensed by the first device, or the number of terminals or devices covered by a specific beam sensed by the first device.

[0420] Optionally, in this embodiment, the NTN-related information includes the reference position or movement trajectory of the cell in the NTN scenario and its changes. For example, in a Low-Earth Orbit (LEO) scenario, the cell on Earth moves as the satellite moves. In a Geostationary Orbit (GEO) scenario, the cell on Earth is fixed as the satellite moves, and a reference position can be considered to exist. The NTN-related information may also include satellite information and its changes.

[0421] Optionally, in this embodiment, the time information can be accurate to a specific time, such as accurate to the second, like 13:25:38. Alternatively, the time information can be a time range, such as 13:00 to 14:00, morning / afternoon, day / night. Alternatively, the time information can be timing information obtained through other radio access technologies (RATs). The RAT can be, for example, Bluetooth, Wi-Fi, and 3G, 4G, 5G, or 6G.

[0422] Optionally, the number of times and / or the time for sending small data in the idle or inactive state includes the number of times small data is sent in the idle or inactive state within a certain period of time, or the number and time of the most recent small data transmission in the idle or inactive state.

[0423] Optionally, the relevant information of the on-demand signal includes at least one of the following:

[0424] Does the first device support the on-demand signal?

[0425] The number of times the on-demand signal is activated or deactivated within a certain period of time;

[0426] The time of the most recent activation or deactivation of the on-demand signal.

[0427] Optionally, the relevant information of the SR includes at least one of the following:

[0428] The number of times SR occurs within a certain period of time;

[0429] The number of RACH-based SRs that occur within a certain period of time;

[0430] The number of PUCCH-based SRs that occur within a certain period of time;

[0431] The time of the most recent SR failure.

[0432] Optionally, the number and / or time of the uplink failure may include the number of uplink failures within a certain period of time, or the time of the most recent uplink failure.

[0433] Optionally, the number and / or time of beam failures may include the number of beam failures occurring within a certain period of time, or the time of the most recent beam failure.

[0434] Optionally, the number of cell handovers and / or the time of occurrence may include the number of cell handovers occurring within a certain period of time, or the time of the most recent cell handover.

[0435] Optionally, the number of times and / or the time of the uplink data arrivals include the number of times uplink data arrives within a certain period of time, or the time of the most recent uplink data arrival.

[0436] Optionally, the number of downlink data arrivals and / or the time of arrival include the number of downlink data arrivals within a certain period of time, or the time of the most recent downlink data arrival.

[0437] Understandably, the first AI model performs model inference based on the first input information to obtain the first output information. In this embodiment, the first input information includes the above-mentioned content, which can help improve the accuracy and flexibility of the first AI model's inference, thereby obtaining more accurate first output information and helping to more accurately predict random access events.

[0438] Optionally, when the first device is a terminal, before the first device obtains the first output information of the first AI model, the method further includes:

[0439] The first device receives first configuration information sent by the network-side device, and the first configuration information is used by the first device to obtain the first output information of the first AI model.

[0440] Understandably, the terminal receives first configuration information sent by the network-side device (base station). For example, the first configuration information includes a first AI model. The terminal can then perform inference based on the first AI model to obtain first output information output by the first AI model. The first output information is related to random access events, which helps the terminal to process random access events. For example, the terminal can reduce unnecessary measurements, reduce the number of random accesses, and reduce terminal power consumption.

[0441] Optionally, the first configuration information includes at least one of the following:

[0442] The first AI model or the model identifier of the first AI model;

[0443] The application scope of the first AI model inference can be, for example, the frequency domain range to which the first AI model inference can be applied, the cell or cell group to which the first AI model inference can be applied, the range of the terminal's location, or the range of the distance between the terminal and the network-side equipment, etc.

[0444] The inference cycle of the first AI model;

[0445] The effective duration of the first AI model's inference;

[0446] The first triggering condition for the inference of the first AI model;

[0447] The configuration information of the first AI model, such as the input information of the first AI model (e.g., the first input information), or the output information of the first AI model (e.g., the first output information);

[0448] Information on whether the first AI model supports (or requires) joint inference between network-side devices and terminals;

[0449] Activate the switch for inference in the first AI model;

[0450] Deactivate the switch for inference in the first AI model.

[0451] In this embodiment of the application, the terminal receives at least one of the first configuration information sent by the network-side device, which enables it to obtain the first AI model or related information for reasoning with the first AI model, thereby helping the first AI model to perform reasoning and prediction.

[0452] Optionally, before the first device acquires the first output information of the first AI model, the method further includes:

[0453] The first device triggers the first AI model to perform inference when the first triggering condition is met.

[0454] The first triggering condition includes at least one of the following:

[0455] Periodic triggering;

[0456] Obtain the first configuration information;

[0457] The activation information for activating the inference of the first AI model is obtained, and the activation information may be a DCI activation indication.

[0458] Timeouts of timers associated with the first AI model, such as timeouts of specific timers used for inference in the first AI model;

[0459] Random access failure exceeds the second preset time;

[0460] The cell handover conditions (or handover event) are met;

[0461] Receive at least one of the first input information;

[0462] Beam failure detected;

[0463] Wireless link failure detected;

[0464] Timer timeout related to beam failure instance (BFI);

[0465] The increase or decrease in the measured value exceeds the first preset threshold;

[0466] The duration during which the increase or decrease in the measured value exceeds the first preset threshold is greater than the third preset duration;

[0467] The duration of no data transmission exceeds the fourth preset duration;

[0468] The movement of the terminal or satellite conforms to the preset conditions;

[0469] The number of times the first event occurs within a certain period of time exceeds a second preset threshold. The first event includes at least one of the following: uplink synchronization failure, downlink synchronization failure, cell handover, beam failure, radio link failure, uplink data arrival, downlink data arrival, random access failure, BFI, LBT success or failure, small data transmission in idle or inactive state, on-demand signal activation or deactivation, SR, RACH-based SR, and PUCCH-based SR.

[0470] The time when the first event most recently occurred is less than or equal to a first preset time.

[0471] It should be noted that the corresponding second preset thresholds for the different first events mentioned above can be the same or different. For example, if the number of uplink synchronization failures exceeds 3 within a certain period (i.e., the second preset threshold is 3), the first AI model will be triggered to perform inference; or, if the number of beam failures exceeds 5 within a certain period (i.e., the second preset threshold is 5), the first AI model will be triggered to perform inference. Of course, the first event can also be other events mentioned above, which will not be listed in detail here.

[0472] Furthermore, the first preset time corresponding to the different first events mentioned above can be the same or different. For example, if the time of the most recent uplink data arrival is less than ten minutes before the current time (i.e., the first preset time), the first AI model will be triggered to perform inference; or, if the time of the most recent cell handover is less than half an hour before the current time (i.e., the first preset time), the first AI model will be triggered to perform inference. Of course, the first event can also be other events mentioned above, which will not be listed in detail here.

[0473] Optionally, in the first triggering condition mentioned above, the movement of the terminal or satellite meets a preset condition, including at least one of the following:

[0474] The terminal or satellite moves at a speed greater than a first preset speed or less than a second preset speed;

[0475] The change in the moving speed of the terminal or satellite exceeds the fifth preset threshold.

[0476] The acceleration of the terminal or satellite is greater than the first preset acceleration;

[0477] The angle by which the terminal or satellite changes direction is greater than the first preset angle.

[0478] In this embodiment, the first AI model is triggered to perform inference when at least one of the first triggering conditions described above is met. For example, the first device periodically triggers the first AI model to perform model inference, or the first AI model is triggered to perform model inference after a random access failure exceeds a second preset time, and so on, without further listing. By triggering the first AI model to perform model inference through the first triggering conditions, the conditions under which the first AI model will be triggered to perform model inference are defined, thereby better realizing the management and use of the first AI model.

[0479] Optionally, the first device triggers the first AI model to perform inference when a first triggering condition is met, including:

[0480] The first device triggers the first AI model to perform inference when the first triggering condition is met and the first instruction information is received.

[0481] Understandably, even if the first device meets at least one of the aforementioned first triggering conditions, it will not immediately trigger the first AI model to perform model inference. Instead, it will only trigger the first AI model to perform model inference upon receiving the first instruction information. For example, the first instruction information is information sent by the network-side device to the first device, and this first instruction information may be information activating (or instructing) the first AI model to perform model inference. This allows for better management and use of the first AI model.

[0482] In some embodiments, the first device can immediately trigger model inference of the first AI model when at least one of the first triggering conditions is met. Alternatively, when at least one of the first triggering conditions is met, the first device can also determine whether to trigger model inference of the first AI model by considering the relevant capabilities of the first AI model, the application scope of the first AI model's inference, etc.

[0483] Optionally, in this embodiment of the application, the first device obtains the first output information of the first AI model, including:

[0484] The first device determines whether the inference performance of the first AI model meets the preset performance requirements based on the first indicator;

[0485] If the inference performance of the first AI model meets the preset performance requirements, the first device acquires the first output information of the inference output of the first AI model.

[0486] The first indicator includes at least one of the following:

[0487] The complexity of the first AI model can be defined differently for different types or capabilities of network-side devices or terminals. For example, for ordinary terminals, the complexity of the first AI model used must not exceed a specific value.

[0488] The inference latency of the first AI model, for example, the time it takes to obtain the first output information by using the first AI model for inference, shall not exceed a specific duration;

[0489] The success rate of the first AI model's reasoning, for example, the success rate of obtaining the first output information by using the first AI model for reasoning, shall not be less than a specific value, or in other words, it shall be greater than a specific value;

[0490] The reliability of the reasoning result of the first AI model, for example, the accuracy of the first output information obtained by reasoning using the first AI model is greater than the preset accuracy.

[0491] In this embodiment of the application, the inference performance of the first AI model is determined by the first indicator to determine whether it meets the preset performance requirements. Only when the inference performance of the first AI model meets the preset performance requirements will the first AI model be used to perform inference to obtain the first output information, thereby effectively ensuring the inference accuracy of the first AI model and the accuracy of the first output information.

[0492] Optionally, the entity performing inference using the first AI model may include at least one of the following: terminal, network-side device, and server-side device.

[0493] Understandably, when using the first AI model for inference on a terminal, the terminal is able to perform model inference based on the latest information, since the information it possesses (especially measurement information) is the most up-to-date.

[0494] In this embodiment, the network-side device includes at least one of a base station, a TRP (Transmission Point Resource Plane), and core network equipment (such as core network equipment specifically used for model training). The base station equipment can be the base station where the terminal is currently camped, the terminal is currently accessing, the terminal target is switched, or the terminal target is reselected. When the network-side device uses the first AI model for inference, the complexity, cost, and power consumption of the terminal can be reduced.

[0495] In this embodiment, the server-side device can be a device specifically designed for training, inference, or providing AI-related information; it can also be a third-party server; or it can be a device provided by an over-the-top (OTT) service provider, a third-party service provider, or the internet. Using a dedicated server-side device for inference with the first AI model can improve the performance of the first AI model and reduce the complexity and cost of network-side and terminal-side devices.

[0496] Optionally, when the terminal uses the first AI model for inference, the method further includes:

[0497] The terminal obtains the trained first AI model from a second device (such as a network-side device, a server-side device, or another terminal) and uses the first AI model to perform inference to obtain the first output information.

[0498] Wherein, at least a portion of the first input information used for inference in the first AI model is sent to the terminal by the network-side device, and the signal or channel for sending the at least a portion of the first input information includes at least one of the following: MAC CE, RRC message, NAS message, user plane data, DCI information, SIB, PDCCH, Physical downlink shared channel (PDSCH), MSG 2 information, MSG 4 information, and MSG B information.

[0499] Optionally, when the network-side device uses the first AI model for inference, the method further includes:

[0500] The network-side device obtains the trained first AI model from the second device (e.g., a terminal or server-side device or other network-side devices), and uses the first AI model to perform inference to obtain the first output information.

[0501] Wherein, at least a portion of the first input information used for inference in the first AI model is reported by the terminal, and the signal or channel that reports the at least portion of the first input information includes at least one of the following: MAC CE, RRC message, NAS message, user plane data, MSG 1 information, MSG 3 information, MSG A information, PUCCH, PUSCH, PRACH, SRS, and other uplink reference signals (e.g., wake-up signal (WUS)).

[0502] Optionally, when the server-side device uses the first AI model for inference, at least a portion of the first input information used for inference by the first AI model is indicated by the terminal or network-side device, for example, by an OTT message.

[0503] In this embodiment of the application, before the first device obtains the first output information of the first AI model, the method further includes:

[0504] The first device trains the first AI model based on the second input information; or...

[0505] The first device acquires the first AI model sent by the second device, wherein the first AI model is an AI model trained by the second device based on the second input information;

[0506] The second input information includes at least one of the following:

[0507] At least one of the first input information;

[0508] The labels used for training the first AI model may also be referred to as truth values ​​in some embodiments. The method for obtaining the labels or truth values ​​may be based on terminal feedback or may be obtained by measurement of network-side devices or terminals.

[0509] The algorithm or algorithm index used for training the first AI model, for example, the algorithm includes the error back propagation (BP) algorithm, stochastic gradient descent algorithm, etc.

[0510] The loss function used for training the first AI model is, in some embodiments, also referred to as the objective function;

[0511] The adjustment information used for training the first AI model, in some embodiments, is also referred to as feedback information or reward information, and includes at least one of the following: the number of times a rollback (i.e., rollback to a non-AI mode) occurs, the number of times the first AI model inferences successfully or fails, and the difference between the result of the first AI model's inference and the label.

[0512] The triggering conditions for training the first AI model.

[0513] In this embodiment, the first AI model may be trained by a first device, and the trained first AI model may be used for inference. Alternatively, the first AI model may be trained by a second device, and the trained first AI model may be sent to the first device, which then uses the trained first AI model for inference. The first device or the second device trains the first AI model based on the aforementioned second input information to ensure that the first AI model can be used for inference.

[0514] Optionally, the labels used for training the first AI model include at least one of the following:

[0515] The time of occurrence of the second event, which includes at least one of uplink synchronization failure, cell handover, beam failure, radio link failure, RACH, LBT failure, small data transmission, SR failure, on-demand signal activation or deactivation, uplink data arrival, and downlink data arrival.

[0516] The period of time during which the second event did not occur;

[0517] The type of RACH, such as CBRA RACH or CFRA RACH, or 2-step RACH or 4-step RACH, or RACH-less access method;

[0518] Timing advance or timing advance offset;

[0519] Channel types used for SR transmission, such as PUSCH or PUCCH;

[0520] PUCCH format used for SR transmission;

[0521] Methods or procedures for beam failure recovery, such as SR-based BFR or RACH-based BFR;

[0522] The cell ID of the target cell for the handover or at least one candidate cell;

[0523] The index of the target beam or at least one candidate beam for beam failure recovery;

[0524] The target beam or the reference signal associated with at least one candidate beam for beam failure recovery;

[0525] The community ID of the added auxiliary community.

[0526] In this embodiment of the application, the labels used for training the first AI model include information related to random access events. After the first AI model is trained based on the labels, the first AI model is able to infer and predict random access events, ensuring that the first AI model can output the first output information related to random access events.

[0527] Optionally, the triggering conditions for training the first AI model include at least one of the following:

[0528] Second trigger condition;

[0529] Periodic triggering;

[0530] Semi-static triggering.

[0531] For example, the training of the first AI model is triggered periodically, and the periodic configuration information includes at least one of the following:

[0532] The starting point for periodic model training could be, for example, the terminal receiving configuration information from the network-side device and then periodically training the first AI model.

[0533] The interval for periodic model training could be, for example, triggering at least one model training session for the first AI model every N time intervals.

[0534] The number of times the model is trained within a cycle or / or the duration of model training.

[0535] Alternatively, the training of the first AI model may be semi-statically triggered, wherein the semi-static triggering includes at least one of the following:

[0536] The issuance and activation of semi-static configuration information are triggered based on specific conditions (or events). The semi-static configuration information includes the starting point of model training, the period, and the duration within the period.

[0537] Semi-static configuration information is configured via RRC, and the training of the first AI model is activated / deactivated via DCI or MAC-CE.

[0538] Alternatively, the training of the first AI model may be triggered by a second triggering condition, which includes at least one of the following:

[0539] At least one of the first triggering conditions (the content of the first triggering conditions is described above and will not be repeated here);

[0540] The device training the first AI model obtains the second instruction information, for example, the terminal (i.e. the device training the first AI model) receives the second instruction information for model training sent by the network-side device through the TA MAC CE command;

[0541] If the first AI model fails to infer, for example, if the first AI model fails to infer, the training of the first AI model will be triggered, or if the number of times the first AI model fails to infer exceeds a certain number, the training of the first AI model will be triggered.

[0542] The first AI model fails inference K times consecutively, where K is an integer greater than 1;

[0543] Use the first AI model for inference, for example, trigger the training of the first AI model whenever inference is performed using the first AI model;

[0544] The timer for training the first AI model timed out for longer than the fifth preset duration;

[0545] L instances of model supervision occur, where L is a positive integer;

[0546] Obtain at least one of the second input information (the content of the second input information is described above and will not be repeated here).

[0547] It should be noted that when the first AI model is trained in accordance with the second triggering condition, the training of the first AI model can be triggered immediately when at least one of the second triggering conditions is met; or the training of the first AI model can be performed after at least one of the second triggering conditions is met. However, whether the training of the first AI model is triggered can be further determined by combining the indication information of the network-side device (such as activation information / deactivation information, etc.). For example, if at least one of the second triggering conditions is met and the network-side device indicates that the training of the model is triggered, then the training of the first AI model is triggered.

[0548] Optionally, the first AI model training is completed when the second condition is met; wherein the second condition includes at least one of the following:

[0549] The loss function satisfies a preset indicator, such as the training error of the first AI model being less than a preset threshold. The loss function includes at least one of the following: the mean square error or normalized mean square error between the predicted and true values ​​of the inference, and the mean absolute error between the predicted and true values ​​of the inference.

[0550] The first AI model has been trained a first preset number of times;

[0551] The first AI model has undergone a second preset number of iterations for fine-tuning.

[0552] At least one of the labels used for training the first AI model meets a preset requirement;

[0553] The training time of the first AI model is greater than or equal to the sixth preset time. For example, if the training time of the first AI model is greater than or equal to the sixth preset time, the training of the first AI model is considered to be complete.

[0554] The training energy consumption of the first AI model is greater than or equal to the preset energy consumption. For example, if the training energy consumption of the first AI model is greater than or equal to the preset energy consumption, then the training of the first AI model is considered to be complete.

[0555] In this embodiment of the application, a second condition is used to determine whether the training of the first AI model is complete, thereby enabling better training of the first AI model.

[0556] Optionally, the execution entity for training the first AI model may include at least one of the following: a terminal, a network-side device, or a server-side device. It should be noted that the execution entity for training the first AI model and the execution entity for inference using the first AI model may be the same device. For example, the first device may be able to train the first AI model and also use the first AI model (e.g., the trained first AI model) for inference.

[0557] The network-side equipment includes at least one of base station equipment, TRP (Transfer-Related Platform), and core network equipment (such as core network equipment specifically used for model training). The base station equipment can be the base station where the current terminal is camped, accessing, or where the terminal target is switched or reselected. The server-side equipment can be a device specifically used for training, inference, or providing AI-related information, or it can be a third-party server, or it can be a device provided by an over-the-top (OTT) service provider, a third-party service provider, or the internet.

[0558] Optionally, the training of the first AI model may be performed by partially training the model on the terminal side, network side device, or server side device, or by joint training on the terminal side, network side device, or server side device. The joint training includes at least one of the following:

[0559] The terminal reports at least part of the output information of the model training to the network-side device or the server-side device, and the network-side device or the server-side device uses the information reported by the terminal (i.e. the output information of the terminal model training) as one of the input contents of its own model training.

[0560] The network-side device sends at least a portion of the output information of the model training to the terminal or server-side device, and the terminal or server-side device uses the information sent by the network-side device (i.e., the output information of the network-side model training) as one of the input contents for its own model training.

[0561] The terminal-side, network-side, or server-side devices perform offline model training, and then perform fine-tuning in the actual network.

[0562] In some embodiments, when the training of the AI ​​model (e.g., the first AI model) occurs or at least partially occurs on the terminal-side or server-side device, the terminal-side or server-side device may have multiple pre-trained first AI models. The terminal-side or server-side device reports auxiliary information to the network-side device, and the network-side device determines and instructs which first AI model to use for AI inference.

[0563] There may be model differences among the multiple sets of first AI models. The model differences include at least one of the following: differences in training datasets, differences in label information, differences in input information, and differences in inference output.

[0564] The auxiliary information may include the identifiers of the multiple sets of first AI models and the classification identifiers of the datasets.

[0565] The network-side device may directly indicate the AI ​​model ID and / or dataset ID and / or data acquisition-related configuration information. The terminal or server-side device determines which AI model to use for model inference based on the received AI model ID and / or dataset ID and / or data acquisition-related configuration information.

[0566] In some embodiments, the training of the first AI model occurs, or at least partially occurs, on the network-side device or the server-side device. After the training of the first AI model is completed, the network-side device or the server-side device sends the trained first AI model to the terminal, and the terminal uses the model for inference.

[0567] In this embodiment of the application, when the first device performs inference on the first AI model based on the first input information, the method further includes:

[0568] The first device performs model supervision on the first AI model based on the second configuration information;

[0569] The second configuration information includes at least one of the following:

[0570] The model identifier of the first AI model;

[0571] The period of the model supervision;

[0572] The number of times the model is supervised;

[0573] The duration of the model supervision;

[0574] Information related to the window monitored by the model;

[0575] The triggering conditions for model supervision;

[0576] The labels of the model supervision;

[0577] The metrics for model supervision.

[0578] The period of model supervision can be the interval between model supervision sessions, for example, model supervision can be performed once a day.

[0579] The number of times the model is supervised refers to the number of times the first AI model needs to be supervised within a model supervision period. For example, the first AI model may be supervised 10 times within a model supervision period. Alternatively, the number of times the model is supervised refers to the total number of times the first AI model needs to be supervised after receiving the second configuration information.

[0580] The duration of model supervision refers to the duration of supervision required each time model supervision is performed within the model supervision cycle. Alternatively, the duration of model supervision refers to the effective duration of the second configuration information, such as Q model supervision cycles within the model supervision duration, or model supervision only being performed within the model supervision duration.

[0581] The information related to the model supervision window refers to the actual duration of model supervision or the number of samples (times) of model supervision within a model supervision period. Alternatively, the information related to the model supervision window refers to the start and end times of the window for which model supervision needs to be performed within a model supervision period.

[0582] The metrics for model supervision include at least one of the following: the error between the inference result and the true value of the first AI model, and communication system performance metrics. The communication system performance metrics include at least one of the following: transmission latency, throughput, and RACH / PUCCH resource collision probability.

[0583] The labels used for model supervision can be the labels used for training the first AI model, or they can be labels specifically used for model supervision.

[0584] In this embodiment of the application, the first device can also determine when and / or how to perform model supervision on the first AI model based on the second configuration information, thereby better ensuring the inference performance of the first AI model and guaranteeing the accuracy of the inference results of the first AI model.

[0585] Optionally, the triggering conditions for model supervision include at least one of the following:

[0586] The third event failure is determined by using the inference result of the first AI model. The third event includes at least one of the following: cell handover, beam failure recovery, and radio link failure recovery.

[0587] A fourth event occurs within a specific time period using the inference results of the first AI model, and the fourth event includes at least one of the following: uplink synchronization failure, downlink synchronization failure, beam failure, and wireless link failure.

[0588] The error of the inference result of the first AI model is greater than the preset error. The error may be the error between the inference result of the first AI model (e.g., the first output information) and the true value, or it may be the error between the output information and the label information under the training of the first AI model.

[0589] The first indicator used to determine whether the performance of the first AI model inference meets the preset performance requirements does not meet the first preset indicator. The specific content of the first indicator is as described above and will not be repeated here.

[0590] The metrics monitored by the model do not meet the second preset metrics;

[0591] The timer used for supervising the first AI model timed out;

[0592] The number of inference failures of the first AI model is greater than the third preset number. For example, the model supervision of the first AI model is triggered as soon as the first AI model fails inference, or the model supervision of the first AI model is triggered after the number of inference failures of the first AI model reaches the threshold value, or the model supervision of the first AI model is triggered after the first AI model fails inference a preset number of times in a row.

[0593] The first AI model is used for inference, for example, the model supervision of the first AI model is triggered whenever the first AI model is used for inference.

[0594] The output information of the first AI model is predefined output information (for example, it can be understood as specific output information), and the probability of the predefined output information appearing is greater than or equal to a first preset probability, or less than or equal to a second preset probability;

[0595] If the deviation value of the predefined output information is greater than the sixth preset threshold, for example, if the time point of beam failure is obtained based on historical information, and the deviation between the beam failure time predicted by the first AI model and the beam failure time obtained based on historical information is greater than the sixth preset threshold, then the model supervision of the first AI model is triggered.

[0596] In this embodiment of the application, when the above-mentioned model supervision triggering conditions are met, the model supervision of the first AI model is triggered, thereby enabling timely model supervision of the first AI model to improve the inference performance of the first AI model and ensure the accuracy of the inference results of the first AI model.

[0597] In this embodiment of the application, the first device has a first capability, which includes at least one of the following:

[0598] The ability to support the training of the first AI model;

[0599] Supports the ability to perform inference using the first AI model;

[0600] When the first device is a terminal or a server, it supports the ability to report at least one type of auxiliary information for training or inference of the first AI model; or when the first device is a network-side device, it supports the ability to indicate at least one type of auxiliary information for training or inference of the first AI model.

[0601] The network-side equipment can be a base station, a TRP, or a core network device. The auxiliary information refers to information used for training or inference of the first AI model; for example, the auxiliary information may include the identifier of the first AI model, the training labels of the first AI model, etc.

[0602] In this embodiment of the application, the first device has the first capability, and thus the first device can train and infer the first AI model, ensuring that the first device can use the first AI model to perform inference and prediction of random access events.

[0603] Optionally, the first capability is determined based on at least one of the following:

[0604] The first capability may differ depending on the type of the device (e.g., Redcap, IoT, etc.); similarly, the first capability may differ depending on the type of network-side device (e.g., NTN, TN).

[0605] Reference signals indicate, for example, specific resources of PRACH (such as specific ROs or preambles) that indicate the terminal has the ability to train the first AI model or to infer the first AI model;

[0606] Uplink control information, such as physical layer control information (e.g., UCI information reported by the terminal to the network);

[0607] RRC signaling;

[0608] Specific interface messages, such as those carried by specific interface messages between the terminal and the server, or between the terminal and the network-side device, or between the server and the network-side device, can be messages related to a specific AI model or messages related to all AI models.

[0609] In some embodiments, if the terminal performs first AI model training and / or inference, the type of terminal needs to be considered, as different types of terminals may have different AI model training and / or inference capabilities.

[0610] The input information used for model training and / or inference varies depending on the type of terminal. For example, for terminals with weaker capabilities, the input information used for model training should be less.

[0611] Different types of terminals use different labels for model training.

[0612] The execution method of model training varies depending on the type of terminal. For example, for terminals with weaker capabilities, model training can be performed only on the network side, or only a small part of the joint model training can be performed on the terminal side (for example, model training involving user privacy data can be performed on the terminal side).

[0613] Different types of terminals require different AI models for training. For example, terminals with weaker capabilities may not be able to apply overly complex AI models.

[0614] In this embodiment, when the terminal performs training and / or inference of the first AI model, different types of terminals can have different AI model training and / or inference capabilities, thereby enabling the training and / or inference of the first AI model to be applicable to different types of terminals and effectively improving the application of the first AI model in different terminal types.

[0615] Please refer to Figure 3 , Figure 3 This is a flowchart of an AI model training method provided in an embodiment of this application. The method is executed by a second device, which may be a terminal, a network-side device, or a server-side device. Figure 3 As shown, the method includes the following steps:

[0616] Step 301: The second device trains the first AI model based on the second input information and sends the trained first AI model to the first device.

[0617] The trained first AI model is used by the first device to obtain first output information, which is related to the random access event; the second input information includes at least one of the following:

[0618] First input information, the first AI model obtains the first output information based on the first input information;

[0619] Labels used for training the first AI model;

[0620] Algorithm or algorithm index used for training the first AI model;

[0621] The loss function used for training the first AI model;

[0622] Adjustment information used for training the first AI model;

[0623] The triggering conditions for training the first AI model.

[0624] Optionally, the first input information includes at least one of the following:

[0625] The measurement value of the second object, the second object including at least one of beam, reference signal, cell, frequency band, carrier, subband, transmit / receive point (TRP), and bandwidth portion (BWP);

[0626] The changes in the measured values ​​of the second object;

[0627] Distance information;

[0628] Path loss information;

[0629] Configuration information of the reference signal;

[0630] Beam configuration and deployment information;

[0631] Identification information for a cell or cell group or TRP or TRP group or timed advance group TAG or tracking area or wireless access network notification area;

[0632] Location-related information for base stations, TRPs, network nodes, terminals, or satellite equipment;

[0633] Mobile-related information of network nodes, terminals, or satellite equipment;

[0634] Business information or load information;

[0635] Transmission power information of base stations, TRPs, network nodes, terminals, or satellite equipment;

[0636] Information regarding the orientation of the antenna panel of a base station, TRP, network node, terminal, or satellite equipment;

[0637] The terminal's battery status;

[0638] Network information supported by the terminal;

[0639] The information obtained by the first device through sensing;

[0640] Weather information;

[0641] NTN related information;

[0642] Multipath information of the channel;

[0643] Channel fading information;

[0644] Time information;

[0645] Frequency band information, frequency zone information, frequency point information, carrier frequency information, frequency band layer information, or bandwidth portion BWP information;

[0646] The number or percentage of successful or failed LBTs over a period of time;

[0647] The number of times and / or the time for sending small data in the idle or inactive state;

[0648] Information related to on-demand signals;

[0649] Information related to SR;

[0650] The number of times and / or the duration of upward step loss;

[0651] The number of beam failures and / or the duration of the failures;

[0652] The number of cell handovers and / or the duration of handovers;

[0653] The number of times and / or the time when upstream data arrives;

[0654] The number of times and / or the time when downlink data arrives.

[0655] Optionally, the measured value is related to at least one of the following: signal strength information, signal quality information, interference signal strength information, and timing advance (TA) related information.

[0656] Optionally, the changes in the measured values ​​include at least one of the following:

[0657] The amount of increase or decrease in the measured value over a period of time;

[0658] The percentage increase or decrease of the measured value over a period of time;

[0659] Whether the increase or decrease in the measured value over a period of time exceeds a first preset threshold;

[0660] The number of times or the duration during which the increase or decrease of the measured value exceeds a first preset threshold within a certain period of time.

[0661] Optionally, the relevant information of the on-demand signal includes at least one of the following:

[0662] Does the first device support the on-demand signal?

[0663] The number of times the on-demand signal is activated or deactivated within a certain period of time;

[0664] The time of the most recent activation or deactivation of the on-demand signal.

[0665] Optionally, the relevant information of the SR includes at least one of the following:

[0666] The number of times SR occurs within a certain period of time;

[0667] The number of SRs based on the random access channel RACH that occur within a certain period of time;

[0668] The number of SRs based on the Physical Uplink Control Channel (PUCCH) that occur within a certain period of time;

[0669] The time of the most recent SR failure.

[0670] Optionally, the labels used for training the first AI model include at least one of the following:

[0671] The time of occurrence of the second event, which includes at least one of uplink synchronization failure, cell handover, beam failure, radio link failure, RACH, LBT failure, small data transmission, SR failure, on-demand signal activation or deactivation, uplink data arrival, and downlink data arrival.

[0672] The period of time during which the second event did not occur;

[0673] Types of RACH;

[0674] Timing advance or timing advance offset;

[0675] Channel type used for SR transmission;

[0676] PUCCH format used for SR transmission;

[0677] Methods or procedures for beam failure recovery;

[0678] The cell ID of the target cell for the handover or at least one candidate cell;

[0679] The index of the target beam or at least one candidate beam for beam failure recovery;

[0680] The target beam or the reference signal associated with at least one candidate beam for beam failure recovery;

[0681] The community ID of the added auxiliary community.

[0682] Optionally, the triggering conditions for training the first AI model include at least one of the following:

[0683] Second trigger condition;

[0684] Periodic triggering;

[0685] Semi-static triggering.

[0686] Optionally, the second triggering condition includes at least one of the following:

[0687] The second device obtains the second instruction information;

[0688] The first AI model failed to infer.

[0689] The first AI model fails inference K times consecutively, where K is an integer greater than 1;

[0690] Inference is performed using the first AI model;

[0691] The timer for training the first AI model timed out for longer than the fifth preset duration;

[0692] L instances of model supervision occur, where L is a positive integer;

[0693] Obtain at least one of the second input information.

[0694] Optionally, the first AI model training is completed when the second condition is met; wherein the second condition includes at least one of the following:

[0695] The loss function satisfies a preset index;

[0696] The first AI model has been trained a first preset number of times;

[0697] The first AI model has undergone a second preset number of iterations for fine-tuning.

[0698] At least one of the labels used for training the first AI model meets a preset requirement;

[0699] The training time of the first AI model is greater than or equal to the sixth preset time.

[0700] The training energy consumption of the first AI model is greater than or equal to the preset energy consumption.

[0701] It should be noted that the relevant probabilities and specific interpretations involved in the embodiments of this application can be referred to the description in the above-mentioned first device-side method embodiments. To avoid repetition, they will not be repeated in this embodiment.

[0702] In this embodiment of the application, the second device can train the first AI model based on the second input information and send the trained first AI model to the first device, thereby ensuring the first device's use of the first AI model and ensuring that the first device can infer and predict potential random access events based on the first AI model.

[0703] The random access event processing method provided in this application can be executed by a random access event processing device. This application uses the execution of the random access event processing method by a random access event processing device as an example to illustrate the random access event processing device provided in this application.

[0704] This application provides a random access event processing device. As an example, the random access event processing device may be a communication device or a component within a communication device, such as a chip. The communication device may be a terminal, a network-side device, or a server, etc. Exemplarily, the terminal may include, but is not limited to, the type of terminal 11 listed above, and the network-side device may include, but is not limited to, the type of network-side device 12 listed above. This application does not impose specific limitations.

[0705] The random access event processing device includes a receiving module, a transmitting module, and a processing module. These modules can be implemented in software or hardware. When implemented in hardware, the processing module can be implemented by a processor. For example, the processor can include general-purpose processors, special-purpose processors, such as a Central Processing Unit (CPU), microprocessor, Digital Signal Processor (DSP), Artificial Intelligence (AI) processor, Graphics Processing Unit (GPU), Application Specific Integrated Circuit (ASIC), Network Processor (NP), Field Programmable Gate Array (FPGA), or other programmable logic devices, gate circuits, transistors, discrete hardware components, etc. The receiving and transmitting modules can be implemented by a communication interface, which can include one or more of the following: transceiver, pins, circuits, bus, radio frequency unit, etc.

[0706] For details, see Figure 4 , Figure 4 This is a structural diagram of a random access event processing device provided in an embodiment of this application, wherein the device is applied to a first device. Figure 4 As shown, the random access event processing device 400 includes:

[0707] The acquisition module 401 is used to acquire the first output information of the first AI model, wherein the first output information is related to the random access event;

[0708] The execution module 402 is used to perform a first operation based on the first output information.

[0709] Optionally, the first operation includes at least one of the following:

[0710] The first information is reported to the training node of the first AI model, and the first information includes at least one of the following: the input information of the first AI model, the output information, and the current status information of the first device;

[0711] Revert to the non-AI random access event triggering process;

[0712] Trigger a first action of the first AI model, the first action including at least one of switching, updating, changing input information, training, fine-tuning and supervision;

[0713] Wherein, if the first device is a terminal, the first operation further includes at least one of the following:

[0714] Measurements are performed on at least one specific beam or reference signal;

[0715] Stop the measurement on at least one beam or reference signal;

[0716] Continuous measurement or cessation of measurement over a period of time;

[0717] The measurement is stopped on at least one beam or reference signal for a period of time;

[0718] Increase the measurement cycle;

[0719] Switch to a new neighborhood;

[0720] Switch to a new first object for data transmission and reception, the first object including at least one of beam, frequency band, carrier, subband, transmit / receive point (TRP), and bandwidth portion (BWP);

[0721] Add auxiliary cells;

[0722] Random access is performed based on the prediction of possible consecutive LBT (Listen Before You Talk) failures.

[0723] Small data transmissions are performed in the idle or inactive state based on the prediction of possible small data transmissions.

[0724] Activate or deactivate on-demand signals;

[0725] Initiate a scheduling request (SR);

[0726] Beam initiation failure recovery BFR;

[0727] Initiate uplink synchronization;

[0728] Initiate the random access procedure corresponding to the radio link failure RLF;

[0729] Initiate an access procedure based on the non-random access channel RACH.

[0730] Optionally, the execution module 402 is further configured to:

[0731] If the first condition is met, the first operation is performed based on the first output information;

[0732] The first condition includes at least one of the following:

[0733] No output information meeting the preset performance requirements was obtained after the first preset time period had elapsed;

[0734] The first AI model either failed to complete AI inference, or it successfully completed AI inference using the first AI model.

[0735] The first AI model was used for N AI inferences, where N is a positive integer.

[0736] Optionally, the first output information includes at least one of the following:

[0737] Information related to upward movement failure;

[0738] RACH related information;

[0739] SR-related information;

[0740] The time period during which beam failure (BF) or beam refraction (RLF) will not occur;

[0741] The probability that BF or RLF will not occur;

[0742] BFR methods or procedures;

[0743] At least one candidate beam or a reference signal corresponding to the candidate beam for BFD or RLFD;

[0744] The pre-compensation information for the first parameter, wherein the first parameter includes at least one of frequency offset, time delay offset, phase offset, Doppler offset, and timing offset;

[0745] The first device may use at least one beam in the future and the probability of using each beam;

[0746] The first device may use at least one timing information in a future second time and the probability of using each timing information;

[0747] Whether the first device will enter the first state and the probability of entering the first state, the first state includes any of the following: idle state, inactive state, disconnected state, connected state and power saving state;

[0748] Information related to the auxiliary community;

[0749] Information related to LBT failure;

[0750] Information related to data transmission in idle or inactive states;

[0751] Information related to triggering on-demand signals;

[0752] Uplink or downlink data arrives.

[0753] Optionally, the first AI model infers the first output information based on the first input information, wherein the first input information includes at least one of the following:

[0754] The measurement value of the second object, the second object including at least one of beam, reference signal, cell, frequency band, carrier, subband, transmit / receive point (TRP), and bandwidth portion (BWP);

[0755] The changes in the measured values ​​of the second object;

[0756] Distance information;

[0757] Path loss information;

[0758] Configuration information of the reference signal;

[0759] Beam configuration and deployment information;

[0760] Identification information for a cell or cell group or TRP or TRP group or timed advance group TAG or tracking area or wireless access network notification area;

[0761] Location-related information for base stations, TRPs, network nodes, terminals, or satellite equipment;

[0762] Mobile-related information of network nodes, terminals, or satellite equipment;

[0763] Business information or load information;

[0764] Transmission power information of base stations, TRPs, network nodes, terminals, or satellite equipment;

[0765] Information regarding the orientation of the antenna panel of a base station, TRP, network node, terminal, or satellite equipment;

[0766] The terminal's battery status;

[0767] Network information supported by the terminal;

[0768] The information obtained by the first device through sensing;

[0769] Weather information;

[0770] Non-terrestrial network (NTN) related information;

[0771] Multipath information of the channel;

[0772] Channel fading information;

[0773] Time information;

[0774] Frequency band information, frequency zone information, frequency point information, carrier frequency information, frequency band layer information, or bandwidth portion BWP information;

[0775] The number or percentage of successful or failed LBTs over a period of time;

[0776] The number of times and / or the time for sending small data in the idle or inactive state;

[0777] Information related to on-demand signals;

[0778] Information related to SR;

[0779] The number of times and / or the duration of upward step loss;

[0780] The number of beam failures and / or the duration of the failures;

[0781] The number of cell handovers and / or the duration of handovers;

[0782] The number of times and / or the time when upstream data arrives;

[0783] The number of times and / or the time when downlink data arrives.

[0784] Optionally, the measured value is related to at least one of the following: signal strength information, signal quality information, interference signal strength information, and TA-related information.

[0785] Optionally, the changes in the measured values ​​include at least one of the following:

[0786] The amount of increase or decrease in the measured value over a period of time;

[0787] The percentage increase or decrease of the measured value over a period of time;

[0788] Whether the increase or decrease in the measured value over a period of time exceeds a first preset threshold;

[0789] The number of times or the duration during which the increase or decrease of the measured value exceeds a first preset threshold within a certain period of time.

[0790] Optionally, the relevant information of the on-demand signal includes at least one of the following:

[0791] Does the first device support the on-demand signal?

[0792] The number of times the on-demand signal is activated or deactivated within a certain period of time;

[0793] The time of the most recent activation or deactivation of the on-demand signal.

[0794] Optionally, the relevant information of the SR includes at least one of the following:

[0795] The number of times SR occurs within a certain period of time;

[0796] The number of SRs based on the random access channel RACH that occur within a certain period of time;

[0797] The number of SRs based on the Physical Uplink Control Channel (PUCCH) that occur within a certain period of time;

[0798] The time of the most recent SR failure.

[0799] Optionally, the device further includes:

[0800] The triggering module is used to trigger the first AI model to perform inference when the first triggering condition is met.

[0801] The first triggering condition includes at least one of the following:

[0802] Periodic triggering;

[0803] Obtain the first configuration information;

[0804] Obtain the activation information that activates the inference of the first AI model;

[0805] The timer associated with the first AI model timed out;

[0806] Random access failure exceeds the second preset time;

[0807] The conditions for cell handover are met;

[0808] Receive at least one of the first input information;

[0809] Beam failure detected;

[0810] Wireless link failure detected;

[0811] BFI-related timers timed out;

[0812] The increase or decrease in the measured value exceeds the first preset threshold;

[0813] The duration during which the increase or decrease in the measured value exceeds the first preset threshold is greater than the third preset duration;

[0814] The duration of no data transmission exceeds the fourth preset duration;

[0815] The movement of the terminal or satellite conforms to the preset conditions;

[0816] The number of times the first event occurs within a certain period of time exceeds a second preset threshold. The first event includes at least one of the following: uplink synchronization failure, downlink synchronization failure, cell handover, beam failure, radio link failure, uplink data arrival, downlink data arrival, random access failure, beam failure instance (BFI), LBT success or failure, small data transmission in idle or inactive state, on-demand signal activation or deactivation, SR, RACH-based SR, and PUCCH-based SR. The second preset threshold may be the same or different for different first events.

[0817] The time of the most recent occurrence of the first event is less than or equal to a first preset time, and the first preset time may be the same or different for different first events.

[0818] Optionally, the movement of the terminal or satellite conforms to preset conditions, including at least one of the following:

[0819] The terminal or satellite moves at a speed greater than a first preset speed or less than a second preset speed;

[0820] The change in the moving speed of the terminal or satellite exceeds the fifth preset threshold.

[0821] The acceleration of the terminal or satellite is greater than the first preset acceleration;

[0822] The angle by which the terminal or satellite changes direction is greater than the first preset angle.

[0823] Optionally, the triggering module is specifically used for:

[0824] The first AI model is triggered to perform inference when the first triggering condition is met and the first instruction information is received.

[0825] Optionally, the first device is a terminal, and the apparatus further includes:

[0826] The receiving module is used to receive first configuration information sent by the network-side device. The first configuration information is used by the first device to obtain the first output information of the first AI model.

[0827] Optionally, the first configuration information includes at least one of the following:

[0828] The first AI model or the model identifier of the first AI model;

[0829] The application scope of the first AI model's reasoning;

[0830] The inference cycle of the first AI model;

[0831] The effective duration of the first AI model's inference;

[0832] The first triggering condition for the inference of the first AI model;

[0833] Configuration information of the first AI model;

[0834] Information regarding whether the first AI model supports joint inference between network-side devices and terminals;

[0835] Activate the switch for inference in the first AI model;

[0836] Deactivate the switch for inference in the first AI model.

[0837] Optionally, the acquisition module 401 is specifically used for:

[0838] The inference performance of the first AI model is determined based on the first indicator to determine whether it meets the preset performance requirements.

[0839] If the inference performance of the first AI model meets the preset performance requirements, obtain the first output information of the inference output of the first AI model;

[0840] The first indicator includes at least one of the following:

[0841] The complexity of the first AI model;

[0842] The inference latency of the first AI model;

[0843] The success rate of the first AI model's inference;

[0844] The reliability of the inference results of the first AI model.

[0845] Optionally, the device further includes:

[0846] The training module is used to train the first AI model based on the second input information; or,

[0847] The device further includes an acquisition module for acquiring the first AI model sent by the second device, wherein the first AI model is an AI model trained by the second device based on the second input information.

[0848] The second input information includes at least one of the following:

[0849] At least one of the first input information;

[0850] Labels used for training the first AI model;

[0851] Algorithm or algorithm index used for training the first AI model;

[0852] The loss function used for training the first AI model;

[0853] Adjustment information used for training the first AI model;

[0854] The triggering conditions for training the first AI model.

[0855] Optionally, the labels used for training the first AI model include at least one of the following:

[0856] The time of occurrence of the second event, which includes at least one of uplink synchronization failure, cell handover, beam failure, radio link failure, RACH, LBT failure, small data transmission, SR failure, on-demand signal activation or deactivation, uplink data arrival, and downlink data arrival.

[0857] The period of time during which the second event did not occur;

[0858] Types of RACH;

[0859] Timing advance or timing advance offset;

[0860] Channel type used for SR transmission;

[0861] PUCCH format used for SR transmission;

[0862] Methods or procedures for beam failure recovery;

[0863] The cell ID of the target cell for the handover or at least one candidate cell;

[0864] The index of the target beam or at least one candidate beam for beam failure recovery;

[0865] The target beam or the reference signal associated with at least one candidate beam for beam failure recovery;

[0866] The community ID of the added auxiliary community.

[0867] Optionally, the triggering conditions for training the first AI model include at least one of the following:

[0868] Second trigger condition;

[0869] Periodic triggering;

[0870] Semi-static triggering.

[0871] Optionally, the second triggering condition includes at least one of the following:

[0872] At least one of the first triggering conditions;

[0873] The device that trained the first AI model acquires the second instruction information;

[0874] The first AI model failed to infer.

[0875] The first AI model fails inference K times consecutively, where K is an integer greater than 1;

[0876] Inference is performed using the first AI model;

[0877] The timer for training the first AI model timed out for longer than the fifth preset duration;

[0878] L instances of model supervision occur, where L is a positive integer;

[0879] Obtain at least one of the second input information.

[0880] Optionally, the first AI model training is completed when the second condition is met; wherein the second condition includes at least one of the following:

[0881] The loss function satisfies a preset index;

[0882] The first AI model has been trained a first preset number of times;

[0883] The first AI model has undergone a second preset number of iterations for fine-tuning.

[0884] At least one of the labels used for training the first AI model meets a preset requirement;

[0885] The training time of the first AI model is greater than or equal to the sixth preset time.

[0886] The training energy consumption of the first AI model is greater than or equal to the preset energy consumption.

[0887] Optionally, the device further includes:

[0888] The supervision module is used to supervise the first AI model based on the second configuration information.

[0889] The second configuration information includes at least one of the following:

[0890] The model identifier of the first AI model;

[0891] The period of the model supervision;

[0892] The number of times the model is supervised;

[0893] The duration of the model supervision;

[0894] Information related to the window monitored by the model;

[0895] The triggering conditions for model supervision;

[0896] The labels of the model supervision;

[0897] The metrics for model supervision.

[0898] Optionally, the triggering conditions for model supervision include at least one of the following:

[0899] The third event failure is determined by using the inference result of the first AI model. The third event includes at least one of the following: cell handover, beam failure recovery, and radio link failure recovery.

[0900] A fourth event occurs within a specific time period using the inference results of the first AI model, and the fourth event includes at least one of the following: uplink synchronization failure, downlink synchronization failure, beam failure, and wireless link failure.

[0901] The error in the inference result of the first AI model is greater than the preset error;

[0902] The first indicator used to determine whether the performance of the first AI model inference meets the preset performance requirements does not meet the first preset indicator;

[0903] The metrics monitored by the model do not meet the second preset metrics;

[0904] The timer used for supervising the first AI model timed out;

[0905] The first AI model failed inference more times than the third preset number of times;

[0906] Inference is performed using the first AI model;

[0907] The output information of the first AI model is predefined output information, and the probability of the predefined output information appearing is greater than or equal to a first preset probability, or less than or equal to a second preset probability;

[0908] The deviation value of the predefined output information is greater than the sixth preset threshold.

[0909] Optionally, the first device has a first capability, which includes at least one of the following:

[0910] The ability to support the training of the first AI model;

[0911] Supports the ability to perform inference using the first AI model;

[0912] When the first device is a terminal or a server, it supports the ability to report at least one type of auxiliary information for training or inference of the first AI model; or when the first device is a network-side device, it supports the ability to indicate at least one type of auxiliary information for training or inference of the first AI model.

[0913] Optionally, the first capability is determined based on at least one of the following:

[0914] The first device type;

[0915] Reference signal indication;

[0916] Uplink control information;

[0917] RRC signaling;

[0918] Specific interface messages.

[0919] The random access event processing device 400 provided in this application embodiment can achieve... Figure 2 The various processes implemented in the method embodiment achieve the same technical effect, and will not be described again here to avoid repetition.

[0920] The AI ​​model training method provided in this application can be executed by an AI model training device. This application uses an AI model training device executing the AI ​​model training method as an example to illustrate the AI ​​model training device provided in this application.

[0921] This application provides an AI model training device. As an example, the AI ​​model training device can be a communication device or a component within a communication device, such as a chip. The communication device can be a terminal, a network-side device, or a server, etc. Exemplarily, the terminal can be, but is not limited to, the type of terminal 11 listed above, and the network-side device can be, but is not limited to, the type of network-side device 12 listed above. This application does not impose specific limitations.

[0922] The AI ​​model training device includes a receiving module, a transmitting module, and a processing module. These modules can be implemented in software or hardware. When implemented in hardware, the processing module can be implemented by a processor. For example, the processor can include general-purpose processors, special-purpose processors, such as a Central Processing Unit (CPU), microprocessor, Digital Signal Processor (DSP), Artificial Intelligence (AI) processor, Graphics Processing Unit (GPU), Application Specific Integrated Circuit (ASIC), Network Processor (NP), Field Programmable Gate Array (FPGA), or other programmable logic devices, gate circuits, transistors, discrete hardware components, etc. The receiving and transmitting modules can be implemented by a communication interface, which can include one or more of the following: transceiver, pins, circuits, bus, radio frequency unit, etc.

[0923] See Figure 5 , Figure 5 This is a structural diagram of an AI model training device provided in an embodiment of this application, wherein the device is applied to a second device. Figure 5 As shown, the AI ​​model training device 500 includes:

[0924] Training module 501 is used to train a first AI model based on second input information and send the trained first AI model to a first device. The trained first AI model is used by the first device to obtain first output information, which is related to a random access event.

[0925] The second input information includes at least one of the following:

[0926] First input information, the first AI model obtains the first output information based on the first input information;

[0927] Labels used for training the first AI model;

[0928] Algorithm or algorithm index used for training the first AI model;

[0929] The loss function used for training the first AI model;

[0930] Adjustment information used for training the first AI model;

[0931] The triggering conditions for training the first AI model.

[0932] Optionally, the labels used for training the first AI model include at least one of the following:

[0933] The time of occurrence of the second event, which includes at least one of uplink synchronization failure, cell handover, beam failure, radio link failure, RACH, LBT failure, small data transmission, SR failure, on-demand signal activation or deactivation, uplink data arrival, and downlink data arrival.

[0934] The period of time during which the second event did not occur;

[0935] Types of RACH;

[0936] Timing advance or timing advance offset;

[0937] Channel type used for SR transmission;

[0938] PUCCH format used for SR transmission;

[0939] Methods or procedures for beam failure recovery;

[0940] The cell ID of the target cell for the handover or at least one candidate cell;

[0941] The index of the target beam or at least one candidate beam for beam failure recovery;

[0942] The target beam or the reference signal associated with at least one candidate beam for beam failure recovery;

[0943] The community ID of the added auxiliary community.

[0944] Optionally, the triggering conditions for training the first AI model include at least one of the following:

[0945] Second trigger condition;

[0946] Periodic triggering;

[0947] Semi-static triggering.

[0948] Optionally, the second triggering condition includes at least one of the following:

[0949] The second device obtains the second instruction information;

[0950] The first AI model failed to infer.

[0951] The first AI model fails inference K times consecutively, where K is an integer greater than 1;

[0952] Inference is performed using the first AI model;

[0953] The timer for training the first AI model timed out for longer than the fifth preset duration;

[0954] L instances of model supervision occur, where L is a positive integer;

[0955] Obtain at least one of the second input information.

[0956] Optionally, the first AI model training is completed when the second condition is met; wherein the second condition includes at least one of the following:

[0957] The loss function satisfies a preset index;

[0958] The first AI model has been trained a first preset number of times;

[0959] The first AI model has undergone a second preset number of iterations for fine-tuning.

[0960] At least one of the labels used for training the first AI model meets a preset requirement;

[0961] The training time of the first AI model is greater than or equal to the sixth preset time.

[0962] The training energy consumption of the first AI model is greater than or equal to the preset energy consumption.

[0963] The random access event processing device 500 provided in this application embodiment can achieve... Figure 3 The various processes implemented in the method embodiment achieve the same technical effect, and will not be described again here to avoid repetition.

[0964] like Figure 6 As shown in the illustration, this application also provides a communication device 600, including a processor 601 and a memory 602. The memory 602 stores programs or instructions that can run on the processor 601. For example, when the communication device 600 is a first device, the program or instructions executed by the processor 601 implement the various steps of the above-described random access event handling method embodiment and achieve the same technical effect. When the communication device 600 is a second device, the program or instructions executed by the processor 601 implement the various steps of the above-described AI model training method embodiment and achieve the same technical effect. To avoid repetition, further details are omitted here.

[0965] This application embodiment also provides a terminal, including a processor and a communication interface, wherein the communication interface is coupled to the processor, and the processor is used to run programs or instructions to implement, for example... Figure 2 The steps in the method embodiments shown, or their implementations, are as follows: Figure 3The steps in the method embodiment shown are illustrated. All implementation processes and methods of the above method embodiments can be applied to this terminal embodiment and achieve the same technical effect. The terminal can be... Figure 4 The random access event handling device shown, or Figure 5 The AI ​​model training device shown. Specifically, Figure 7 A schematic diagram of the hardware structure of a terminal to implement an embodiment of this application.

[0966] The terminal 700 includes, but is not limited to, at least some of the following components: radio frequency unit 701, network module 702, audio output unit 703, input unit 704, sensor 705, display unit 706, user input unit 707, interface unit 708, memory 709, and processor 710.

[0967] Those skilled in the art will understand that the terminal 700 may also include a power supply (such as a battery) for supplying power to various components. The power supply may be logically connected to the processor 710 through a power management system, thereby enabling functions such as managing charging, discharging, and power consumption through the power management system. Figure 7 The terminal structure shown does not constitute a limitation on the terminal. The terminal may include more or fewer components than shown, or combine certain components, or have different component arrangements, which will not be elaborated here.

[0968] It should be understood that, in this embodiment, the input unit 704 may include a graphics processor 7041 and a microphone 7042. The graphics processor 7041 processes image data of still images or videos obtained by an image capture device (such as a camera) in video capture mode or image capture mode. The display unit 706 may include a display panel 7061, which may be configured in the form of a liquid crystal display, an organic light-emitting diode, or the like. The user input unit 707 includes at least one of a touch panel 7071 and other input devices 7072. The touch panel 7071 is also called a touch screen. The touch panel 7071 may include a touch detection device and a touch controller. Other input devices 7072 may include, but are not limited to, physical keyboards, function keys (such as volume control buttons, power buttons, etc.), trackballs, mice, and joysticks, which will not be described in detail here.

[0969] In this embodiment, after receiving downlink data from the network-side device, the radio frequency unit 701 can transmit it to the processor 710 for processing; in addition, the radio frequency unit 701 can send uplink data to the network-side device. Typically, the radio frequency unit 701 includes, but is not limited to, antennas, amplifiers, transceivers, couplers, low-noise amplifiers, duplexers, etc.

[0970] The memory 709 can be used to store software programs or instructions, as well as various data. The memory 709 may primarily include a first storage area for storing programs or instructions and a second storage area for storing data. The first storage area may store the operating system, application programs or instructions required for at least one function (such as sound playback, image playback, etc.). Furthermore, the memory 709 may include volatile memory or non-volatile memory. The non-volatile memory may be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. Volatile memory can be random access memory (RAM), static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDRSDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM), and direct memory bus RAM (DRRAM). The memory 709 in the embodiments of this application includes, but is not limited to, these and any other suitable types of memory.

[0971] Processor 710 may include one or more processing units; optionally, processor 710 integrates an application processor and a modem processor, wherein the application processor mainly handles operations involving the operating system, user interface, and applications, and the modem processor mainly handles wireless communication signals, such as a baseband processor. It is understood that the aforementioned modem processor may also not be integrated into processor 710.

[0972] In one embodiment, the processor 710 is configured to:

[0973] Obtain the first output information of the first AI model, which is related to the random access event;

[0974] The first operation is performed based on the first output information.

[0975] In this implementation, terminal 700 can achieve the above. Figure 2The entire process of the method embodiment described herein, and the same or corresponding technical effects achieved, will not be repeated here to avoid duplication.

[0976] In another embodiment, the processor 710 is configured to: train a first AI model based on second input information, and send the trained first AI model to a first device, wherein the trained first AI model is used by the first device to obtain first output information, and the first output information is related to a random access event;

[0977] The second input information includes at least one of the following:

[0978] First input information, the first AI model obtains the first output information based on the first input information;

[0979] Labels used for training the first AI model;

[0980] Algorithm or algorithm index used for training the first AI model;

[0981] The loss function used for training the first AI model;

[0982] Adjustment information used for training the first AI model;

[0983] The triggering conditions for training the first AI model.

[0984] In this implementation, terminal 700 can achieve the above. Figure 3 The entire process of the method embodiment described herein, and the same or corresponding technical effects achieved, will not be repeated here to avoid duplication.

[0985] This application embodiment also provides a network-side device, including a processor and a communication interface, wherein the communication interface is coupled to the processor, and the processor is used to run programs or instructions to implement, for example... Figure 2 or Figure 3 The steps of the method embodiment shown are illustrated. All implementation processes and methods of the above method embodiments can be applied to this network-side device embodiment and achieve the same technical effect.

[0986] Specifically, embodiments of this application also provide a network-side device, which can be... Figure 4 The random access event handling device shown, or Figure 5 The AI ​​model training device shown is as follows. Figure 8As shown, the network-side device 800 includes: an antenna 81, a radio frequency (RF) device 82, a baseband device 83, a processor 84, and a memory 85. The antenna 81 is connected to the RF device 82. In the uplink direction, the RF device 82 receives information through the antenna 81 and transmits the received information to the baseband device 83 for processing. In the downlink direction, the baseband device 83 processes the information to be transmitted and sends it to the RF device 82. The RF device 82 processes the received information and transmits it through the antenna 81.

[0987] The method executed by the network-side device in the above embodiments can be implemented in the baseband device 83, which includes a baseband processor.

[0988] Baseband device 83 may include, for example, at least one baseband board on which multiple chips are disposed, such as... Figure 8 As shown, one of the chips is, for example, a baseband processor, which is connected to the memory 85 via a bus interface to call the program in the memory 85 and execute the network device operation shown in the above method embodiment.

[0989] The network-side device may also include a network interface 86, such as a Common Public Radio Interface (CPRI).

[0990] Specifically, the network-side device 800 in this application embodiment further includes: instructions or programs stored in memory 85 and executable on processor 84, wherein processor 84 calls the instructions or programs in memory 85 to execute. Figure 4 or Figure 5 The methods executed by each module shown achieve the same technical effect, and to avoid repetition, they will not be described in detail here.

[0991] Specifically, embodiments of this application also provide a network-side device. For example... Figure 9 As shown, the network-side device 900 includes a processor 901, a network interface 902, and a memory 903. This network-side device can be... Figure 4 The random access event handling device shown, or Figure 5 The AI ​​model training device shown is an example of a network interface 902, such as a common public radio interface (CPRI).

[0992] Specifically, the network-side device 900 in this application embodiment further includes: instructions or programs stored in memory 903 and executable on processor 901, wherein processor 901 calls the instructions or programs in memory 903 to execute. Figure 4 or Figure 5The methods executed by each module shown achieve the same technical effect, and to avoid repetition, they will not be described in detail here.

[0993] This application also provides a readable storage medium storing a program or instructions. When the program or instructions are executed by a processor, they implement the various processes of the above-described random access event handling method embodiment or the various processes of the above-described AI model training method embodiment, and can achieve the same technical effect. To avoid repetition, they will not be described again here.

[0994] The processor mentioned above is the processor in the terminal described in the above embodiments. The readable storage medium includes computer-readable storage media, such as computer read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk. In some examples, the readable storage medium may be a non-transient readable storage medium.

[0995] This application embodiment also provides a chip, which includes a processor and a communication interface. The communication interface is coupled to the processor. The processor is used to run programs or instructions to implement the various processes of the above-described random access event handling method embodiment or the various processes of the above-described AI model training method embodiment, and can achieve the same technical effect. To avoid repetition, it will not be described again here.

[0996] It should be understood that the chip mentioned in the embodiments of this application may also be referred to as a system-on-a-chip, system chip, chip system, or system-on-a-chip, etc.

[0997] This application also provides a computer program / program product, which is stored in a storage medium and executed by at least one processor to implement the various processes of the above-described random access event handling method embodiment or the various processes of the above-described AI model training method embodiment, and can achieve the same technical effect. To avoid repetition, it will not be described again here.

[0998] This application also provides a communication system, including a first device and a second device. The first device can be used to perform the steps of the random access event processing method described above, and the second device can be used to perform the steps of the AI ​​model training method described above.

[0999] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element. Furthermore, it should be noted that the scope of the methods and apparatuses in the embodiments of this application is not limited to performing functions in the order shown or discussed, but may also include performing functions substantially simultaneously or in the reverse order, depending on the functions involved. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.

[1000] From the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of computer software products plus necessary general-purpose hardware platforms, and of course, they can also be implemented by hardware. The computer software product is stored in a storage medium (such as ROM, RAM, magnetic disk, optical disk, etc.) and includes several instructions to cause the terminal or network-side device to execute the methods described in the various embodiments of this application.

[1001] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other implementations under the guidance of this application without departing from the spirit and scope of the claims. All of these implementations are within the protection scope of this application.

Claims

1. A method for handling random access events, characterized in that, include: The first device acquires the first output information of the first artificial intelligence (AI) model, and the first output information is related to the random access event; The first device performs a first operation based on the first output information.

2. The method according to claim 1, characterized in that, The first operation includes at least one of the following: The first information is reported to the training node of the first AI model, and the first information includes at least one of the following: the input information of the first AI model, the output information, and the current status information of the first device; Revert to the non-AI random access event triggering process; Trigger a first action of the first AI model, the first action including at least one of switching, updating, changing input information, training, fine-tuning and supervision; Wherein, if the first device is a terminal, the first operation further includes at least one of the following: Measurements are performed on at least one specific beam or reference signal; Stop the measurement on at least one beam or reference signal; Continuous measurement or cessation of measurement over a period of time; The measurement is stopped on at least one beam or reference signal for a period of time; Increase the measurement cycle; Switch to a new neighborhood; Switch to a new first object for data transmission and reception, the first object including at least one of beam, frequency band, carrier, subband, transmit / receive point (TRP), and bandwidth portion (BWP); Add auxiliary cells; Random access is performed based on the prediction of possible consecutive LBT (Listen Before You Talk) failures. Small data transmissions are performed in the idle or inactive state based on the prediction of possible small data transmissions. Activate or deactivate on-demand signals; Initiate a scheduling request (SR); Beam initiation failure recovery BFR; Initiate uplink synchronization; Initiate the random access procedure corresponding to the radio link failure RLF; Initiate an access procedure based on the non-random access channel RACH.

3. The method according to claim 2, characterized in that, The first device performs a first operation based on the first output information, including: If the first condition is met, the first device performs the first operation based on the first output information; The first condition includes at least one of the following: No output information meeting the preset performance requirements was obtained after the first preset time period had elapsed; The first AI model either failed to complete AI inference, or it successfully completed AI inference using the first AI model. The first AI model was used for N AI inferences, where N is a positive integer.

4. The method according to any one of claims 1-3, characterized in that, The first output information includes at least one of the following: Information related to upward movement failure; RACH related information; SR-related information; The time period during which beam failure (BF) or beam refraction (RLF) will not occur; The probability that BF or RLF will not occur; BFR methods or procedures; At least one candidate beam or a reference signal corresponding to the candidate beam for beam failure detection (BFD) or radio link failure detection (RLFD); The pre-compensation information for the first parameter, wherein the first parameter includes at least one of frequency offset, time delay offset, phase offset, Doppler offset, and timing offset; The first device may use at least one beam in the future and the probability of using each beam; The first device may use at least one timing information in a future second time and the probability of using each timing information; Whether the first device will enter the first state and the probability of entering the first state, the first state includes any of the following: idle state, inactive state, disconnected state, connected state and power saving state; Information related to the auxiliary community; Information related to LBT failure; Information related to data transmission in idle or inactive states; Information related to triggering on-demand signals; Uplink or downlink data arrives.

5. The method according to any one of claims 1-4, characterized in that, The first AI model infers the first output information based on the first input information, wherein the first input information includes at least one of the following: The measurement value of the second object, the second object including at least one of beam, reference signal, cell, frequency band, carrier, subband, transmit / receive point (TRP), and bandwidth portion (BWP); The changes in the measured values ​​of the second object; Distance information; Path loss information; Configuration information of the reference signal; Beam configuration and deployment information; Identification information for a cell or cell group or TRP or TRP group or timed advance group TAG or tracking area or wireless access network notification area; Location-related information for base stations, TRPs, network nodes, terminals, or satellite equipment; Mobile-related information of network nodes, terminals, or satellite equipment; Business information or load information; Transmission power information of base stations, TRPs, network nodes, terminals, or satellite equipment; Information regarding the orientation of the antenna panel of a base station, TRP, network node, terminal, or satellite equipment; The terminal's battery status; Network information supported by the terminal; The information obtained by the first device through sensing; Weather information; Non-terrestrial network (NTN) related information; Multipath information of the channel; Channel fading information; Time information; Frequency band information, frequency zone information, frequency point information, carrier frequency information, frequency band layer information, or bandwidth portion BWP information; The number or percentage of successful or failed LBTs over a period of time; The number of times and / or the time for sending small data in the idle or inactive state; Information related to on-demand signals; Information related to SR; The number of times and / or the duration of upward step loss; The number of beam failures and / or the duration of the failures; The number of cell handovers and / or the duration of handovers; The number of times and / or the time when upstream data arrives; The number of times and / or the time when downlink data arrives.

6. The method according to claim 5, characterized in that, The measured value is related to at least one of the following: signal strength information, signal quality information, interference signal strength information, and timing advance (TA) related information.

7. The method according to claim 5, characterized in that, The changes in the measured values ​​include at least one of the following: The amount of increase or decrease in the measured value over a period of time; The percentage increase or decrease of the measured value over a period of time; Whether the increase or decrease in the measured value over a period of time exceeds a first preset threshold; The number of times or the duration during which the increase or decrease of the measured value exceeds a first preset threshold within a certain period of time.

8. The method according to claim 5, characterized in that, The relevant information of the on-demand signal includes at least one of the following: Does the first device support the on-demand signal? The number of times the on-demand signal is activated or deactivated within a certain period of time; The time of the most recent activation or deactivation of the on-demand signal.

9. The method according to claim 5, characterized in that, The relevant information of the SR includes at least one of the following: The number of times SR occurs within a certain period of time; The number of SRs based on the random access channel RACH that occur within a certain period of time; The number of SRs based on the Physical Uplink Control Channel (PUCCH) that occur within a certain period of time; The time of the most recent SR failure.

10. The method according to claim 5, characterized in that, Before the first device acquires the first output information of the first AI model, the method further includes: The first device triggers the first AI model to perform inference when the first triggering condition is met; The first triggering condition includes at least one of the following: Periodic triggering; Obtain the first configuration information; Obtain the activation information that activates the inference of the first AI model; The timer associated with the first AI model timed out; Random access failure exceeds the second preset time; The conditions for cell handover are met; Receive at least one of the first input information; Beam failure detected; Wireless link failure detected; BFI-related timers timed out; The increase or decrease in the measured value exceeds the first preset threshold; The duration during which the increase or decrease in the measured value exceeds the first preset threshold is greater than the third preset duration; The duration of no data transmission exceeds the fourth preset duration; The movement of the terminal or satellite conforms to the preset conditions; The number of times the first event occurs within a certain period of time exceeds a second preset threshold. The first event includes at least one of the following: uplink synchronization failure, downlink synchronization failure, cell handover, beam failure, radio link failure, uplink data arrival, downlink data arrival, random access failure, beam failure instance (BFI), LBT success or failure, small data transmission in idle or inactive state, on-demand signal activation or deactivation, SR, RACH-based SR, and PUCCH-based SR. The second preset threshold may be the same or different for different first events. The time of the most recent occurrence of the first event is less than or equal to a first preset time, and the first preset time may be the same or different for different first events.

11. The method according to claim 10, characterized in that, The movement of the terminal or satellite conforms to a preset condition, including at least one of the following: The terminal or satellite moves at a speed greater than a first preset speed or less than a second preset speed; The change in the moving speed of the terminal or satellite exceeds the fifth preset threshold. The acceleration of the terminal or satellite is greater than the first preset acceleration; The angle by which the terminal or satellite changes direction is greater than the first preset angle.

12. The method according to claim 10, characterized in that, The first device triggers the first AI model to perform inference when a first triggering condition is met, including: The first device triggers the first AI model to perform inference when the first triggering condition is met and the first instruction information is received.

13. The method according to claim 1, characterized in that, The first device is a terminal. Before the first device obtains the first output information of the first AI model, the method further includes: The first device receives first configuration information sent by the network-side device, and the first configuration information is used by the first device to obtain the first output information of the first AI model.

14. The method according to claim 10 or 13, characterized in that, The first configuration information includes at least one of the following: The first AI model or the model identifier of the first AI model; The application scope of the first AI model's reasoning; The inference cycle of the first AI model; The effective duration of the first AI model's inference; The first triggering condition for the inference of the first AI model; Configuration information of the first AI model; Information regarding whether the first AI model supports joint inference between network-side devices and terminals; Activate the switch for inference in the first AI model; Deactivate the switch for inference in the first AI model.

15. The method according to claim 1, characterized in that, The first device acquires the first output information from the inference output of the first AI model, including: The first device determines whether the inference performance of the first AI model meets the preset performance requirements based on the first indicator; If the inference performance of the first AI model meets the preset performance requirements, the first device acquires the first output information of the inference output of the first AI model. The first indicator includes at least one of the following: The complexity of the first AI model; The inference latency of the first AI model; The success rate of the first AI model's inference; The reliability of the inference results of the first AI model.

16. The method according to claim 10, characterized in that, Before the first device acquires the first output information of the first AI model, the method further includes: The first device trains the first AI model based on the second input information; or... The first device acquires the first AI model sent by the second device, wherein the first AI model is an AI model trained by the second device based on the second input information; The second input information includes at least one of the following: At least one of the first input information; Labels used for training the first AI model; Algorithm or algorithm index used for training the first AI model; The loss function used for training the first AI model; Adjustment information used for training the first AI model; The triggering conditions for training the first AI model.

17. The method according to claim 16, characterized in that, The labels used for training the first AI model include at least one of the following: The time of occurrence of the second event, which includes at least one of uplink synchronization failure, cell handover, beam failure, radio link failure, RACH, LBT failure, small data transmission, SR failure, on-demand signal activation or deactivation, uplink data arrival, and downlink data arrival. The period of time during which the second event did not occur; Types of RACH; Timing advance or timing advance offset; Channel type used for SR transmission; PUCCH format used for SR transmission; Methods or procedures for beam failure recovery; The cell ID of the target cell for the handover or at least one candidate cell; The index of the target beam or at least one candidate beam for beam failure recovery; The target beam or the reference signal associated with at least one candidate beam for beam failure recovery; The community ID of the added auxiliary community.

18. The method according to claim 16, characterized in that, The triggering conditions for training the first AI model include at least one of the following: Second trigger condition; Periodic triggering; Semi-static triggering.

19. The method according to claim 18, characterized in that, The second triggering condition includes at least one of the following: At least one of the first triggering conditions; The device that trained the first AI model acquires the second instruction information; The first AI model failed to infer. The first AI model fails inference K times consecutively, where K is an integer greater than 1; Inference is performed using the first AI model; The timer for training the first AI model timed out for longer than the fifth preset duration; L instances of model supervision occur, where L is a positive integer; Obtain at least one of the second input information.

20. The method according to claim 17, characterized in that, The first AI model training is completed when the second condition is met; wherein the second condition includes at least one of the following: The loss function satisfies a preset index; The first AI model has been trained a first preset number of times; The first AI model has undergone a second preset number of iterations for fine-tuning. At least one of the labels used for training the first AI model meets a preset requirement; The training time of the first AI model is greater than or equal to the sixth preset time. The training energy consumption of the first AI model is greater than or equal to the preset energy consumption.

21. The method according to claim 5, characterized in that, When the first device performs inference on the first AI model based on the first input information, the method further includes: The first device performs model supervision on the first AI model based on the second configuration information; The second configuration information includes at least one of the following: The model identifier of the first AI model; The period of the model supervision; The number of times the model is supervised; The duration of the model supervision; Information related to the window monitored by the model; The triggering conditions for model supervision; The labels of the model supervision; The metrics for model supervision.

22. The method according to claim 21, characterized in that, The triggering conditions for model supervision include at least one of the following: The third event failure is determined by using the inference result of the first AI model. The third event includes at least one of the following: cell handover, beam failure recovery, and radio link failure recovery. A fourth event occurs within a specific time period using the inference results of the first AI model, and the fourth event includes at least one of the following: uplink synchronization failure, downlink synchronization failure, beam failure, and wireless link failure. The error in the inference result of the first AI model is greater than the preset error; The first indicator used to determine whether the performance of the first AI model inference meets the preset performance requirements does not meet the first preset indicator; The metrics monitored by the model do not meet the second preset metrics; The timer used for supervising the first AI model timed out; The first AI model failed inference more times than the third preset number of times; Inference is performed using the first AI model; The output information of the first AI model is predefined output information, and the probability of the predefined output information appearing is greater than or equal to a first preset probability, or less than or equal to a second preset probability; The deviation value of the predefined output information is greater than the sixth preset threshold.

23. The method according to any one of claims 1-22, characterized in that, The first device has a first capability, which includes at least one of the following: The ability to support the training of the first AI model; Supports the ability to perform inference using the first AI model; When the first device is a terminal or a server, it supports the ability to report at least one type of auxiliary information for training or inference of the first AI model; or when the first device is a network-side device, it supports the ability to indicate at least one type of auxiliary information for training or inference of the first AI model.

24. The method according to claim 23, characterized in that, The first capability is determined based on at least one of the following: The first device type; Reference signal indication; Uplink control information; Radio Resource Control (RRC) signaling; Specific interface messages.

25. An AI model training method, characterized in that, include: The second device trains the first AI model based on the second input information and sends the trained first AI model to the first device. The trained first AI model is used by the first device to obtain first output information, which is related to the random access event. The second input information includes at least one of the following: First input information, the first AI model obtains the first output information based on the first input information; Labels used for training the first AI model; Algorithm or algorithm index used for training the first AI model; The loss function used for training the first AI model; Adjustment information used for training the first AI model; The triggering conditions for training the first AI model.

26. The method according to claim 25, characterized in that, The labels used for training the first AI model include at least one of the following: The time of occurrence of the second event, which includes at least one of uplink synchronization failure, cell handover, beam failure, radio link failure, RACH, LBT failure, small data transmission, SR failure, on-demand signal activation or deactivation, uplink data arrival, and downlink data arrival. The period of time during which the second event did not occur; Types of RACH; Timing advance or timing advance offset; Channel type used for SR transmission; PUCCH format used for SR transmission; Methods or procedures for beam failure recovery; The cell ID of the target cell for the handover or at least one candidate cell; The index of the target beam or at least one candidate beam for beam failure recovery; The target beam or the reference signal associated with at least one candidate beam for beam failure recovery; The community ID of the added auxiliary community.

27. The method according to claim 25, characterized in that, The triggering conditions for training the first AI model include at least one of the following: Second trigger condition; Periodic triggering; Semi-static triggering.

28. The method according to claim 27, characterized in that, The second triggering condition includes at least one of the following: The second device obtains the second instruction information; The first AI model failed to infer. The first AI model fails inference K times consecutively, where K is an integer greater than 1; Inference is performed using the first AI model; The timer for training the first AI model timed out for longer than the fifth preset duration; L instances of model supervision occur, where L is a positive integer; Obtain at least one of the second input information.

29. The method according to claim 25, characterized in that, The first AI model training is completed when the second condition is met; wherein the second condition includes at least one of the following: The loss function satisfies a preset index; The first AI model has been trained a first preset number of times; The first AI model has undergone a second preset number of iterations for fine-tuning. At least one of the labels used for training the first AI model meets a preset requirement; The training time of the first AI model is greater than or equal to the sixth preset time. The training energy consumption of the first AI model is greater than or equal to the preset energy consumption.

30. A random access event processing apparatus, the apparatus being applied to a first device, characterized in that, The device includes: The acquisition module is used to acquire the first output information of the first AI model, which is related to the random access event; An execution module is used to perform a first operation based on the first output information.

31. The apparatus according to claim 30, characterized in that, The first operation includes at least one of the following: The first information is reported to the training node of the first AI model, and the first information includes at least one of the following: the input information of the first AI model, the output information, and the current status information of the first device; Revert to the non-AI random access event triggering process; Trigger a first action of the first AI model, the first action including at least one of switching, updating, changing input information, training, fine-tuning and supervision; Wherein, if the first device is a terminal, the first operation further includes at least one of the following: Measurements are performed on at least one specific beam or reference signal; Stop the measurement on at least one beam or reference signal; Continuous measurement or cessation of measurement over a period of time; The measurement is stopped on at least one beam or reference signal for a period of time; Increase the measurement cycle; Switch to a new neighborhood; Switch to a new first object for data transmission and reception, the first object including at least one of beam, frequency band, carrier, subband, transmit / receive point (TRP), and bandwidth portion (BWP); Add auxiliary cells; Random access is performed based on the prediction of possible consecutive LBT (Listen Before You Talk) failures. Small data transmissions are performed in the idle or inactive state based on the prediction of possible small data transmissions. Activate or deactivate on-demand signals; Initiate a scheduling request (SR); Beam initiation failure recovery BFR; Initiate uplink synchronization; Initiate the random access procedure corresponding to the radio link failure RLF; Initiate an access procedure based on the non-random access channel RACH.

32. The apparatus according to claim 30, characterized in that, The first output information includes at least one of the following: Information related to upward movement failure; RACH related information; SR-related information; The time period during which beam failure (BF) or beam refraction (RLF) will not occur; The probability that BF or RLF will not occur; BFR methods or procedures; At least one candidate beam or a reference signal corresponding to the candidate beam for BFD or RLFD; The pre-compensation information for the first parameter, wherein the first parameter includes at least one of frequency offset, time delay offset, phase offset, Doppler offset, and timing offset; The first device may use at least one beam in the future and the probability of using each beam; The first device may use at least one timing information in a future second time and the probability of using each timing information; Whether the first device will enter the first state and the probability of entering the first state, the first state includes any of the following: idle state, inactive state, disconnected state, connected state and power saving state; Information related to the auxiliary community; Information related to LBT failure; Information related to data transmission in idle or inactive states; Information related to triggering on-demand signals; Uplink or downlink data arrives.

33. An AI model training device, wherein the device is applied to a second device, characterized in that, The device includes: The training module is used to train the first AI model based on the second input information and send the trained first AI model to the first device. The trained first AI model is used by the first device to obtain first output information, which is related to the random access event. The second input information includes at least one of the following: First input information, the first AI model obtains the first output information based on the first input information; Labels used for training the first AI model; Algorithm or algorithm index used for training the first AI model; The loss function used for training the first AI model; Adjustment information used for training the first AI model; The triggering conditions for training the first AI model.

34. The apparatus according to claim 33, characterized in that, The labels used for training the first AI model include at least one of the following: The time of occurrence of the second event, which includes at least one of uplink synchronization failure, cell handover, beam failure, radio link failure, RACH, LBT failure, small data transmission, SR failure, on-demand signal activation or deactivation, uplink data arrival, and downlink data arrival. The period of time during which the second event did not occur; Types of RACH; Timing advance or timing advance offset; Channel type used for SR transmission; PUCCH format used for SR transmission; Methods or procedures for beam failure recovery; The cell ID of the target cell for the handover or at least one candidate cell; The index of the target beam or at least one candidate beam for beam failure recovery; The target beam or the reference signal associated with at least one candidate beam for beam failure recovery; The community ID of the added auxiliary community.

35. The apparatus according to claim 33, characterized in that, The first AI model training is completed when the second condition is met; wherein the second condition includes at least one of the following: The loss function satisfies a preset index; The first AI model has been trained a first preset number of times; The first AI model has undergone a second preset number of iterations for fine-tuning. At least one of the labels used for training the first AI model meets a preset requirement; The training time of the first AI model is greater than or equal to the sixth preset time. The training energy consumption of the first AI model is greater than or equal to the preset energy consumption.

36. A communication device, characterized in that, It includes a processor and a memory, the memory storing a program or instructions that can run on the processor, the program or instructions being executed by the processor to implement the steps of the random access event handling method as described in any one of claims 1-24, or to implement the steps of the AI ​​model training method as described in any one of claims 25-29.

37. A readable storage medium, characterized in that, The readable storage medium stores a program or instructions that, when executed by a processor, implement the steps of the random access event handling method as described in any one of claims 1-24, or the steps of the AI ​​model training method as described in any one of claims 25-29.

38. A computer program product, characterized in that, The computer program product is stored in a storage medium and is executed by at least one processor to implement the steps of the random access event handling method as described in any one of claims 1-24, or the steps of the AI ​​model training method as described in any one of claims 25-29.

Images