Beam failure detection and recovery

Abstract

Embodiments of the present disclosure relate to beam failure detection and recovery. Embodiments relate to an apparatus and a method. The method comprises detecting a beam failure detection (BFD) prediction start condition related to a first radio access technology (RAT) cell providing a serving beam; obtaining first radio resource management (RRM) measurement samples from the first RAT cell within a first observation window; obtaining second RRM measurement samples from a second RAT cell within a second observation window; using a plurality of trained machine learning models to: predict a first BFD status of the serving beam within a first prediction window based on the first RRM measurement samples; and predict a second BFD status of a strongest beam of the second RAT cell within a second prediction window based on the second RRM measurement samples; trigger performance of a plurality of early beam failure recovery (BFR) actions based on the first BFD status and / or the second BFD status; and transmitting, to the first RAT cell and / or the second RAT cell, a BFD prediction report comprising the first BFD status and / or the second BFD status after performance of the plurality of early BFR actions.

Description

Technical Field

[0001] Various example implementations relate to beam fault detection and recovery. Background Technology

[0002] 5G communication, especially at higher frequencies, uses massive MIMO with beamforming. Beamforming is a signal processing technique that allows the gNB to send a targeted data beam to the user to reduce interference and utilize the spectrum more efficiently, while improving spectral efficiency. When users are indoors or mobile, the radio link between the user equipment (UE) and the gNB is susceptible to congestion and degradation of radio frequency (RF) signals, which can suddenly interrupt the communication link, leading to beam failure.

[0003] In traditional UE-side beam fault detection (BFD), the UE monitors the radio link quality (RLQ) of the serving beam. The number of beam fault instances increments as the RLQ deviates from the out-of-synchronization threshold. A beam fault is detected when the maximum number of consecutive beam fault instances is reached, where two consecutive instances are separated by a maximum of one BFD timer. Upon detection of a beam fault, the UE triggers a RACH procedure for beam fault recovery (BFR). This type of BFD / BFR is inherently time-consuming and slow.

[0004] 5G / 6G Multiple RAT Spectrum Sharing (MRSS) is a key feature for the migration to 6G because it allows NR and 6G cells to share (multiple) the same carriers to dynamically adapt to service requirements. New designs and enhancements are needed for inter-RAT beam management and handover processes in MRSS deployment. Summary of the Invention

[0005] The independent claims define the scope of protection sought for each embodiment. Embodiments and features described in this specification that are not within the scope of the independent claims (if any) are to be interpreted as examples useful for understanding the various embodiments.

[0006] Various aspects include a method, an apparatus, and a computer-readable medium comprising a computer program or signals stored therein, characterized by the content described in the independent claim. Various details of the embodiments are disclosed in the dependent claims, as well as in the corresponding drawings and description.

[0007] According to a first aspect, an apparatus includes: components for detecting beam fault detection (BFD) prediction start conditions associated with a first Radio Access Technology (RAT) cell providing a serving beam; components for acquiring a first Radio Resource Management (RRM) measurement sample from the first RAT cell within a first observation window; components for acquiring a second RRM measurement sample from a second RAT cell within a second observation window; components for performing the following using one or more trained machine learning models: predicting a first BFD state for the serving beam within the first prediction window based on the first RRM measurement sample; and predicting a second BFD state for the strongest beam of the second RAT cell within the second prediction window based on the second RRM measurement sample; components for triggering the execution of one or more early beam fault recovery (BFR) actions based on the first BFD state and / or the second BFD state; and components for sending a BFD prediction report to the first RAT cell and / or the second RAT cell after executing one or more early BFR actions, wherein the BFD prediction report includes the first BFD state and / or the second BFD state.

[0008] According to one embodiment, the apparatus includes components for triggering the execution of one or more early BFR actions independently of instructions from a first RAT cell or a second RAT cell.

[0009] According to one embodiment, detecting the BFD prediction start condition includes monitoring the radio link quality (RLQ) of the serving beam.

[0010] According to one embodiment, the BFD prediction start condition includes a threshold for the RLQ of the serving beam.

[0011] According to one embodiment, the threshold is set to the out-of-sync threshold Q of the first RAT cell. out,LR .

[0012] According to one embodiment, detecting the BFD prediction start condition includes receiving a beam fault instance (BFI) from a first RAT cell.

[0013] According to one embodiment, the BFD prediction start condition includes the BFI counter being equal to 1.

[0014] According to one embodiment, detecting the start condition for BFD prediction includes detecting the incrementing of the BFI counter from 0 to 1.

[0015] According to one embodiment, the prediction is performed using a direct prediction method, and the BFD prediction report includes probability confidence levels associated with a first BFD state and / or a second BFD state.

[0016] According to one embodiment, the prediction is performed using an indirect prediction method, and the BFD prediction report includes the mean absolute error used when predicting the first BFD state and / or the second BFD state.

[0017] According to one embodiment, the apparatus includes components for transmitting capability information to a first RAT cell indicating support for BFD prediction and / or early BFR actions.

[0018] According to one embodiment, the apparatus includes components for receiving a configuration of one or more of the following: BFD prediction start conditions, the length of a first observation window, the length of a second observation window, the length of a first prediction window, the length of a second prediction window, a prediction method to be used to predict a first BFD state and / or a second BFD state, or one or more early BFR actions.

[0019] According to one embodiment, the configuration of the BFD prediction start condition includes a threshold for the Radio Link Quality (RLQ) of the serving beam.

[0020] According to one embodiment, the threshold of the RLQ of the serving beam in the configuration of the BFD prediction start condition is set to the out-of-sync threshold Q of the first RAT cell. out,LR .

[0021] According to one embodiment, the apparatus includes components for: determining a predicted reliability score for a predicted first BFD state and a second BFD state; comparing the predicted reliability score with a reliability threshold; and triggering the execution of one or more early BFR actions based on the predicted reliability score meeting the reliability threshold.

[0022] According to one embodiment, the apparatus includes components for continuing prediction until the predicted reliability score meets a reliability threshold.

[0023] According to a second aspect, an apparatus includes at least one processor; and at least one memory storing instructions, which, when executed by the at least one processor, cause the apparatus to at least: detect a beam fault detection (BFD) prediction start condition associated with a first Radio Access Technology (RAT) cell providing a serving beam; acquire a first Radio Resource Management (RRM) measurement sample from the first RAT cell within a first observation window; acquire a second RRM measurement sample from a second RAT cell within a second observation window; use one or more trained machine learning models to: predict a first BFD state for the serving beam within the first prediction window based on the first RRM measurement sample; and predict a second BFD state for the strongest beam of the second RAT cell within the second prediction window based on the second RRM measurement sample; trigger the execution of one or more early beam fault recovery (BFR) actions based on the first BFD state and / or the second BFD state; and after executing one or more early BFR actions, send a BFD prediction report to the first RAT cell and / or the second RAT cell, wherein the BFD prediction report includes the first BFD state and / or the second BFD state.

[0024] According to a third aspect, a method includes: detecting beam fault detection (BFD) prediction start conditions associated with a first Radio Access Technology (RAT) cell providing the serving beam; acquiring a first Radio Resource Management (RRM) measurement sample from the first RAT cell within a first observation window; acquiring a second RRM measurement sample from a second RAT cell within a second observation window; using one or more trained machine learning models to: predict a first BFD state for the serving beam within the first prediction window based on the first RRM measurement sample; and predict a second BFD state for the strongest beam of the second RAT cell within the second prediction window based on the second RRM measurement sample; triggering the execution of one or more early beam fault recovery (BFR) actions based on the first BFD state and / or the second BFD state; and after executing one or more early BFR actions, sending a BFD prediction report to the first RAT cell and / or the second RAT cell, wherein the BFD prediction report includes the first BFD state and / or the second BFD state.

[0025] According to a fourth aspect, a computer program product includes instructions that, when executed by at least one processor, cause an apparatus or system to at least: detect a beam fault detection (BFD) prediction start condition associated with a first Radio Access Technology (RAT) cell providing the serving beam; acquire a first Radio Resource Management (RRM) measurement sample from the first RAT cell within a first observation window; acquire a second RRM measurement sample from a second RAT cell within a second observation window; use one or more trained machine learning models to: predict a first BFD state for the serving beam within the first prediction window based on the first RRM measurement sample; and predict a second BFD state for the strongest beam of the second RAT cell within the second prediction window based on the second RRM measurement sample; trigger the execution of one or more early beam fault recovery (BFR) actions based on the first BFD state and / or the second BFD state; and after executing one or more early BFR actions, send a BFD prediction report to the first RAT cell and / or the second RAT cell, wherein the BFD prediction report includes the first BFD state and / or the second BFD state.

[0026] According to one embodiment, the computer program product is embodied on a non-transitory computer-readable medium. Attached Figure Description

[0027] In the following description, various embodiments will be described in more detail with reference to the accompanying drawings, in which:

[0028] Figure 1 A schematic block diagram of an apparatus for use in conjunction with an embodiment is shown;

[0029] Figure 2 The layout of the apparatus according to an example embodiment is illustrated schematically;

[0030] Figure 3 A portion of an exemplary wireless access network is shown;

[0031] Figure 4 An example beam failure recovery process is shown;

[0032] Figure 5 and Figure 6 An example beam fault detection process is shown;

[0033] Figure 7 An example of a contention-based random access (CBRA) procedure for beam fault recovery is shown;

[0034] Figure 8 An example of a contention-free random access (CFRA) procedure for beam failure recovery is shown;

[0035] Figure 9 An example MRSS framework is shown;

[0036] Figure 10 An example of the first beam set of the first RAT (6G cell) and the second beam set of the second RAT (5G cell) in MRSS is shown;

[0037] Figure 11 This is a flowchart of the method according to the embodiment;

[0038] Figure 12 This is an example time series plot of BFD prediction in an MRSS scenario;

[0039] Figure 13 (with partial view) Figure 13A and Figure 13B The document provides example implementation diagrams of embodiments related to direct prediction. Specifically, Figure 13 shows a view... Figure 13A An example training module is shown, for example, in Figure 13. Figure 13B Examples of an inference module and an recovery module are shown. The approximate relationship between the local views is shown in Figure 13, and the alignment guide on the individual local views allows for alignment of the local views as shown in Figure 13;

[0040] Figure 14 (with partial view) Figure 14A and Figure 14B The document provides example implementation diagrams of embodiments related to indirect prediction. Specifically, Figure 14 shows a view... Figure 14A Examples of training and inference modules are shown in Figure 14. Figure 14B Examples of a fault detection module and an example of a recovery module are shown. The approximate relationship between the local views is illustrated in Figure 14, and alignment guidelines on individual local views allow for alignment of the local views as guided by Figure 14.

[0041] Figure 15 This is a flowchart of the method according to the embodiment;

[0042] Figure 16 (with partial view) Figure 16A and Figure 16B This provides example signaling steps for UE-autonomous BFD prediction and BFR under the MRSS framework. The approximate relationship between local views is shown in Figure 16, and the alignment guide on individual local views allows for alignment of local views under the guidance in Figure 16;

[0043] Figure 17 This is a flowchart of the method according to the embodiments; and

[0044] Figure 18 (with partial view) Figure 18A and Figure 18BThis provides example signaling steps for BFD prediction and BFR in network control under the MRSS framework. The approximate relationship between local views is shown in Figure 18, and the alignment guide on individual local views allows for alignment of local views under the guidance of Figure 18. Detailed Implementation

[0045] The suitable apparatus and possible mechanisms for performing system information acquisition are described in more detail below. Although the following focuses primarily on 5G networks, the embodiments further described below are by no means limited to implementation in the aforementioned networks, but are applicable to any network that supports system information acquisition, particularly 6G and any next-generation networks.

[0046] In this regard, first refer to Figure 1 and Figure 2 ,in Figure 1 A schematic block diagram of an exemplary device or electronic device 50 is shown, which may be combined with an arrangement according to an embodiment. Figure 2 The layout of the device according to an example embodiment is shown. The following will explain... Figure 1 and Figure 2 The elements in.

[0047] Electronic device 50 may be, for example, a mobile terminal or user equipment of a wireless communication system. Device 50 may include a housing 30 for attaching to and protecting the device. Device 50 may also include a display 32 and a keypad 34. Instead of a keypad, the user interface may be implemented as a virtual keyboard or data input system as part of a touch-sensitive display.

[0048] The device may include a microphone 36 or any suitable audio input, which may be a digital or analog signal input. The device 50 may also include an audio output device, such as headphones 38, a speaker, or any analog or digital audio output connection. The device 50 may also include a battery 40 (or the device may be powered by any suitable mobile energy device, such as a solar cell, fuel cell, or clockwork generator). The device may also include a camera 42 capable of recording or capturing images and / or video. The device 50 may also include an infrared port 41 for short-range line-of-sight communication with other devices. In other embodiments, the device 50 may also include any suitable short-range communication solution, such as a Bluetooth wireless connection or a USB / FireWire wired connection.

[0049] Device 50 may include a controller 56 or a processor for controlling device 50. Controller 56 may be connected to memory 58, which may store user data and instructions to be implemented on controller 56. Memory may be random access memory (RAM) and / or read-only memory (ROM). Memory may store computer-readable, computer-executable software, including instructions that, when executed, cause the controller / processor to perform the various functions described herein. In some cases, the software may not be directly executable by the processor, but may cause the computer (e.g., at compile and execution time) to perform the functions described herein. Controller 56 may also be connected to codec circuitry 54, which is adapted to perform encoding and decoding of audio and / or video data or to assist encoding and decoding performed by the controller.

[0050] Device 50 may include a wireless interface circuitry system 52 connected to a controller and adapted to generate wireless communication signals, for example, for communicating with a cellular communication network, a wireless communication system, or a wireless local area network. Device 50 may also include an antenna 44 connected to the wireless interface circuitry system 52 for transmitting radio frequency signals generated at the wireless interface circuitry system 52 to other devices and receiving radio frequency signals from other devices.

[0051] In the following description, radio access architectures based on Advanced Long Term Evolution (LTE Advanced, LTE-A) or New Radio (NR, 5G) will be used as examples of access architectures to which embodiments may be applied, but the embodiments are not limited to such architectures. Those skilled in the art will understand that the embodiments can also be applied to other types of communication networks with appropriate means by suitably adjusting parameters and processes. Some examples of other options for suitable systems are Universal Mobile Telecommunications System (UMTS) Radio Access Network (UTRAN or E-UTRAN), Long Term Evolution (LTE, the same as E-UTRA), Wireless Local Area Network (WLAN or WiFi), Global Microwave Access Interoperability (WiMAX), Bluetooth®, Personal Communication Services (PCS), ZigBee®, Wideband Code Division Multiple Access (WCDMA), systems using Ultra Wideband (UWB) technology, sensor networks, Mobile Ad Hoc Networks (MANET), and Internet Protocol Multimedia Subsystem (IMS), or any combination thereof.

[0052] Figure 3 An example of a simplified system architecture is described, showing only some elements and functional entities, which are logical units whose implementations may differ from those shown. Figure 3 The connections shown are logical connections; the actual physical connections may differ. Those skilled in the art will understand that the system typically also includes, in addition to... Figure 3Other functions and structures beyond those shown. However, the embodiments are not limited to the system given as an example, and those skilled in the art can apply this solution to other communication systems with the necessary properties.

[0053] Figure 3 The example illustrates a portion of an exemplary wireless access network.

[0054] Figure 3 User equipments 300 and 302 are shown, configured to be in a wireless connection state with an access node (such as (e / g)NodeB) 304 providing the cell on one or more communication channels within the cell. The physical link from the user equipment to the (e / g)NodeB is called an uplink or reverse link, while the physical link from the (e / g)NodeB to the user equipment is called a downlink or forward link. It should be understood that the (e / g)NodeB, or its functionality, can be implemented using any entity suitable for this purpose, such as a node, host, server, or access point.

[0055] A communication system typically includes more than one (e / g)NodeB. In this case, the (e / g)NodeBs can also be configured to communicate with each other via wired or wireless links designed for this purpose. These links can be used for signaling purposes. An (e / g)NodeB is a computing device configured to control the radio resources of the communication system coupled to it. A NodeB can also be referred to as a base station, access point, or any other type of interface device, including relay stations capable of operating in a wireless environment. An (e / g)NodeB includes or is coupled to a transceiver. From the transceiver of the (e / g)NodeB, a connection is provided to an antenna element, which establishes a bidirectional radio link to the user equipment. The antenna element may include multiple antennas or antenna elements. The (e / g)NodeB is also connected to the core network 310 (CN or Next Generation Core NGC). Depending on the system, the counterpart on the CN side may be a Serving Gateway (S-GW, used for routing and forwarding user data packets), a Packet Data Network Gateway (P-GW) for providing connectivity between the user equipment (UE) and external packet data networks, or a Mobility Management Entity (MME), etc. A CN can include network entities or nodes that may be referred to as management entities. Examples of network entities include at least the Access and Mobility Management Function (AMF).

[0056] User equipment (also known as user equipment (UE), user terminal, terminal equipment, radio equipment, mobile station (MS), etc.) illustrates a type of device to which resources on the air interface are allocated and assigned, and therefore any features of the user equipment described herein can be implemented by corresponding network devices such as relay nodes, eNBs, and gNBs. An example of such a relay node is a Layer 3 relay (self-backhaul relay) toward a base station.

[0057] User equipment (UE) generally refers to portable computing devices, including wireless mobile communication devices with or without a Subscriber Identity Module (SIM), including but not limited to: mobile stations (mobile phones), smartphones, personal digital assistants (PDAs), cell phones, devices using wireless modems (such as alarm or measuring devices), laptops and / or touchscreen computers, tablets, game consoles, and multimedia devices. It should be understood that UE can also be a virtually exclusively uplink-only device, an example of which is a camera or camcorder that loads images or video clips onto a network. UE can also be a device capable of operating in an Internet of Things (IoT) network, where objects can transmit data over the network without human-to-human or human-to-computer interaction. Therefore, UE can be an IoT device. UE can also utilize the cloud. In some applications, UE may include small portable devices with wireless components (such as watches, headphones, or glasses), and computation may be performed in the cloud. UE (or, in some embodiments, a Layer 3 relay node) is configured to perform one or more UE functions. UE may also be referred to as a subscriber unit, mobile station, remote terminal, access terminal, user terminal, or user equipment (UE), to name just a few.

[0058] The various techniques described in this paper can also be applied to cyber-physical systems (CPS) (systems that collaboratively control computing elements of physical entities). CPS can realize and utilize a large number of interconnected ICT devices (sensors, actuators, processors, microcontrollers, etc.) embedded in different locations within physical objects. Mobile cyber-physical systems, in which the physical systems discussed have inherent mobility, are a subcategory of cyber-physical systems. Examples of mobile physical systems include mobile robots and electronic devices transported by humans or animals.

[0059] Furthermore, although the device is depicted as a single entity, different units, processors, and / or memory units can be implemented. Figure 1 (Not all of them are shown in the image).

[0060] 5G supports the use of multiple-input multiple-output (MIMO) antennas, and far more base stations or nodes than LTE (the so-called small cell concept), including macro sites that cooperate with small cells and employ multiple wireless technologies, depending on service requirements, use cases, and / or available spectrum. Access nodes in a wireless network form transmit / receive (TX / Rx) points (TRPs), and the UE is expected to access networks with at least partially overlapping multi-TRPs, such as macro cells, small cells, pico cells, femtocells, remote radio heads, relay nodes, etc. Access nodes can be equipped with massive MIMO antennas, i.e., very large antenna arrays consisting of, for example, hundreds of antenna elements, implemented in a single antenna panel or multiple antenna panels, capable of communicating with the UE using multiple simultaneous radio beams. The UE can be equipped with MIMO antennas, which have antenna arrays consisting of, for example, dozens of antenna elements, implemented in a single antenna panel or multiple antenna panels. Therefore, a UE can use one beam to access a TRP, use multiple beams to access a TRP, use one (common) beam to access multiple TRPs, or use multiple beams to access multiple TRPs.

[0061] While 4G / LTE networks support some multi-TRP schemes, in 5G NR, multi-TRP features are enhanced, such as the transmission of multiple control signals via multi-TRP, thereby improving link diversity gain. Furthermore, for multi-TRP technology, high carrier frequencies (such as millimeter waves) and massive MIMO antennas require new multi-TRP beam management processes.

[0062] 5G mobile communication supports a wide range of use cases and related applications, including video streaming, augmented reality, different data sharing methods, and various forms of machine-type applications, such as (massive) machine-type communication (mMTC), including vehicle safety, various sensors, and real-time control. 5G is expected to have multiple radio interfaces, namely sub-6GHz, cmWave, and mmWave, and will be able to integrate with existing legacy radio access technologies such as LTE. Integration with LTE can be implemented, at least in the early stages, as a system where LTE provides macro coverage and 5G radio interface access is aggregated to LTE from small cells. In other words, 5G is planned to simultaneously support inter-RAT interoperability (such as LTE-5G) and inter-RI interoperability (inter-radio interface interoperability, such as sub-6GHz-cmWave and sub-6GHz-cmWave-mmWave). One concept considered for use in 5G networks is network slicing, where multiple independent and dedicated virtual subnets (network instances) can be created within the same infrastructure to run services with different requirements for latency, reliability, throughput, and mobility.

[0063] 5G NR frequency bands are divided into two frequency ranges: Frequency Range 1 (FR1), which includes the sub-6GHz bands, the bands traditionally used in previous standards, as well as new bands extended to cover potential new spectrum offerings from 410MHz to 7125MHz; and Frequency Range 2 (FR2), which includes the bands from 24.25GHz to 52.6GHz. Therefore, FR2 encompasses bands within the mmWave range, and due to its shorter range and higher available bandwidth, these bands require some different approaches to radio resource management compared to the bands in FR1.

[0064] The architecture in LTE networks is entirely distributed across the radio and entirely centralized in the core network. Low-latency applications and services in 5G require content to be closer to the radio, leading to local breakout and multi-access edge computing (MEC). 5G enables analytics and knowledge generation to occur at the data source. This approach requires leveraging resources that may not be continuously connected to the network, such as laptops, smartphones, tablets, and sensors. MEC provides a distributed computing environment for application and service hosting. It also has the ability to store and process content near cellular subscribers to accelerate response times. Edge computing encompasses a wide range of technologies, such as wireless sensor networks, mobile data acquisition, mobile signature analytics, collaborative distributed peer-to-peer self-organizing networks and processing (which can also be categorized as local cloud / fog computing and grid / mesh computing), dew computing, mobile edge computing, cloudlets, distributed data storage and retrieval, autonomous self-healing networks, remote cloud services, augmented and virtual reality, data caching, the Internet of Things (IoT) (massive connectivity and / or latency critical), and critical communications (autonomous vehicles, traffic safety, real-time analytics, time-critical control, healthcare applications).

[0065] The communication system can also communicate with other networks such as the public switched telephone network or the Internet 312, or utilize the services provided by them. The communication network may also be able to support the use of cloud services; for example, at least a portion of the core network operation can be performed as a cloud service (this is in...). Figure 3 (As described in “Cloud” 314). The communication system may also include a central control entity that provides facilities for different operators’ networks to cooperate, for example, in spectrum sharing.

[0066] Edge cloud can be introduced into the radio access network (RAN) by leveraging Network Functions Virtualization (NFV) and Software-Defined Networking (SDN). Using an edge cloud means that access node operations are performed, at least partially, in servers, hosts, or nodes that are operationally coupled to remote radio headends or base stations, including the radio portion. Node operations can also be distributed across multiple servers, nodes, or hosts. The application of the cloudRAN architecture enables real-time RAN functions to be executed on the RAN side (in the distributed unit DU) and non-real-time functions to be executed in a centralized manner (in the centralized unit CU 308).

[0067] It should also be understood that the workload allocation between core network operations and base station operations may differ from, or even not exist, in LTE. Some other technological advancements that may be used include big data and all-IP, which could potentially change how networks are built and managed. 5G (or New Radio, NR) networks are designed to support a multi-tiered architecture, where MEC servers can be placed between the core and base stations or gNBs. It should be understood that MEC can also be applied to 4G networks. gNBs are next-generation gNBs (or new gNBs) that support 5G networks (i.e., NR).

[0068] 5G can also leverage non-terrestrial nodes 306 (e.g., access nodes) to enhance or supplement 5G service coverage, for example by providing backhaul, wireless access to wireless devices, service continuity for machine-to-machine (M2M) communications, service continuity for Internet of Things (IoT) devices, service continuity for passengers on board vehicles, ensuring service availability for critical communications, and / or ensuring service availability for future rail / sea / air communications. Non-terrestrial nodes can have a fixed location relative to the Earth's surface, or they can be mobile non-terrestrial nodes that can move relative to the Earth's surface. Non-terrestrial nodes can include satellites and / or HAPS. Satellite communications can utilize geostationary Earth orbit (GEO) satellite systems, or low Earth orbit (LEO) satellite systems, particularly mega-constellations (systems deploying hundreds of (nano) satellites). Each satellite in a mega-constellation can cover several satellite-enabled network entities that create a ground cell. Ground cells can be created via ground relay nodes 304 or gNBs located on the ground or in satellites.

[0069] Those skilled in the art will understand that the described system is merely an example of a part of a wireless access system, and in practice, the system may include multiple (e / g) NodeBs, user equipment may access multiple wireless cells, and the system may also include other devices, such as physical layer relay nodes or other network elements. At least one (e / g) NodeB may be a home (e / g) nodeB. Furthermore, multiple different types of wireless cells and multiple wireless cells may be provided within the geographical area of ​​the wireless communication system. Wireless cells may be macrocells (or umbrella cells), which are large cells typically tens of kilometers in diameter, or smaller cells such as micro, femtocells, or picocells. Figure 1 The (e / g) NodeB can provide any type of these cells. Cellular wireless systems can be implemented as multi-layer networks comprising several cells. Typically, in a multi-layer network, one access node provides one or more cells, and therefore multiple (e / g) NodeBs are required to provide this network structure.

[0070] To meet the need for improved deployment and performance of communication systems, the concept of "plug and play" (e / g) NodeBs was introduced. Typically, in addition to home (e / g) NodeBs (H(e / g) nodeBs), networks capable of using "plug and play" (e / g) NodeBs also include home node B gateways or HNB-GWs ( Figure 1 (Not shown in the image). HNB gateways (HNB-GWs), typically installed within a carrier's network, can aggregate services from a large number of HNBs back to the core network.

[0071] The Radio Resource Control (RRC) protocol is used in various wireless communication systems to define the air interface between the UE and the base station (such as eNB / gNB). 3GPP specifies this protocol in TS 36.331 for LTE and TS 38.331 for 5G. Regarding RRC, the UE can operate in idle or connected mode in LTE and 5G, where the available radio resources depend on the mode in which the UE is currently camped. In 5G, the UE can also operate in inactive mode. In RRC idle mode, the UE has no connection for communication, but can listen for paging messages. In RRC connected mode, the UE can operate in different states, such as CELL_DCH (Dedicated Channel), CELL_FACH (Forward Access Channel), CELL_PCH (Cell Paging Channel), and URA_PCH (URA Paging Channel). The UE can communicate with the eNB / gNB via various logical channels, such as the Broadcast Control Channel (BCCH), Paging Control Channel (PCCH), Common Control Channel (CCCH), Dedicated Control Channel (DCCH), and Dedicated Traffic Channel (DTCH).

[0072] The transitions between states are controlled by the RRC state machine. When the UE powers on, it is in disconnected / idle mode. The UE can enter RRC connected mode via initial attachment or connection establishment. If the UE is inactive for a short period, the eNB / gNB can suspend its session by moving to RRC inactive mode and resume its session by moving to RRC connected mode. The UE can move from RRC connected mode or RRC inactive mode to RRC idle mode.

[0073] The actual user and control data from the network to the UE is transmitted via the downlink physical channel. In 5G, the downlink physical channel includes the physical downlink control channel (PDCCH) carrying the necessary downlink control information (DCI), the physical downlink sharing channel (PDSCH) carrying user data and system information, and the physical broadcast channel (PBCH) carrying the necessary system information to enable the UE to access the 5G network.

[0074] User and control data from the UE to the network is transmitted via uplink physical channels. In 5G, uplink physical channels include the Physical Uplink Control Channel (PUCCH) for uplink control information (including HARQ feedback confirmation, scheduling requests, and downlink channel state information for link adaptation), the Physical Uplink Shared Channel (PUSCH) for uplink data transmission, and the Physical Random Access Channel (PRACH) used by the UE to request connection establishment (referred to as random access).

[0075] The classic random access (RA) procedure consists of multiple steps between the UE and the gNB, which are used by the UE to obtain uplink (UL) synchronization and acquire resources for subsequent communication. The random access procedure has several purposes, but the primary purpose is to (a) inform the gNB that the UE needs to establish a connection, or (b) to provide a signal from the UE to the gNB that allows the gNB to adjust / correct the UE's transmission timing.

[0076] The random access (RA) procedure can be performed as either contention-free random access (CFRA) or contention-based random access (CBRA) configured by the network. The traditional CBRA procedure used from 4G / LTE to 5G is called a 4-step RA procedure, which includes four messages (Msg1, Msg2, Msg3, and Msg4). The following section combines... Figure 7 and Figure 8 Provide example details for these two processes.

[0077] According to the current 5G NR specification, the network configures random access resources in a semi-static manner through broadcasting via System Information Block (SIB) or by using UE-specific RRC configuration.

[0078] The transitions between states are controlled by the RRC state machine. When the UE powers on, it is in disconnected / idle mode. The UE can enter RRC connected mode via initial attachment or connection establishment. If the UE is inactive for a short period, the eNB / gNB can suspend its session by moving to RRC inactive mode and resume its session by moving to RRC connected mode. The UE can move from RRC connected mode or RRC inactive mode to RRC idle mode.

[0079] When the UE is in RRC connection mode, it must periodically perform beam management related operations, such as Tracking Reference Signal / Channel State Information Reference Signal (TRS / CSI-RS) measurement, Synchronization Signal Block (SSB) measurement, and periodically report the results of these measurements to the network.

[0080] When a UE is in RRC idle / inactive mode and is controlled to transition to RRC connected mode, it must acquire System Information (SI) to access the network. The UE's System Information (S1) includes a set of Master Information Blocks (MIBs) and System Information Blocks (SIBs). The MIB contains basic system information and is periodically broadcast on the PBCH. The SIB further includes various scheduling and cell access information broadcast on the PDSCH.

[0081] Beam Fault Detection (BFD) and Beam Fault Recovery (BFR) in NR

[0082] 5G communication, especially at higher frequencies, uses massive MIMO with beamforming. Beamforming is a signal processing technique that allows gNBs to send targeted data beams to users to reduce interference and utilize the spectrum more efficiently, while improving spectral efficiency.

[0083] When a user is indoors or moving, the radio link between the UE and the gNB is susceptible to radio frequency (RF) signal congestion and degradation, which can abruptly interrupt the communication link, leading to beam failure. Detecting this is commonly referred to as beam failure detection (BFD). Therefore, to detect beam failure at the right time, the UE can measure this sudden and rapid change in the communication link and simultaneously recover from it to continue service. The UE can do this with the help of the beam failure recovery (BFR) procedure. The BFR procedure is a combination of UE PHY (physical layer) and MAC (media access control) procedures without involving any higher-layer signaling.

[0084] From a high-level perspective, the beam fault recovery process includes the following four steps: Figure 4 As shown: • Beam fault detection 400 • Optimal beam selection for recovery 402 (SSB or CSI-RS). 3GPP Technical Specification 38.331 specifies candidateBeamRSList, candidateBeamRSListExt-v1610 as a set of reference signals (CSI-RS and / or SSB) for identifying candidate beams and associated RA parameters for recovery. • Beam fault recovery request 404 • Beam fault recovery response 406

[0085] Beam Fault Detection (BFD)

[0086] Figure 5 and Figure 6 The diagram illustrates an example procedure for beam failure detection. Beam failure detection is a combined L1 / L2 procedure, where L1 provides a MAC layer indication of a beam failure instance (BFI). The MAC layer counts the indications and declares failure when the maximum configured number of BFI indications is reached. Whenever the PHY layer detects that the RSRP of the reference signal for the serving beam is below a threshold, assuming 10% BLER for the PDCCH, it triggers a beam failure instance (BFI) and sends it to the MAC.

[0087] The MAC layer starts a timer immediately upon receiving a BFI and increments the counter by 1 for each BFI. When a certain threshold of the BFI is reached, i.e., when... BFI_COUNTER >= beamFailureInstanceMaxCount At this time, MAC will trigger BeamFailure and begin the recovery process, such as Figure 5 As shown.

[0088] There is a situation where, for example Figure 6 As shown, this occurs when L1 stops sending beam fault instances to the MAC layer, or when signal quality improves and L1 no longer detects any problems. In handling such cases, the UE MAC layer runs a timer every time the PHY reports a BFI, and for each occurrence, it will... BFI_COUNTER Increment by 1 and restart the timer. If no beam failure instance is received at the MAC layer and the timer expires, the MAC layer will reset. BFI_COUNTER And it is assumed that there are no more instances of beam failure.

[0089] Beam Fault Recovery (BFR)

[0090] The UE is provided with a set of resources for the recovery process via an RRC message in BeamFailureRecoveryConfig. Beam recovery is achieved by performing random access (RACH) on the best candidate beam selected during the BFR procedure.

[0091] The UE can perform the following two RACH procedures in response to a BFR request: • Contendency-free random access – CFRA RACH • Contention-based random access – CBRA RACH

[0092] BFR's CBRA: When the UE performs CBRA RACH for the recovery procedure, it will act as... Figure 7 The typical contention-based RACH procedure shown is executed as follows: • The UE is in RRC connection mode •UE detects beam faults at the PHY layer • The UE performs candidate beam selection during the recovery process. • The UE triggers the RACH procedure and sends the RACH preamble to the gNB. • gNB sends the Random Access Response (RAR) and UL Authorization for msg3 within the RAR window. • The UE sends a MAC CE (msg3) for contention resolution. • gNB sends UL authorization to UE •BFR completed

[0093] BFR's CFRA: Figure 8 The contention-free RACH procedure for BFR is illustrated. In the case of a CFRA BFR request, the UE will receive a RAR scrambled with C-RNTI instead of RA-RNTI during the regular RACH procedure. The gNB will also configure a dedicated CORSET for BFR, which the UE monitors after sending the RACH preamble. The procedure includes the following steps: • The UE is configured with CORSET-BFR and also with a dedicated preamble for BFR. • If the UE detects a beam fault, it will begin the optimal beam selection process. • If the optimal beam is within any beam configured in the beam failure recovery configuration, the UE selects CFRA; otherwise, it selects the CBRA RACH procedure (in which case, the UE selects CFRA-RACH). • The UE triggers RACH using a dedicated preamble. • The UE monitors the presence of a PDCCH scrambled with C-RNTI in the CORSET-BFR—the UE receives a RAR scrambled with C-RNTI-UE. •BFR completed

[0094] MRSS framework

[0095] Figure 9This demonstrates a Multi-RAT Spectrum Sharing (MRSS) framework with both 5G Standalone (SA) and 6G SA. 5G / 6G MRSS is a key feature for the migration to 6G because it allows NR and 6G cells to share (multiple) the same carriers to dynamically adapt to service requirements.

[0096] MRSS is the preferred enabler for the 5G to 6G migration. The following assumptions are made for Day 1 of 6G: • 6G independent design • Assume that inter-RAT handover (HO) is the process used to move a UE from a 6G cell (or 5G cell) to a 5G cell (or 6G cell).

[0097] In MRSS scenarios, it is assumed that 5G and 6G cells share the same RU (see...). Figure 9 The coverage of 5G MRSS cells and 6G MRSS cells is expected to be the same. Therefore, it is clear that a new design / enhancement is needed for inter-RAT beam management and HO procedures for MRSS deployment.

[0098] Figure 10 Examples of the first beamsets for the first RAT (6G cell) 1010, 1012, 1014 and the second beamsets for the second RAT (5G cell) 1020, 1022, 1024, 1026 in MRSS are shown. Figure 10 As shown, the UE communicating with the first RAT determines the optimal subset of beams to configure for the first RAT cell when measuring the beams of the first and second RAT cells. The second RAT cell shares the same frequency band and the same RU as the first RAT cell. The signaling aspects that enable this solution are described.

[0099] In MRSS scenarios, the UE communicating with the first RAT should consider the beams 1020, 1022, 1024, and 1026 of the second RAT as additional beams to the beam set 1010, 1012, and 1014 of the first RAT. The UE may consider the complete beam. ... , ... The set is measured at least as a combination of the first beam set of the first RAT and the second beam set of the second RAT.

[0100] This paper considers beam management and mobility supporting artificial intelligence / machine learning (AI / ML) in 6G use cases, particularly for fault and event prediction use cases. This is especially important for the MRSS framework.

[0101] Various AI / ML enabling solutions can be used to enhance L3 mobility processes. The solutions described in this paper are specifically designed to address predictive BFD and BFR processes within the MRSS framework. These solutions can be implemented in both single RAT and MRSS scenarios.

[0102] BFRs triggered by iterative processes, such as when the UE must detect a certain number of BFD instances to trigger beam fault recovery, are inherently time-consuming and slow. AI / ML methods offer a valuable solution if the BFD state is predictable, enabling the UE to proactively handle beam faults by leveraging the MRSS framework.

[0103] The embodiments described herein provide a beam failure prediction method designed to address the aforementioned outstanding challenges. This method is well-suited for enhancing beam management and mobility in 5G NR and above (i.e., 6G MRSS).

[0104] The embodiments described herein provide a BFD prediction method utilizing an AI / ML model. To align with the ongoing 3GPP rel.19 research project (SI) on AI / ML mobility measurement events and RLF prediction use cases, prediction can be achieved through two approaches: direct prediction using a classification model and indirect prediction using a regression model. For the direct prediction method, the classification model directly predicts the BFD state (true / false, or 0 / 1 in one-bit representation) within a defined time window. For the indirect prediction method, the UE first predicts the time series of RRM measurements (e.g., including Layer 1 or Layer 3 measurements). Subsequently, the BFD state (true / false, or 0 / 1 in one-bit representation) is derived based on the predicted RRM measurements within the time window. In MRSS scenarios, BFD prediction is performed for both the serving RAT (first RAT) and the adjacent RAT (second RAT).

[0105] To facilitate BFD prediction at the UE, the following steps are proposed. These steps may cover various embodiments as follows.

[0106] RRC configuration parameters and information elements

[0107] The UE may send and / or receive configurations, including but not limited to one or more of the following. These configurations may be sent to and / or received from the network, such as serving / first RAT and / or neighboring / second RAT. These configurations are preferably sent / received by the UE as RRC messages.

[0108] The UE can send UE capability information to support direct or indirect BFD prediction in MRSS.

[0109] The UE can receive predicted start conditions. Predicted start conditions can be determined by a threshold of the serving RAT and an optional offset.

[0110] The UE can receive configuration for prediction timing. Prediction timing can be configured using, for example, two sets of parameters, which include the observation window and prediction window of both the first RAT and the second RAT.

[0111] The UE can receive thresholds used for indirect BFD prediction, such as: • If the average predicted measurement within the prediction window is below this threshold, the indirect BFD prediction is considered true, indicating a potential fault. • If the average predicted measurement remains above the threshold, the indirect BFD prediction is considered false, indicating that there is no impending failure. • A single threshold or two separate thresholds can be used for different RATs.

[0112] The UE can receive UE actions for predictive BFD processing. Alternative UE actions may include, but are not limited to, at least one of the following: • Operation option 1a: Start RACH to the same or different RATs in advance. • Operational Option 1b: Perform a no-RACH handover under Low-Level Triggered Mobility (LTM) to achieve beam fault recovery to candidate beams of different RATs. • Operation Option 2: Reduce the traditional beam fault detection timer for the same or different RATs. • Operation Option 3: Increase RACH uplink transmission power for the same or different RATs.

[0113] The UE can receive UE report configuration and format. The configuration and / or format may include, but is not limited to, at least one of the following: • Reporting criteria: The reporting criteria are most likely to be event-triggered and should be consistent with the configuration conditions that initiate the prediction. • Report content: The content of the report depends on the forecasting method used, which can be direct or indirect forecasting. • Report format: The prediction results can be quantized as a bitmap (at least 2 bits), similar to the format used for L1 SS-RSRP reports.

[0114] Signaling enhancement between UE and MRSS network

[0115] The following outlines two alternative embodiments that can provide signaling enhancement methods between the UE and the MRSS network.

[0116] Example 1: UE Autonomous Prediction and Recovery in MRSS, where the UE is configured to perform prediction of the predicted BFD for both RATs and to perform early BFR actions without network instructions. For more details, please refer to [link to relevant documentation]. Figure 15 The method described in Figure 16.

[0117] Example 2: Network-controlled UE prediction in MRSS, where the UE is configured to perform prediction and then send the prediction report to the network so that the network can instruct the UE to perform early BFR actions on the predicted BFD of these two RATs. See below for more details. Figure 17 The method described in Figure 18.

[0118] The features of Embodiment 1 can be combined with the features of Embodiment A and / or Embodiment B as described below. The features of Embodiment 2 can be combined with the features of Embodiment A and / or Embodiment B as described below.

[0119] Implementation algorithm of BFD prediction for multi-class classification

[0120] The following outlines two alternative implementation algorithms for multi-class classification using BFD prediction. • Example A: Multi-class classification (using direct prediction), where direct prediction involves a multi-class classification model that predicts whether both the service RAT and the second RAT will occur in the upcoming time window. • Example B: Multi-class classification (using indirect prediction), where indirect prediction involves using a regression model to predict the beam measurements of both the serving beam (of the first RAT) and the strongest beam of the second RAT, and then processing the predictions to determine the BFD state.

[0121] The above steps will be described in detail below, including corresponding example signaling diagrams. Other embodiments and implementation details will also be described.

[0122] Implementation of BFD prediction on the UE side (MRSS scenario)

[0123] In the traditional process, as described above, the UE monitors the Radio Link Quality (RLQ) of the serving beam for beam fault detection. When RLQ is greater than Q... out,LR When the time difference is reached, the number of beam fault instances increases. If the maximum number of consecutive beam fault instances is reached, a beam fault is detected, where two consecutive indications are at most... beamFailureDetectionTimer Separation. When a beam fault is detected, the UE will trigger the RACH procedure to recover the beam fault.

[0124] Below, a method for beam fault detection and recovery aims to predict the beam fault detection (BFD) of multiple RATs under the MRSS framework. It is assumed that the device or UE is communicating with the serving beam from the first RAT. It can also be assumed that the UE or device performs measurements on configuration reference signals from two RATs.

[0125] This method is as follows Figure 11 The flowchart illustrates the operation of a terminal device (such as a user equipment (UE)). The method includes: 1100 detecting a beam fault detection (BFD) prediction start condition associated with a first Radio Access Technology (RAT) cell providing the serving beam (for the device); 1102 acquiring a first Radio Resource Management (RRM) measurement sample from the first RAT cell within a first observation window; 1104 acquiring a second RRM measurement sample from a second RAT cell within a second observation window; using one or more trained machine learning models to: predict a first BFD state for the serving beam within the first prediction window based on the first RRM measurement sample; and predict a second BFD state for the strongest beam of the second RAT cell within the second prediction window based on the second RRM measurement sample; and 1108 performing one or more early beam fault recovery (BFR) actions based on the first BFD state and / or the second BFD state.

[0126] An apparatus (such as a UE) includes a means for performing Figure 11 The method or any embodiment thereof.

[0127] The component may include at least one processor; and at least one memory storing instructions that, when executed by the at least one processor, cause the device to perform at least one of the following: Figure 11 The method or any embodiment thereof.

[0128] A computer program product (optionally stored on a non-transitory computer-readable medium) includes instructions that, when executed by at least one processor, cause the device or system to perform at least the following: Figure 11 The method or any embodiment thereof.

[0129] Such a device may include, for example Figures 1-3 The functional units disclosed in any one of the embodiments for implementing the embodiments.

[0130] This method, apparatus, and computer program product can offer various advantages. For example, it can save PUCCH resources and UE power consumption.

[0131] Now for reference Figure 12 describe Figure 11 Implementation examples and aspects of the method, Figure 12Time series plots of prediction start time, observation window, and prediction window for BFD prediction in MRSS scenarios are presented. Although Figure 12 The observation stop time and prediction start time synchronized with T0 are shown, but the prediction start times of the two RATs may be the same or different, and / or the prediction stop times of the two RATs may be the same or different. Alternatively or additionally, the lengths of the observation windows and / or prediction windows of the two RATs may be the same or different.

[0132] In one example, there may be a gap between the observation stop time and the prediction start time of one or both RATs, and / or a gap between the observation stop time of one RAT and the prediction start time of another RAT. This can be implemented, for example, in a manner such that the prediction of RAT1 begins at T0_RAT1, and the prediction of RAT2 begins at T0_RAT2. In one example, T0_RAT1 and T0_RAT2 are different, i.e., they are not synchronized at the same time step. To improve the effectiveness of prediction, a time difference threshold can be defined if one RAT starts prediction earlier. For example, the absolute value of the time difference can be less than the time difference threshold, i.e., abs(T0_RAT1-T0_RAT2) < time difference threshold (in ms). The time difference threshold can be configured by the network (e.g., RAT1 or RAT2). The UE can receive configurations including the time difference threshold, such as RRC messages including the threshold.

[0133] BFD Prediction Start Conditions

[0134] Various criteria can be used to trigger BFD prediction. In some embodiments, the BFD prediction start condition is related to the serving beam. In one embodiment, detecting the BFD prediction start condition includes monitoring the radio link quality (RLQ) of the serving beam. In some embodiments, the BFD prediction start condition includes a threshold for the RLQ of the serving beam, such as the out-of-sync threshold Q of the first RAT cell. out,LR (See Standard 1 below), and the threshold can be set to the out-of-sync threshold Q in the first RAT cell. out,LR The predetermined offset is reached (see Standard 2 below). Here, "loss of synchronization" refers to link loss of synchronization. In one embodiment, the BFD prediction start condition includes dedicated indication signaling received from the first RAT cell (see Standard 3 below). The triggering condition (i.e., the BFD prediction start condition) is preferably based on the serving RAT radio conditions. In one embodiment, the method includes receiving configuration for the BFD prediction start condition (from the network, such as the first RAT). Some example standards or BFD prediction start conditions are outlined below.

[0135] Standard 1: UE out of step (RLQ vs. Q) out,LR (Poor) or the UE receives the first BFI from the serving RAT.

[0136] Standard 2: The UE is about to experience a loss of synchronization (RLQ compared to Q). out,LR -Offset difference). This standard ensures that the UE can begin prediction before losing synchronization with the serving beam. In an example implementation, Q out,LR The offset has the same sign (usually negative, or positive). For example, Q out,LR The threshold can be -8dB, and the offset can be -2dB. Therefore, the resulting standard is -8dB - (-2dB) = -6dB. Thus, prediction can start 2dB earlier than in standard 1. Typically, the threshold can be set to be a certain offset above the out-of-sync threshold of the RAT cell. Therefore, the offset is in Q... out,LR Starting prediction before reaching (i.e., decreasing to) a specific value provides a margin. These examples apply to the various embodiments described herein, which may utilize Standard 2 or variations thereof as the starting condition for BFD prediction.

[0137] Standard 3: The UE receives dedicated indication signaling from the network, such as MAC CE beam / cell handover commands, DCI commands for TCI state activation / handover, or dedicated indication signaling for triggering prediction execution. In this case, "dedicated" means indication signaling specifically used to trigger prediction execution. The indication signaling may be dedicated solely to this purpose or may have other purposes.

[0138] Forecast time range:

[0139] The first observation window / observe window of the first RAT (T_observe_rat1, in ms): The UE monitors past RRM measurement samples, such as L1 / L3 measurement samples from the serving RAT within the defined observation window, ranging from T0 to T_observe_rat1 to T0. The choice of the observation window can be flexible, for example, 1000ms configured by the network.

[0140] The second observation window / observe window of the second RAT (T_observe_rat2, in ms): The UE monitors past RRM measurement samples, such as L1 / L3 measurement samples from the second RAT within the defined observation window, ranging from T0 to T_observe_rat2 and back to T0. The choice of the observation window can also be flexible, for example, 1000ms configured by the network. RRM measurement samples can indicate which beam is the strongest beam; therefore, the UE can determine the strongest beam based on the RRM measurement samples.

[0141] The first prediction window / prediction window of the first RAT (T_predict_rat1, in ms): Based on observation data from the first RAT, the UE predicts the serving beam BFD state within a future prediction window, ranging from T0 to T0+T_predict_rat1 or T0 to T1. This prediction window determines the time range of the predicted BFD state from the serving RAT. The selection of the prediction window is preferably adapted to the duration of the continuous BFI counter, for example, the BFD timer length multiplied by the maximum number of configured BFIs.

[0142] The second prediction window / prediction window of the second RAT (T_predict_rat2, in ms): Based on observation data from the second RAT, the UE predicts the BFD state of the strongest beam of the second RAT within a future prediction window, ranging from T0 to T0+T_predict_rat2 or T0 to T2. This prediction window determines the time range of the predicted BFD state from the second RAT.

[0143] It is worth noting that, compared to the first RAT, the second RAT can have different BFD configurations, thus allowing for more flexibility in configuring the observation window and prediction window in different ways.

[0144] In one embodiment, the method includes receiving (from a network, such as a first RAT) configuration for one or more of the following: the length of a first observation window, the length of a second observation window, the length of a first prediction window, and the length of a second prediction window.

[0145] Prediction results:

[0146] Direct Prediction: The UE uses probability values ​​to directly predict the BFD state (0 / 1) of the serving beam (first RAT) to indicate the confidence level (e.g., X%). Simultaneously, the UE uses probability values ​​to directly predict the BFD state (0 / 1) of the strongest beam (second RAT) to indicate the confidence level (e.g., X%). Finally, these two numbers are generated from the direct multi-class classification.

[0147] Indirect Prediction: In one embodiment, prediction includes: predicting a first time-series beam measurement for the serving beam of a first RAT cell within a first prediction window based on a first RRM measurement sample, and deriving a first BFD state based on the first time-series beam measurement; and / or predicting a second time-series beam measurement for the strongest beam of a second RAT cell within a second prediction window based on a second RRM measurement sample, and deriving a second BFD state based on the second time-series beam measurement. For example, the UE uses a precision metric such as Mean Absolute Error (MAE) (in dB) to predict the time series of beam-level measurements based on RS transmitted from the first RAT. The BFD state (0 / 1) is then derived based on these predicted measurements. Simultaneously, the UE predicts the time series of beam-level measurements based on RS transmitted from the second RAT, and then derives the BFD state (0 / 1). Finally, both figures are generated from indirect multi-class classification.

[0148] Prediction Report: The UE reports prediction results to the network based on the configured BFD prediction start criteria. Example steps for each of the above BFD criteria are provided below.

[0149] Standard 1: UEs may not report synchronization problems immediately, but may report them after the BFR process is complete.

[0150] Standard 2 or 3: The UE can report the prediction results to the network, and then the network can instruct the UE to perform an early BFR.

[0151] UE actions for early BFR:

[0152] The UE performs one or more early BFR actions based on the first BFD state and / or the second BFD state. For example, if the state of the serving beam indicates a beam failure, the UE should move to a different beam. If the state of the strongest beam indicates a beam failure, the UE should not move to that beam. The selection of which early BFR actions to perform can be performed by the UE and / or the network and can depend on various configurations. Multiple early BFR actions can be performed on the serving beam of the first RAT, another beam of the first RAT, the strongest beam of the second RAT, another beam of the second RAT, and / or a beam of another RAT. The beam selection can depend on the specific scenario.

[0153] Action Option 1a: In one embodiment, one or more early BFR actions include terminating an ongoing BFD procedure before reaching the maximum number of BFI attempts. In one embodiment, one or more early BFR actions may include initiating a Random Access Channel (RACH) procedure by transmitting a preamble to one or more candidate beams. The one or more candidate beams may include one or more beams of a second RAT cell. For example, the network instructs the UE to terminate the ongoing BFD procedure early before reaching the maximum number of BFI attempts and initiates a BFR RACH procedure by transmitting a preamble to a target candidate beam on the current RAT or a different RAT.

[0154] Action Option 1b: In one embodiment, one or more early BFR actions include performing a RACH-free handover to a candidate beam. The candidate beam can be a beam from a second RAT cell, such as the strongest beam. For example, when LTM is configured, the candidate cells include both the beams from the first RAT (serving) and the second RAT. The UE can then perform a RACH-free handover to restore the beam failure to a candidate beam from a different RAT.

[0155] Action Option 2: In one embodiment, one or more early BFR actions include adjusting the BFD timer for the first RAT and / or the second RAT. For example, the network provides the UE with a list of T_delta_rat1 values ​​(in milliseconds), which can be used to adjust the conventional BFD timer for the serving RAT. beamFailureDetectionTimer This allows for a faster response to predicted BFD events. Another list of T_delta_rat2 values ​​(in milliseconds) can be used to adjust the conventional BFD timer for the second RAT. The values ​​of T_delta_rat1 and T_delta_rat2 can be dynamically adjusted based on the prediction confidence level score (expressed as a percentage).

[0156] Action Option 3: In one embodiment, one or more early BFR actions include increasing uplink transmission power during the RACH process for the first RAT cell and / or the second RAT cell. For example, the network provides the UE with a list of P_delta_rat1 values ​​(in dB), which can be used to increase UL transmit power during the RACH process serving the RAT, thereby increasing the likelihood of a successful BFR process initiation. The UE can use P_delta_rat2 to increase UL transmit power from the second RAT to the candidate beam. The values ​​of P_delta_rat1 and P_delta_rat2 can be dynamically adjusted based on the prediction confidence level score (expressed as a percentage).

[0157] Preferably, only one early BFR action / action option is performed. However, in some embodiments, more than one action may be performed.

[0158] Example implementation details of two prediction methods, direct prediction (Example A) and indirect prediction (Example B), will now be described with reference to Figures 13 and 14. These two prediction techniques are compatible with and composable with all methods disclosed herein, including... Figure 11 , Figure 15 and Figure 17 The method in the middle.

[0159] Example A: Multi-class classification (direct prediction)

[0160] Direct prediction involves a multi-class classification model used to predict whether both the service RAT and the second RAT will experience BFD within an upcoming time window. Referring to Figure 13, the steps to implement direct prediction may include one or more of the following:

[0161] Initialization and RRC configuration reception: The UE initializes and receives RRC configuration parameters to enable beam fault detection and recovery for AI / ML.

[0162] Check the prediction start conditions: • Start prediction time: The UE assesses whether the current conditions meet the requirements for starting the prediction process. • Conditions: The UE monitors Radio Link Quality (RLQ) according to Standard 1 or Standard 2, or receives a trigger indication from the network according to Standard 3. If the conditions are met, data collection begins; otherwise, the system defaults to the traditional BFD and recovery procedure. • Alternatively, the network can use dedicated DCI signaling to trigger the UE to begin prediction.

[0163] Data collection: L1 or L3 beam-level measurements were collected from the serving RAT beam and the second RAT beam during observation windows T_observe_rat1 and T_observe_rat2, respectively.

[0164] Training module: • Data labeling: If a serving beam fault detection occurs within the prediction window (T_predict_rat1), the serving RAT BFD state is marked as true; otherwise, it is marked as false. Similarly, if a beam fault detection occurs within the second RAT within the prediction window (T_predict_rat2), the second RAT BFD state is marked as true; otherwise, it is marked as false. • Model Training: Train the ML model, for example, using labeled data for random forest or gradient boosting. Once completed, the model is ready for inference.

[0165] Reasoning module: • Predict beam failure: Use a trained model to predict whether beam failure detection (true / false) will occur in the first and second RATs. • Decision (BFD prediction result): Based on the prediction result, if BFD occurs in the serving RAT, the second RAT, or even both RATs, the UE will proceed to the early recovery phase. Otherwise, the UE will follow the traditional BFD and BFR procedures.

[0166] Recovery module: Based on the configured operation options: • Option 1a: Start RACH to the same or different RATs in advance. • Option 1b: Perform a RACH-free handover under LTM to achieve beam fault recovery to candidate beams of different RATs. • Option 2: Reduce the number of traditional beam fault detection timers for the same or different RATs. • Option 3: Increase RACH uplink transmission power for the same or different RATs.

[0167] Example B: Multi-class classification (indirect prediction)

[0168] Indirect prediction involves using a regression model to predict the beam measurements of both the serving beam (of the first RAT) and the strongest beam of the second RAT, and then processing the predictions to determine the BFD state. Referring to Figure 14, the steps for implementing direct prediction may include one or more of the following:

[0169] Initialization and RRC configuration reception: The UE initializes and receives RRC configuration parameters to enable beam fault detection and recovery for AI / ML.

[0170] Check the prediction start conditions: • Start prediction time: The UE assesses whether the current conditions meet the requirements for starting the prediction process. • Conditions: The UE monitors Radio Link Quality (RLQ) according to Standard 1 or Standard 2, or receives a trigger indication from the network according to Standard 3. If the conditions are met, data collection begins; otherwise, the system defaults to the traditional BFD and recovery procedure. • Alternatively, the network can use dedicated DCI signaling to trigger the UE to begin prediction.

[0171] Data collection: L1 or L3 beam-level measurements were collected from the serving RAT beam and the second RAT beam during observation windows T_observe_rat1 and T_observe_rat2, respectively.

[0172] Training module: • Training the model: Train a regression model to predict future measurements of the serving beam during prediction window T_predict_rat1, and future measurements of the strongest neighboring beam of the second RAT during prediction window T_predict_rat2. • Training complete: The trained model is ready for inference.

[0173] Reasoning module: • Predictive beam measurements: Uses a trained model to predict future beam measurements of the serving beam of the first RAT and the strongest neighboring beam of the second RAT.

[0174] Fault detection module: • Process Prediction: Evaluate the predictive measurement. If the SINR of the serving beam is below a predefined threshold, BFD is marked as true; otherwise, it is marked as false. Similarly, if the SINR of the strongest neighboring beam of the second RAT is below a predefined threshold, BFD is marked as true; otherwise, it is marked as false. • Decision (BFD Outcome): Based on the prediction outcome, if BFD occurs in the serving RAT, the second RAT, or even both RATs, the UE will proceed to the early recovery phase. Otherwise, the UE will follow the traditional BFD and BFR procedures.

[0175] Recovery module: Based on the configured operation options: • Option 1a: Start RACH to the same or different RATs in advance. • Option 1b: Perform a RACH-free handover under LTM to achieve beam fault recovery to candidate beams of different RATs. • Option 2: Reduce the number of traditional beam fault detection timers for the same or different RATs. • Option 3: Increase RACH uplink transmission power for the same or different RATs.

[0176] Signaling procedures for UE-side BFD prediction and early BFR (MRSS)

[0177] RRC Configuration Parameters (MRSS)

[0178] Similar to the single RAT scenario, some configuration parameters for the MRSS scenario can be summarized as follows. The UE can send and / or receive (multiple) configurations, including but not limited to one or more of the following. These configurations can be sent to and / or received from the network, such as serving / first RAT and / or neighboring / second RAT. These configurations are preferably sent / received by the UE as RRC messages.

[0179] The UE can transmit UE capability information to support direct or indirect BFD prediction for MRSS. In one embodiment, the method includes transmitting capability information to a first RAT cell indicating support for BFD prediction and / or early BFR actions.

[0180] The UE can receive predicted start conditions. Predicted start conditions can be determined by a threshold of the serving RAT and an optional offset:

[0181] Standard 1: If Standard 1 is used, the threshold is set to the out-of-step threshold Q. out,LR .

[0182] Standard 2: If Standard 2 is used, the threshold is also set to Q. out,LR However, an additional offset (in dB) can be configured. This offset ensures that the prediction is triggered before the UE actually loses synchronization.

[0183] The UE can receive configuration for predictive timing, which includes, but is not limited to, at least one of the following parameters. Predictive timing can be configured using, for example, two parameters of a RAT cell: • The first RAT's observation window (T_observe_rat1, in ms): The UE monitors past L1 / L3 measurement samples from the serving RAT. • The observation window of the second RAT (T_observe_rat2, in ms): The UE monitors past L1 / L3 measurement samples from the second RAT. • Prediction window of the first RAT (T_predict_rat1, in ms): This prediction window determines the time range for predicting the BFD state of the serving RAT. The prediction window should be chosen to at least accommodate the duration of the continuous BFI counter, for example, the BFD timer length multiplied by the maximum number of configured BFIs. • Prediction window of the second RAT (T_predict_rat2, in ms): Based on the observation data from the second RAT, the UE predicts the strongest beam of the second RAT within the future prediction window according to its BFD state.

[0184] It is worth noting that the second RAT can have a different BFD configuration compared to the first RAT, so the observation window and prediction window can be configured differently. Alternatively, the same configuration can be used for both RATs.

[0185] The UE can receive thresholds used for indirect BFD prediction, such as: • If the average predicted measurement within the prediction window is below this threshold, the indirect BFD prediction is considered true, indicating a potential fault. • If the average predicted measurement is still above the threshold, the indirect BFD prediction is considered false, indicating that there is no impending failure.

[0186] A single threshold or two separate thresholds can be used for different RATs.

[0187] The UE can receive UE actions for predictive BFD processing. UE actions can include, but are not limited to, at least one of the following: • Operation Option 1a: The network instructs the UE to terminate the ongoing BFD procedure before reaching the maximum number of BFI attempts, and initiates a BFR RACH procedure by transmitting a preamble to the target candidate beam on the current RAT or a different RAT. The network can send instructions based on the BFD prediction report after receiving it from the UE. • Operation Option 1b: When LTM is configured, candidate cells include both the first RAT (serving) and the second RAT's beams. The UE can then perform a RACH-free handover to restore a beam failure to a candidate beam from a different RAT. • Operation Option 2: The network provides the UE with a list of T_delta_rat1 values ​​(in milliseconds), which can be used to adjust the traditional BFD timer for the serving RAT. beamFailureDetectionTimer This allows for a faster response to predicted BFD events. Another list of T_delta_rat2 values ​​(in milliseconds) can be used to adjust the conventional BFD timer for the second RAT. The values ​​of T_delta_rat1 and T_delta_rat2 can be dynamically adjusted based on the prediction confidence level score (expressed as a percentage). • Operation Option 3: The network provides the UE with a list of P_delta_rat1 values ​​(in dB), which can be used to increase UL transmit power during the RACH process of the serving RAT, thereby increasing the likelihood of a successful BFR procedure initiation. The UE can use P_delta_rat2 to increase UL transmit power from the second RAT to the candidate beam. The values ​​of P_delta_rat1 and P_delta_rat2 can be dynamically adjusted based on the prediction confidence level score (expressed as a percentage).

[0188] The UE can receive UE report configuration and format. The configuration and / or format may include, but is not limited to, at least one of the following: • Reporting criteria: The reporting criteria are most likely to be event-triggered and preferably consistent with the configuration conditions that initiate the prediction. ○Standard 1: If this standard is configured, the UE may not report to the network immediately. Reporting is only allowed after the BFR procedure is completed. ○ Standard 2 or 3: If any of these standards are configured or applied, the UE can report the prediction results to the network. Based on this report, the network can instruct the UE to perform an early BFR. • Report Content: In one embodiment, the method includes sending a BFD prediction report to a first RAT cell and / or a second RAT cell, wherein the BFD prediction report includes a first BFD state and / or a second BFD state. For example, the content of the report depends on the prediction method used, which may be direct prediction or indirect prediction. ○ Direct prediction: The report includes the predicted BFD status (2 digits - the first digit indicates the serving beam, and the second digit indicates the strongest beam of different RATs) and the probability confidence level (e.g., X%). ○ Indirect Prediction: The report includes the predicted BFD state derived from time-series measurements of two RATs (2 digits – the first digit indicates the serving beam, and the second digit indicates the strongest beam of the different RATs) and the mean absolute error (MAE) in dB. ○ Traditional BFD monitoring status: The report also includes metrics such as the percentage of timers running, the number of consecutive BFIs, and other relevant indicators. • Report Format: In one embodiment, the BFD prediction report includes a bitmap, which comprises at least two bits. For example, the prediction results may be formatted as a bitmap (minimum 2 bits), similar to the format used for L1 SS-RSRP reports.

[0189] Now for reference Figure 15 A second method for beam fault detection and recovery is presented. This method provides UE-autonomous prediction and recovery steps within the MRSS framework. Figure 15 Some characteristics of the method and Figure 11 The method is the same and has been described in detail above. Generally, the methods described in this article... Figure 11 The features, assumptions, and embodiments of the method can be combined with Figure 15 In the method. Accordingly, Figure 15 The features and embodiments of the method can be combined with Figure 11 In the method.

[0190] This method is as follows Figure 15The flowchart illustrates the operation of a terminal device (such as a User Equipment (UE)). The method includes: 1500 detecting a beam fault detection (BFD) prediction start condition associated with a first Radio Access Technology (RAT) cell providing the serving beam (for the device); 1502 acquiring a first Radio Resource Management (RRM) measurement sample from the first RAT cell within a first observation window; 1504 acquiring a second RRM measurement sample from a second RAT cell within a second observation window; using one or more trained machine learning models to: predict a first BFD state for the serving beam within the first prediction window based on the first RRM measurement sample; and predict a second BFD state for the strongest beam of the second RAT cell within the second prediction window based on the second RRM measurement sample; triggering one or more early beam fault recovery (BFR) actions based on the first and / or second BFD states within 1508; and after performing one or more early BFR actions, sending a BFD prediction report to the first and / or second RAT cell within 1510, wherein the BFD prediction report includes the first and / or second BFD states.

[0191] An apparatus (such as a UE) includes a means for performing Figure 15 The method or any embodiment thereof.

[0192] The device may include at least one processor; and at least one memory storing instructions that, when executed by the at least one processor, cause the device to perform at least one of the following: Figure 15 The method or any embodiment thereof.

[0193] A computer program product (optionally stored on a non-transitory computer-readable medium) includes instructions that, when executed by at least one processor, cause the device or system to perform at least the following: Figure 15 The method or any embodiment thereof.

[0194] Such a device may include, for example Figures 1-3 The functional units disclosed in any one of the embodiments for implementing the embodiments.

[0195] Now refer to Figure 16 for description. Figure 15 Regarding the implementation and examples of the method, Figure 16 illustrates an example signaling process for the UE autonomous prediction and recovery steps under the MRSS framework. This process includes the following steps, some of which may be omitted depending on the specific implementation:

[0196] Step 0: The network (first RAT) sends an indication to the UE that it is in an MRSS scenario based on the SS / PBCH indication.

[0197] Step 1: Exchange of UE capability information between the UE and the network under the MRSS framework. • In one embodiment, the method includes sending capability information to a first RAT cell indicating support for BFD prediction and / or early BFR actions. The UE notifies the network of its capabilities to support BFD prediction and BFR. For example, this includes UE capabilities to perform specific actions related to BFD prediction and BFR, such as monitoring beam measurements, performing prediction methods (e.g., direct prediction versus indirect prediction), and performing early BFR actions.

[0198] Step 2: The network configures parameters related to BFD prediction and BFR for the UE via RRC signaling. • In one embodiment, the method includes receiving configuration for one or more of the following: BFD prediction start conditions, the length of a first observation window, the length of a second observation window, the length of a first prediction window, the length of a second prediction window, a prediction method to be used to predict a first BFD state and / or a second BFD state, and one or more early BFR actions. For example, this includes setting a threshold for triggering BFD prediction; defining the prediction method to be used; and specifying the actions the UE should take during early BFR. Details are summarized in the above discussion of RRC configuration parameters (MRSS). • For example, the configuration of BFD prediction start conditions may include a threshold for the radio link quality (RLQ) for the serving beam, such as the out-of-sync threshold Q for the first RAT cell. out,LR .

[0199] Step 3: The network (first RAT) performs a beam scan to identify potential target beams for the UE. This involves transmitting reference signals (RS) on different beams.

[0200] Step 4: The network (second RAT) performs a beam scan to identify potential target beams for the UE. This involves transmitting reference signals (RS) on different beams.

[0201] Step 5: The UE continuously monitors service measurements to verify the prediction start condition. In one embodiment, detecting the BFD prediction start condition includes monitoring the radio link quality (RLQ) of the serving beam. In one embodiment, the BFD prediction start condition includes a threshold for the RLQ of the serving beam, such as the out-of-sync threshold Q of the first RAT cell. out,LR In this case, Standard 1 is used to check if the Radio Link Quality (RLQ) is below a predefined threshold Q. out,LR This indicates that the UE has lost synchronization and triggered the BFD prediction process. • Alternatively or additionally, the BFD prediction start condition may include receiving a beam fault instance (BFI) from a first RAT cell (not shown in Figure 16). The BFD prediction start condition may include the BFI counter being equal to 1. The BFD prediction start condition may include detecting the BFI counter incrementing from 0 to 1. This corresponds to the UE receiving the first BFI from the serving RAT.

[0202] Step 6: During the observation windows T_observe_rat1 and T_observe_rat2, the UE collects RRM measurements, such as L1 or L3 beam-level measurements, from the serving RAT beam and the second RAT beam, respectively. Here, we use T_observe to refer to both T_observe_rat1 and T_observe_rat2 collectively.

[0203] Step 7: The UE performs a first RAT BFD prediction in the future prediction window T_predict_rat1 and a second RAT BFD prediction in the future prediction window T_predict_rat2 based on the collected measurements. Here, we use T_predict to represent both T_predict_rat1 and T_predict_rat2 collectively. There are two main implementation options: • Direct Prediction: The UE directly predicts the BFD state for the upcoming prediction windows T_predict_rat1 and T_predict_rat2 based on historical measurements of the input. As shown and described with respect to Figure 13, multi-class classification models such as random forest classifiers can be used to perform the prediction. • Indirect Prediction: The UE predicts time-series measurements for upcoming prediction windows T_predict_rat1 and T_predict_rat2 based on historical input measurements, and then passes the predictions through a post-processing unit to detect beam faults using network-configured thresholds. As shown and described with respect to Figure 14, the predictions can be performed using regression models such as CNN or LSTM.

[0204] Steps 8-9 (optional): The UE monitors the performance of its BFD prediction method. In one embodiment, the method includes determining prediction reliability scores for predicted first and second BFD states; and comparing the prediction reliability scores with a reliability threshold. The prediction reliability score reflects the reliability of the predicted BFD state and can be, for example, a prediction confidence level score or another classification metric, such as an F1 score. For example, this involves comparing the prediction confidence level score or other classification metrics (such as the F1 score) with a network configuration reliability threshold if necessary. In one embodiment, the method also includes triggering the execution of one or more early BFR actions based on the prediction reliability score meeting the reliability threshold. When the prediction result reaches the reliability threshold, the UE can initiate an early BFR action.

[0205] Step 10: When the prediction result is sufficiently reliable, the UE performs an early BFR action. In one embodiment, the method includes continuing prediction until the prediction reliability score reaches a reliability threshold. Different operational options exist for handling BFD predictions, and some relevant options are discussed in this paper. • This depends on the network configuration used by the UE to process the early BFR after the prediction of BFD. In this scenario, the UE can directly process BFR actions without further interaction with the network. The UE uses a first BFD state and / or a second BFD state to trigger and execute one or more early BFR actions. In one embodiment, the method includes triggering the execution of one or more early BFR actions independently of instructions from a first RAT cell or a second RAT cell. That is, even if the UE receives such instructions, it is configured to trigger one or more BFR actions independently, regardless of such instructions. See below for reference. Figure 17 In another scenario, as depicted in Figure 18, the UE (always) requires instructions from the network to perform early recovery.

[0206] Step 11: The UE performs an early BFR action to switch to the new beam.

[0207] Step 12: In one embodiment, the method includes sending a BFD prediction report to a first RAT cell and / or a second RAT cell after performing one or more early BFR actions. Therefore, after early BFR to a new serving beam, the UE sends a BFD prediction report to the network (first RAT and second RAT). Typically, the BFD prediction report may include a first BFD state and / or a second BFD state. For direct prediction, the BFD prediction report may include probability confidence levels associated with the first BFD state and / or the second BFD state. For indirect prediction, the BFD prediction report may include the mean absolute error used to predict the first BFD state and / or the second BFD state. As a detailed example, the report may include one or more of the following: • For direct prediction: The report includes the predicted BFD state (binary: 0 / 1) and the probability confidence level (e.g., X%). • For indirect forecasts: The report includes the predicted BFD state (binary: 0 / 1) derived from the time series measurement forecast, and the mean absolute error (MAE) in dB. • Traditional BFD monitoring status: The report also includes metrics such as the percentage of timers running, the number of consecutive BFIs, and other relevant indicators.

[0208] Now for reference Figure 17 A third method for beam fault detection and recovery is presented. This method provides UE prediction and network control recovery steps within the MRSS framework. Figure 17 Some characteristics of the method and Figure 11 The method is the same and has been described in detail above. Generally, the methods described in this article... Figure 11 The features, assumptions, and embodiments of the method can be combined with Figure 17 In the method. Accordingly, Figure 17 The features and embodiments of the method can be combined with Figure 11 In the method. Similarly. Figure 17 Some characteristics of the method and Figure 15 The method is the same and has been described in detail above. Generally, the methods described in this article... Figure 15 The features and embodiments of the method can be combined with Figure 17 In the method. Accordingly, Figure 17 The features and embodiments of the method can be combined with Figure 15 In the method.

[0209] This method is as follows Figure 17The flowchart illustrates the operation of a terminal device (such as a User Equipment (UE)). The method includes: 1700 detecting a beam fault detection (BFD) prediction start condition associated with a first Radio Access Technology (RAT) cell providing the serving beam (for the device); 1702 acquiring a first Radio Resource Management (RRM) measurement sample from the first RAT cell within a first observation window; 1704 acquiring a second RRM measurement sample from a second RAT cell within a second observation window; using one or more trained machine learning models to: predict a first BFD state for the serving beam within the first prediction window based on the first RRM measurement sample; and predict a second BFD state for the strongest beam of the second RAT cell within the second prediction window based on the second RRM measurement sample; sending a BFD prediction report to the first RAT cell, wherein the BFD prediction report includes the first and second BFD states; receiving an early beam fault recovery (BFR) instruction from the first RAT cell; and performing one or more early BFR actions based on the early BFR instruction.

[0210] An apparatus (such as a UE) includes a means for performing Figure 17 The method or any embodiment thereof.

[0211] The device may include at least one processor; and at least one memory storing instructions that, when executed by the at least one processor, cause the device to perform at least one of the following: Figure 17 The method or any embodiment thereof.

[0212] A computer program product (optionally stored on a non-transitory computer-readable medium) includes instructions that, when executed by at least one processor, cause the device or system to perform at least the following: Figure 17 The method or any embodiment thereof.

[0213] Such a device may include, for example Figures 1-3 The functional units disclosed in any one of the embodiments for implementing the embodiments.

[0214] Now refer to Figure 18 for description. Figure 17 Regarding the implementation and examples of the method, Figure 18 illustrates an example signaling procedure for UE prediction and network control recovery steps under the MRSS framework. This procedure includes the following steps, some of which may be omitted depending on the specific implementation:

[0215] Step 0: The network (first RAT) sends an indication to the UE that it is in an MRSS scenario based on the SS / PBCH indication.

[0216] Step 1: Exchange of UE capability information between the UE and the network. • In one embodiment, the method includes sending capability information to a first RAT cell indicating support for BFD prediction and / or early BFR actions. The UE notifies the network of its capabilities to support BFD prediction and BFR. This includes UE capabilities to perform specific actions related to BFD prediction and BFR, such as monitoring beam measurements, performing prediction methods (e.g., direct prediction versus indirect prediction), and performing early BFR actions.

[0217] Step 2: The network configures parameters related to BFD prediction and BFR for the UE via RRC signaling. • In one embodiment, the method includes receiving configuration for one or more of the following: BFD prediction start conditions, the length of a first observation window, the length of a second observation window, the length of a first prediction window, the length of a second prediction window, a prediction method to be used to predict a first BFD state and / or a second BFD state, and one or more early BFR actions. For example, this includes setting a threshold for triggering BFD prediction; defining the prediction method to be used; and specifying the actions the UE should take during early BFR. Details are summarized in the above discussion of RRC configuration parameters (MRSS).

[0218] Step 3: The network (first RAT) performs a beam scan to identify potential target beams for the UE. This involves transmitting reference signals (RS) on different beams.

[0219] Step 4: The network (second RAT) performs a beam scan to identify potential target beams for the UE. This involves transmitting reference signals (RS) on different beams.

[0220] Step 5: Predict or verify the starting conditions. • Step 5a: Apply Standard 3, where the network sends a dedicated signaling indication to the UE to initiate prediction. In one embodiment, the BFD prediction initiation condition includes a dedicated indication signaling received from a first RAT cell. Detecting the BFD prediction initiation condition may include receiving the dedicated indication signaling from the first RAT cell. For example, the dedicated indication signaling may include: a Media Access Control (MAC) control element (CE) beam / cell handover command; a Downlink Control Information (DCI) command for transmitting configuration indication TCI state activation / handover; and / or a dedicated indication signaling command for triggering prediction execution. • Step 5b: In one embodiment, detecting the BFD prediction start condition includes monitoring the radio link quality (RLQ) of the serving beam. In one embodiment, the BFD prediction start condition includes a threshold for the RLQ of the serving beam. The threshold may be set to a loss-of-synchronization threshold Q in the first RAT cell. out,LRThis is achieved by reaching a predetermined offset. For example, as shown in Figure 18, Standard 2 is used for measurements in UE continuous monitoring services to verify whether the radio link quality (RLQ) is below a predefined threshold Q. out,LR Subtract the offset value. This indicates that the UE triggers the BFD prediction process before losing synchronization. • Preferably, only one criterion from the above steps is applied.

[0221] Step 6: During the observation windows T_observe_rat1 and T_observe_rat2, the UE collects RRM measurements, such as L1 or L3 beam-level measurements, from the serving RAT beam and the second RAT beam, respectively. Here, we use T_observe to refer to both T_observe_rat1 and T_observe_rat2 collectively.

[0222] Step 7: The UE performs a first RAT BFD prediction in the future prediction window T_predict_rat1 and a second RAT BFD prediction in the future prediction window T_predict_rat2 based on the collected measurements. Here, we use T_predict to represent both T_predict_rat1 and T_predict_rat2 collectively. There are two main implementation options: • Direct Prediction: The UE directly predicts the BFD state for the upcoming prediction windows T_predict_rat1 and T_predict_rat2 based on historical measurements of the input. As shown and described with respect to Figure 13, multi-class classification models such as random forest classifiers can be used to perform the prediction. • Indirect Prediction: The UE predicts time-series measurements for upcoming prediction windows T_predict_rat1 and T_predict_rat2 based on historical input measurements, and then passes the predictions through a post-processing unit to detect beam faults using network-configured thresholds. As shown and described with respect to Figure 14, the predictions can be performed using regression models such as CNN or LSTM.

[0223] Steps 8-9 (optional): The UE monitors the performance of its BFD prediction method. In one embodiment, the method includes determining prediction reliability scores for predicted first and second BFD states; and comparing the prediction reliability scores with a reliability threshold. The prediction reliability score reflects the reliability of the predicted BFD state and can be, for example, a prediction confidence level score or another classification metric, such as an F1 score. For example, this involves comparing the prediction confidence level score or other classification metrics (such as the F1 score) with a network configuration reliability threshold if necessary. In one embodiment, the method also includes triggering the transmission of a BFD prediction report based on the prediction reliability score meeting the reliability threshold (see step 10 below). When the prediction result reaches the reliability threshold, the UE can send a report to the network and then initiate an early BFR action based on network instructions. In one embodiment, the method includes continuing prediction until the prediction reliability score reaches the reliability threshold. This allows the UE to wait before sending a report so that the prediction results included in the report are sufficiently reliable.

[0224] Step 10: The UE sends a BFD prediction report to the network to request an early BFR action and waits for network instructions. In one embodiment, the method includes waiting after sending the BFD prediction report until an early BFR instruction is received. In another embodiment, the method includes preventing the execution of an early BFR action until an early BFR instruction is received. This prevents the UE from acting independently during the waiting period, thereby facilitating network control. Typically, the report includes at least a first BFD state and a second BFD state. For direct prediction, the BFD prediction report may include probability confidence levels associated with the first BFD state and / or the second BFD state. For indirect prediction, the BFD prediction report may include the mean absolute error used to predict the first BFD state and / or the second BFD state. As a detailed example, the report may include one or more of the following: • For direct predictions: The report includes the predicted first RAT BFD state (binary: 0 / 1) and the probability confidence level (e.g., X%) and the predicted second RAT BFD state (binary: 0 / 1) and the probability confidence level (e.g., X%). • For indirect forecasts: The report includes the first RAT BFD state (binary: 0 / 1) and the second RAT BFD state (binary: 0 / 1) derived from the time series measurement forecast, as well as the mean absolute error (MAE) in dB. • Traditional BFD monitoring status: The report also includes metrics such as the percentage of timers running, the number of consecutive BFIs, and other relevant indicators from both the service RAT and the second RAT.

[0225] Step 11: The network (first RAT) receives the BFD prediction report and verifies the prediction results. This may involve comparing the predictions with other network data and assessing the reliability of the predictions.

[0226] Step 12: The UE receives an early BFR instruction from the first RAT cell. In one embodiment, the early BFR instruction includes one or more early BFR actions. For example, the network sends an early BFR indication to the UE with network-preferred actions for beam recovery. • L1 / L2 signaling, such as DCI indication, can be used to communicate network-preferred BFR actions to the UE in order to process predicted BFD.

[0227] Step 13: After receiving an indication signal from the network, the UE performs an early BFR action. When one or more early BFR actions are received in an early BFR instruction, the method may include performing one or more early BFR actions based on the early BFR instruction. The UE performs the action(s) based on the received instruction; since the network receives (multiple) predicted BFD states for generating the instruction, the early BFR actions performed by the UE are based on the predicted BFD states. Different operational options exist for handling BFD predictions, and some of these options are discussed herein.

[0228] Step 14: The UE performs an early BFR action to switch to the new beam.

[0229] In any of the methods and embodiments described herein, early BFR actions can be naturally omitted if no beam fault is detected. Early BFR actions can only be triggered and executed if a beam fault is detected based on (multiple) predicted BFD states.

[0230] Generally, various embodiments of the present invention can be implemented in hardware or dedicated circuitry or any combination thereof. Although various aspects of the invention may be illustrated and described as block diagrams or represented using some other graphical representation, it is well known, by way of non-limiting example, that the blocks, devices, systems, techniques or methods described herein can be implemented in hardware, software, firmware, dedicated circuitry or logic, general-purpose hardware or controllers or other computing devices or some combination thereof.

[0231] Embodiments of the present invention can be practiced in various components such as integrated circuit modules. The design of integrated circuits is largely a highly automated process. Complex and powerful software tools can be used to send logic-level designs to semiconductor circuit designs ready for etching and formation on semiconductor substrates.

[0232] Programs such as those offered by Synopsys in Mountain View, California, and CadenceDesign in San Jose, California, use well-established design rules and pre-stored libraries of design modules to automatically route conductors and position components on semiconductor chips. Once the semiconductor circuit design is complete, the final design in a standardized electronic format (e.g., Opus, GDSII, etc.) can be sent to a semiconductor manufacturing facility or "wafer fab" for fabrication.

[0233] The term “non-transient” as used in this article refers to a limitation on the medium itself (i.e., tangible, not signaling), rather than a limitation on the persistence of data storage (e.g., RAM vs. ROM).

[0234] The foregoing description provides a complete and informative description of exemplary embodiments of the invention through exemplary and non-limiting examples. However, when read in conjunction with the accompanying drawings and examples, various modifications and adaptations can be found by those skilled in the art in light of the foregoing description. Nevertheless, all such and similar modifications to the teachings of the invention will still fall within the scope of the invention.

[0235] For completeness, some abbreviations used in this article, along with their unshortened forms, are listed below. UE: User Equipment RSRP: Reference Signal Received Power HO: Switch BFD: Beam Fault Detection RLM: Wireless Link Monitoring BFI: Examples of Beam Failure BFR: Beam Fault Recovery SSB: Synchronization Signal Block CSI: Channel State Information AI / ML: Artificial Intelligence / Machine Learning CNN: Convolutional Neural Network LSTM: Long Short-Term Memory gNB: Next-generation NodeB RRC: Radio Resource Control MAC: Media Access Control PHY: Physical Layer DCI: Downlink Control Information RACH: Random Access FR1: Frequency range 1 FR2: Frequency range 2 RLF: Wireless Link Failure MRSS: Multi-RAT Spectrum Sharing

Claims

1. A device for communication, comprising: At least one processor; as well as At least one memory storing instructions, which, when executed by the at least one processor, cause the means to perform at least the following: Beam Fault Detection (BFD) Prediction Start Conditions Related to the First Radio Access Technology (RAT) Cell for Detecting and Providing Service Beams; Within the first observation window, acquire the first Radio Resource Management (RRM) measurement sample from the first RAT cell; Acquire the second RRM measurement sample from the second RAT cell within the second observation window; Use one or more trained machine learning models to: Based on the first RRM measurement sample, predict the first BFD state for the serving beam within the first prediction window; as well as Based on the second RRM measurement sample, predict the second BFD state of the strongest beam for the second RAT cell within the second prediction window; Based on the first BFD state and / or the second BFD state, trigger the execution of one or more Early Beam Fault Recovery (BFR) actions. as well as After performing one or more early BFR actions, a BFD prediction report is sent to the first RAT cell and / or the second RAT cell, wherein the BFD prediction report includes the first BFD status and / or the second BFD status.

2. The apparatus of claim 1, wherein the apparatus is further configured to perform: The execution of one or more early BFR actions is triggered independently of instructions from the first RAT cell or the second RAT cell; The detection of the BFD prediction start condition includes monitoring the radio link quality (RLQ) of the serving beam; The BFD prediction start condition includes a threshold for the RLQ used for the serving beam; and The threshold is set to the out-of-sync threshold Q of the first RAT cell. out,LR .

3. The apparatus according to any one of claims 1 to 2, wherein detecting the BFD prediction start condition includes receiving a beam fault instance (BFI) from the first RAT cell; The BFD prediction start condition includes the BFI counter being equal to 1; and The detection of the BFD prediction start condition includes detecting the incrementing of the BFI counter from 0 to 1.

4. The apparatus according to any one of claims 1 to 2, wherein the prediction is performed using a direct prediction method, and wherein the BFD prediction report includes a probability confidence level associated with the first BFD state and / or the second BFD state.

5. The apparatus according to any one of claims 1 to 2, wherein the prediction is performed using an indirect prediction method, and wherein the BFD prediction report includes the mean absolute error used in predicting the first BFD state and / or the second BFD state.

6. The apparatus according to any one of claims 1 to 2, wherein the apparatus is further caused to perform: Send capability information to the first RAT cell indicating support for BFD prediction and / or early BFR actions.

7. The apparatus according to any one of claims 1 to 2, wherein the apparatus is further caused to perform: Receive configurations for one or more of the following: The BFD prediction start condition The length of the first observation window, The length of the second observation window, The length of the first prediction window, The length of the second prediction window, A prediction method for predicting the first BFD state and / or the second BFD state, or The one or more early BFR actions.

8. The apparatus of claim 7, wherein the configuration for the BFD prediction start condition includes a threshold for the Radio Link Quality (RLQ) of the serving beam; and The threshold for the RLQ of the serving beam in the configuration used for the BFD prediction start condition is set to the out-of-step threshold Q of the first RAT cell. out,LR .

9. The apparatus according to any one of claims 1 to 2, wherein the apparatus is further caused to perform: Determine the prediction reliability score for the predicted first BFD state and the second BFD state; The predicted reliability score is compared with a reliability threshold; and Based on the predicted reliability score meeting the reliability threshold, trigger the execution of one or more early BFR actions; and Continue the prediction until the prediction reliability score meets the reliability threshold.

10. A method for communication, comprising: Beam Fault Detection (BFD) Prediction Start Conditions Related to the First Radio Access Technology (RAT) Cell for Detecting and Providing Service Beams; Within the first observation window, acquire the first Radio Resource Management (RRM) measurement sample from the first RAT cell; Acquire the second RRM measurement sample from the second RAT cell within the second observation window; Use one or more trained machine learning models to: Based on the first RRM measurement sample, predict the first BFD state for the serving beam within the first prediction window; as well as Based on the second RRM measurement sample, predict the second BFD state of the strongest beam for the second RAT cell within the second prediction window; Based on the first BFD state and / or the second BFD state, trigger the execution of one or more Early Beam Fault Recovery (BFR) actions. as well as After performing one or more early BFR actions, a BFD prediction report is sent to the first RAT cell and / or the second RAT cell, wherein the BFD prediction report includes the first BFD status and / or the second BFD status.

Images