Radio resource management using machine learning
A neural network-based RRM scheme for UE dynamically adjusts RRM actions using sensor and radio measurements, addressing inefficiencies and power consumption issues in conventional static configurations.
Patent Information
- Application Number
- CN202380038578.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Priority Date
- 2022-05-10
- Filing Date
- 2023-05-03
- Publication Date
- 2024-12-03
AI Technical Summary
Conventional radio resource management (RRM) configurations for user equipment (UE) in wireless networks are static or algorithmic, leading to inefficient UE utilization and excessive power consumption, requiring significant design and implementation effort.
Implementing a neural network-based RRM scheme that fuses sensor data and radio measurements to dynamically determine RRM actions, adapting to the UE's current context and environment, reducing unnecessary operations and optimizing power usage.
Enhances RRM efficiency by adapting to the UE's current circumstances, reducing power consumption and eliminating unnecessary operations, while minimizing design and implementation efforts.
Smart Images

Figure CN119072944A_ABST
Abstract
Description
Background Art
[0001] Cellular networks and other wireless networks typically employ radio resource management (RRM) to manage network capacity issues at a large scale level (e.g., multi-user or multi-cell level) rather than addressing point-to-point network capacity issues. Therefore, RRM utilizes a wide variety of techniques to provide efficient overall network throughput while seeking efficient power consumption on various networking components. These techniques may include, for example, techniques for power control, scheduling, cell search, cell reselection, handover, radio link or connection monitoring, connection establishment / reestablishment, co-interference management, etc. User equipment (UE) in a wireless network typically plays an important role in RRM, such as by collecting various radio measurements and other system observations to report to the network for implementing network-side RRM actions, and implementing certain procedures based on these measurements. In a conventional approach, a serving base station (BS) or other components of the network infrastructure directs the UE to adopt a static or algorithmic configuration for the UE's role in the entire RRM process, such as by configuring the UE to use static scheduling to perform intra-frequency, inter-frequency, or cross-RAT (radio access technology) scanning of a serving cell or neighboring cells, or by specifying a specific algorithm or fixed threshold set for the UE to use in making a conditional handover decision (CHO) for switching between cells. This static approach to configuring the UE's role in RRM generally fails to take into account the UE's special circumstances and may therefore result in non-optimal RRM behavior at the UE. This generally results in inefficiencies in obtaining RRM-related information at the UE and may therefore affect the overall RRM decision process that utilizes such information. Non-optimal RRM configuration of the UE may also result in unnecessary power consumption at the UE as the UE performs various RRM actions that are less relevant or less timely for the overall efficiency of the RRM process. BRIEF DESCRIPTION OF THE DRAWINGS
[0002] By referring to the accompanying drawings, the present disclosure will be better understood, and its numerous features and advantages will become apparent to those skilled in the art.The use of the same reference numbers in different drawings indicates similar or identical items.
[0003] Figure 1 is a diagram illustrating an example wireless system employing neural network-based RRM management at a wireless device in accordance with some embodiments.
[0004] Figure 2 is a diagram illustrating an example configuration of a wireless device implementing a neural network for UE-side RRM management in accordance with some embodiments.
[0005] Figure 3 is a diagram illustrating an example configuration of infrastructure components that implement a UE neural network selection and update scheme in accordance with some embodiments.
[0006] Figure 4 is a diagram illustrating a machine learning module that employs a neural network for a neural network-based sensor and transceiver fusion scheme in accordance with some embodiments.
[0007] Figure 5 is a diagram illustrating an example operating environment for an RRM scheme at a wireless device according to some embodiments.
[0008] Figure 6 is a diagram illustrating an example operating environment for an RRM scheme at an infrastructure component according to some embodiments.
[0009] Figure 7 and Figure 8 Together, flow charts are shown of example methods for configuring and implementing an RRM neural network at a wireless device in accordance with some embodiments.
[0010] Fig. 9 It is shown Figure 7 and Figure 8 A ladder diagram of an example of the method.
[0011] Fig.10 is a diagram showing a method according to some embodiments Figure 7 and Figure 8 Flowchart of an alternative implementation of the method.
[0012] Fig.11 is a flow chart illustrating an example method of configuring and implementing a conditional handover (CHO) neural network at a wireless device for managing CHO decision making at the wireless device in accordance with some embodiments.
[0013] Fig.12 is a diagram showing a method according to some embodiments Fig.11 A ladder diagram of an example of the method. DETAILED DESCRIPTION
[0014] Static or algorithmic configuration of UEs for RRM operations, as found in many conventional wireless systems, may result in inefficient UE utilization of the entire RRM process at the expense of excessive power consumption at the UE. Further, configuring the UE to employ these static RRM configurations typically requires considerable design, testing, and implementation effort. Figures 1 to 12The static or algorithmic approach to UE-side RRM operations can be replaced or supplemented by a neural network (NN)-based approach that is used to fuse sensor data from available sensors of the UE with radio measurements taken by the UE to derive one or more RRM actions to be taken by the UE. Such actions may include, for example, configuring the frequency or type of radio measurements to be performed, determining certain conditional RRM actions (such as conditional handover (CHO) decisions or conditional primary cell changes (CPC)), performing specific types of measurements or other RRM actions, etc.
[0015] In at least one embodiment, a base station (BS) (or other infrastructure component) may have access to a set of neural network architecture configurations that have been trained using various training data sets reflecting different sensor capabilities, different sensor data, different RRM-related measurements taken by a UE or other wireless device, etc. During or after a wireless connection is established between the BS and the UE, the UE provides the BS with a representation of its sensor capabilities, or in particular, sensor capabilities related to sensor types used to train the neural network architecture configurations accessible to the BS. The BS then selects a neural network architecture configuration based on the indicated sensor capabilities of the UE and directs the UE to implement the selected neural network architecture configuration at the UE's RRM neural network (e.g., a deep neural network (DNN)).
[0016] Where the UE is so configured, the UE captures a series of one or more sensor measurements from a sensor set and supplies them to the RRM neural network as a sensor data set. Similarly, the UE captures a series of one or more radio measurements (also referred to as radio parameters or radio metrics), such as signal power measurements, of one or both of a serving cell or one or more target or neighboring cells via a radio interface distinct from the sensor set and supplies them to the RRM neural network as a radio measurement set. The RRM neural network then uses these inputs to generate outputs representing one or more RRM actions to be taken by the UE in response to these input data sets. The RRM actions specified by the outputs of the RRM neural network may include directing the taking of specific RRM actions (such as taking specified RRM-related measurements), or directing the implementation of a CHO decision (e.g., initiating a handover to a target cell, or conversely, suspending any handover), or directing the avoidance of specific RRM actions. The RRM action may additionally or alternatively include configuring some aspect of the RRM action, such as configuring the frequency of performing a specific RRM action (e.g., the frequency of checking for data when in a connected state, or the frequency of evaluating cell reselection when in an idle state), configuring parameters for the RRM action (e.g., specifying a specific frequency band to be used for inter-frequency measurements), etc.
[0017] Further, the UE may report one or both of the sensor data or radio measurements used by the RRM neural network in determining the resulting RRM actions to the BS, which may then use the feedback to retrain a model or copy of the RRM neural network and provide an updated version of the RRM neural network to the UE for subsequent use. Alternatively, the UE may utilize the same data and other information to retrain or modify its local copy of the RRM neural network.
[0018] The combination or fusion of the UE's recent local sensor data with recent radio measurements by the RRM neural network facilitates local RRM processes employed by the UE that are more easily adapted to the current transmission environment or other current circumstances of the UE than would be provided via a static / algorithmic RRM configuration of the UE. For example, because of the neural network architecture implemented by training, sensor data indicating the proximity of a body, building, or other interference source can trigger the RRM neural network to prescribe an RRM action that reduces the frequency of certain measurements made using the UE's millimeter wave (mmW) antenna (thereby saving power); or can trigger the RRM neural network to adopt a lower threshold for a signal power parameter measured via the mmW antenna before triggering CHO, and thereby potentially eliminate unnecessary handover processes. In addition, by utilizing a neural network configured with an architecture trained based on sensor data consistent with the UE's indicated sensor capabilities, the UE can effectively implement various UE-side RRM actions without the extensive design, testing, and implementation effort that would be required to implement conventional static / algorithmic RRM configurations.
[0019] The systems and techniques for neural network-based RRM detailed herein utilize coordination between infrastructure components of a wireless network and wireless devices. For ease of illustration, these systems and techniques are described in the example context of a cellular network, in which a base station (BS) acts as the aforementioned infrastructure component and a UE acts as the aforementioned wireless device. However, these systems and techniques are not limited to this example implementation. For example, in the same cellular context, the infrastructure component may be a server or other component separate from the BS, or may involve multiple infrastructure components, such as collaborating servers and BSs. As another example, in a wireless local area network (WLAN) implementation, the aforementioned infrastructure component may be a wireless access point (AP) or other component "upstream" of the aforementioned wireless device connected to the wireless AP.
[0020] Figure 1An example wireless communication network 100 employing a neural network-based RRM scheme in accordance with some embodiments is shown. In the depicted example, the wireless communication network 100 is a cellular network that includes a network infrastructure 102 that is wirelessly connected to one or more wireless devices, such as UE 104. The network infrastructure 102 includes a core network 106 that is coupled to one or more wide area networks (WANs) 108 or other packet data networks (PDNs) such as the Internet. The core network 106 is further connected to one or more BSs 110, such as BS 110-1 and BS 110-2. Each BS 110 supports wireless communication with one or more wireless devices, such as UE 104, via radio frequency (RF) signaling using one or more available RATs as specified by one or more communication protocols or standards. As such, each BS 110 operates as a wireless interface between one or more wireless devices and various networks and services provided by the network infrastructure 102, such as packet switched (PS) data services, circuit switched (CS) services, etc. Conventionally, the communication of signaling from the BS 110 to the UE 104 is referred to as the "downlink" or "DL," while the communication of signaling from the UE 104 to the BS 110 is referred to as the "uplink" or "UL."
[0021] Each BS 110 may employ any of a variety of RATs, such as operating as a NodeB (or base transceiver station (BTS)) of a Universal Mobile Telecommunications System (UMTS) RAT (also referred to as "3G"), operating as an enhanced NodeB (eNodeB) of a 3rd Generation Partnership Project (3GPP) Long Term Evolution (LTE) RAT, operating as a 5G NodeB ("gNB") of a 3GPP fifth generation (5G) New Radio (NR) RAT, etc. UE 104 may in turn represent any of a variety of electronic devices operable to communicate with BS 110 via a suitable RAT, including, for example, a mobile cellular telephone, a cellular-enabled tablet or laptop computer, a desktop computer, a cellular-enabled video game system, a server, a cellular-enabled home appliance, a cellular-enabled automotive communication system, a cellular-enabled smart watch or other wearable device, etc.
[0022] For the following purposes, Figure 11, BS 110-1 is currently wirelessly connected to UE 104 and providing network services via the resulting wireless connection, and is therefore referred to herein as the "serving" BS 110-1 (also referred to in the art as a "master" BS). In this example, BS 110-2 is not currently providing network services to UE 104, but is available for switching and beginning to provide network services, and is therefore referred to as the "target" BS 110-1 (also referred to in the art as a "neighboring" BS or a "secondary" BS).
[0023] although Figure 1 The example wireless communication network 100 of 100 depicts only a single UE 104 and two BSs 110 for ease of illustration, but in a real-world implementation, such a wireless system would have numerous BSs and UEs 104 operating in adjacent geographic areas and thus provide many opportunities for RF interference and contention for shared RF resources or network resources. Therefore, in at least one embodiment, the wireless communication network 100 utilizes an RRM scheme to manage network capacity and network resource contention at a large scale. The RRM scheme involves the UE 104 in both taking various measurements that can be used for various RRM processes and performing some or all of some of these RRM processes. As described above, conventional RRM schemes rely on algorithms or static approaches to utilize UEs for RRM-related measurements and operations.
[0024] In contrast, in at least one embodiment, the wireless communication network 100 utilizes a neural network-based RRM scheme, wherein the UE 104 is configured to utilize a trained RRM neural network (NN) 112 to adaptively and non-statically implement RRM actions based on information about the current context of the UE 104. In particular, in at least one embodiment, the RRM NN 112 receives as separate inputs local sensor data 114 and radio measurement data 116, and provides outputs indicative of at least one RRM action 118 from at least these inputs.
[0025] Local sensor data 114 includes data from a set of sensors of UE 104 (see Figure 2) and may be provided as a single point in time sample of sensor data / sensor state from the involved sensors, or as a time series of samples of sensor data from the involved sensors within a fixed or variable sliding time window. It should be understood that the information captured by the sensors in the sensor set of UE 104 may reflect the present operating state of UE 104, both with reference to the physical local RF transmission environment of UE 104 and with reference to the internal operating state of UE 104. For example, present conditions involving UE 104 (and the absence of such conditions) that may affect RF signaling between UE 104 and serving BS 110-1 may be detectable from, or otherwise represented in, sensor data generated by, for example, object detection sensors (such as radar, lidar, or imagers (e.g., imaging cameras)) that generate sensor data reflecting the presence or absence of interfering objects in a line of sight (LOS) propagation path between serving BS 110-1 and UE 104. Similarly, positioning data (such as from a global positioning system (GPS) sensor, a gyroscope, an accelerometer, or a camera-based visual odometry sensor system) locates one of the position and / or motion of the UE 104 relative to, for example, the serving BS 110-1, and thus may represent the current RF signal propagation environment. As another example, a light sensor, an image sensor, or a touch sensor may provide sensor data indicating the posture of the UE 104 relative to the user's body, and thus serve as an indication of the likely current RF signal propagation environment of the UE 104. With respect to the current internal operating environment of the UE 104, a battery power sensor may indicate the amount of battery power remaining, and thus the ability of the UE to continue to perform various tasks without affecting the overall operation of the UE 104, while a thermal sensor of the UE 104 may indicate whether the UE 104 is approaching a thermal limit, and thus the extent to which the UE 104 may continue to perform various RRM actions that may contribute to the thermal output of the UE 104. Thus, inputting local sensor data 114 into the trained RRM NN 112 may facilitate the decision-making process of the RRMNN 112 such that the resulting input is adapted to reflect the current operating environment of the UE 104.
[0026] For radio measurement data 116, the information may also be provided as a single point in time measurement sample, a time series of measurement samples within a fixed or variable sliding window, or a combination thereof. The radio measurement data 116 reflects measurements of various parameters or metrics of either or both of the received RF signaling or the transmitted RF signaling. Examples of such radio measurements may include reference signal received power (RSRP) measurements, reference signal received quality (RSRP) measurements, signal-to-noise ratio (SNR) measurements, signal-to-noise plus interference ratio (SNIR), signal-to-interference plus noise ratio (SINR) measurements, received signal strength indicator (RSSI) measurements, reference signal received quality (RSRQ) measurements, interference-to-thermal (IoT) ratio measurements, and the like.
[0027] As described above, the output of the RRM NN 112 in response to the input of the local sensor data 114 and the radio measurement data 116 represents one or more RRM actions 118 to be performed by the UE 104. The RRM actions 118 may belong to at least one of the following categories: (1) directing the taking or avoiding of an RRM process at the UE 104; (2) directing the modification or setting of parameters of an RRM process to be performed at the UE; or (3) determining the decision of a conditional RRM process of the UE 104. The first category of RRM actions represents RRM actions 118 to which the UE 104 responds by performing an RRM process or avoiding an RRM process that the UE 104 would otherwise perform. For example, based on the local sensor data 114 indicating that the UE 104 is stationary and in a relatively signal-free position, the RRMNN 112 may output an RRM action 118 that directs the UE 104 to perform a one-time RRM measurement (e.g., a scan of a particular frequency band) and report the results back to the serving BS 110-1. Conversely, based on local sensor data 114 indicating that UE 104 is in very fast motion or indicating that UE 104 is indoors and surrounded by a large number of RF blocking objects, RRM NN 112 may output an RRM action 118 that directs UE 104 to skip the next scheduled RRM measurement to reflect that attempting a scheduled RRM measurement may be futile in the UE's current operating environment (and thereby potentially saving power and computing resources of the UE).
[0028] The second category of RRM actions represents an RRM action 118 to which the UE 104 responds by modifying an RRM procedure that the UE 104 has been scheduled or directed to take. For example, the UE 104 may be configured to perform certain RRM measurements according to a certain timing, and the RRM action 118 may represent a modification of that timing. To illustrate, the 3GPP Fifth Generation New Radio (5G NR) specification specifies a comparison of the relative signal strength of a serving cell (e.g., as provided by a serving BS 110-1) with a neighboring cell (e.g., as provided by a target BS 110-2) using a cell signal measurement procedure using a synchronization signal (SS) / physical broadcast channel (PBCH) block or SSB, wherein the timing or periodicity of each consecutive measurement using the SSB is controlled via an SSB-based RRM measurement timing configuration (SMTC) window. Thus, one or both of the input local sensor data 114 or the current radio measurement data 116 may trigger the trained RRM NN 112 to provide an output representing an RRM action 118 that directs the UE 104 to adjust the SMTC window, thereby increasing or decreasing (depending on the adjustment) the periodicity of SSB-based cell measurements of the serving cell and neighboring cells.
[0029] The third category of RRM actions represents RRM actions 118 that represent decisions made for one or more conditional RRM procedures that can be performed by the UE 104. For example, the 3GPP 5G NR Release 16 specification specifies a conditional handover (CHO) RRM procedure, in which a handover command is sent to the UE along with one or more conditions that are monitored by the UE in association with the handover command. Instead of performing the handover immediately, the handover command is stored and the UE monitors the specified one or more conditions. When the monitored conditions are met, the UE initiates the previously received handover. Therefore, for this example, the RRM action 118 can mimic the CHO procedure without the need to statically or algorithmically define the specific handover conditions. In contrast, the RRM NN 112 is trained using a training data set having similar training sensor data and similar training radio measurement data to make CHO-like decisions, and thus, the RRM actions 118 outputted in the field by the RRM NN 112 may include CHO decisions to initiate a handover or to avoid initiating a handover based on the current network and UE context as reflected in the input local sensor data 114 and radio measurement data 116. Other such conditional RRM process decisions that may be designated as RRM actions 118 may include, for example, a conditional primary cell change (CPC), in which the UE performs a change of a primary cell (Pcell) when one or more specified conditions are met.
[0030] The RRM action 118 may also be a combination of two or more of the three categories identified above; that is, the RRM action 118 may be a hybrid RRM action. To illustrate, the output of the RRM NN 112 may represent an RRM action 118 that activates a previously deactivated RRM measurement process and configures one or more parameters of the RRM measurement process (e.g., activates inter-RAT scanning and configures frequencies to scan).
[0031] The RRM NN 112 may also receive other inputs that help generate outputs representing RRM actions 118. For example, the 5G NR specification specifies that certain RRM measurements are specific to a combination of the UE's RRC state (IDLE, INACTIVE, CONNECTED) and whether the UE's 5G NRRAT is in standalone (SA) mode or non-standalone (NSA) mode. For example, if the UE is in the RRC IDLE state and in SA mode, the UE is allowed to perform cell selection or cell reselection, but if the UE is in NSA mode while in the RRC IDLE state, the 5G NR specification specifies that the UE is expected to abandon cell selection or cell reselection. Similarly, if the UE is in the RRC CONNECTED state, the UE can perform random access to the primary cell regardless of whether the 5G NR RAT is in SA mode or NSA mode, but the 5G NR specification specifies that the UE is expected to perform a handover only if it is in SA mode while in that RRC state. Thus, the UE 104 may provide operating state data 120 as input to the RRM NN 112, wherein the operating state data 120 represents specific operating state parameters of the UE 104 and / or the RAT of the UE 104, such as, for example, the RRC state of the UE 104 and the SA / NSA mode of the UE's 5G NR RAT. In this example, the RRM NN 112 may be trained using the training data to avoid providing outputs that trigger RRM actions 118 representing RRM procedures that would be inconsistent with RRC state and SA / NSA mode restrictions regarding certain RRM measurements.
[0032] In order for UE 104 to perform RRM actions using RRM NN 112, RRM NN 112 is first configured for UE 104. Further, in some embodiments, the infrastructure configuration of RRM NN 112 may be updated by local update or remote update via UE 104 reporting relevant information. Figure 1121 shows a general overview of the process of initial configuration and updating of a UE 104 using an RRM NN 112. In at least one embodiment, the RRM NN 112 employed by the UE 104 has been trained using training data similar to data inputs expected to be provided by the UE 104. As explained above, the data inputs to the RRM NN 112 include local sensor data 114 generated by a particular set of sensors available to the UE 104, and radio measurement data 116 representing particular types of radio measurements that may be taken by the UE 104. As such, it is advantageous to select a trained NN architecture configuration for the RRM NN 112 to be employed by the UE 104 that has been trained using training data from similar sensors and for similar types of radio measurements. For example, employing a NN architecture configuration for the RRM NN 112 that was trained using training data that was heavily influenced by radar sensor data will generally provide inferior results at UEs that do not have radar sensors. To this end, the serving BS 110-1 stores or otherwise accesses a collection of various NN architecture configurations, each of which is trained according to a particular sensor configuration and / or radio measurement configuration. During the initialization process between the serving BS 110-1 and the UE 104, the UE 104 provides a representation of its capabilities (UE capability message 122) to the serving BS 110-1. The UE capability message 122 includes representations of various capabilities of the UE 104, including one or both of a representation of the sensor capabilities of the UE 104 or a representation of the radio measurement capabilities of the UE 104. The representation of the sensor capabilities of the UE 104 may include, for example, a representation of the types of sensors included in the sensor set of the UE 104, the capabilities of each of some or all of these sensors, or other parameters, and the like. For example, during the attachment process, the serving BS 110-1 may send a UE capability query radio resource control (RRC) message to the UE 104, to which the UE 104 responds with a UECapabilitiesInformation RRC message, wherein one or more fields of the UECapabilitiesInformation RRC message include data or other information representing the type, number, and parameters of sensors of the sensor set of the UE 104.
[0033] like Figure 1As shown, the serving BS 110-1 (or another infrastructure component, such as a server of the core network 106) then uses the sensor and / or radio measurement capabilities of the UE 104 as indicated in the UE capability message 122 to select an RRM NN architecture configuration 124 for the UE 104 that is appropriate for the indicated sensor capabilities and / or radio measurement capabilities. The selection may be performed using any of a variety of techniques, such as using one or more lookup tables (LUTs) indexed by the indicated capabilities, a weighted ranking comparing the indicated capabilities to the sensor types / parameters and / or radio measurement types used to train the corresponding candidate RRM NN architecture configurations, and the like. The serving BS 110-1 then signals the UE 104 to adopt the selected RRM NN architecture configuration 124. For example, in some embodiments, the serving BS 110-1 may send one or more messages to the UE 104 having data representing the actual RRM NN architecture configuration 124. In other embodiments, UE 104 may have been provisioned with a set of stored RRM NN architecture configurations, and serving BS 110-1 may then send a message including an index or other identifier of a particular RRM NN architecture configuration in the stored set to be used by UE 104.
[0034] In response to the identification or provision of the RRM NN architecture configuration 124, the UE 104 configures the RRM NN 112 to utilize the identified RRM NN architecture configuration 124, and begins operation of the RRM NN 112 as so configured. In the course of operation, and consistent with many RRM schemes, the UE 104 may provide various RRM reports 126 back to the serving BS 110-1. The RRM reports 126 may include reports of RRM measurements made by the UE 104, including those directed or configured by the RRM NN 112. The RRM reports 126 may further include reports of decisions or configurations reflected in the RRM actions 118 output by the RRM NN 112. For example, the UE 104 may report via the RRM reports 126 when it has changed the frequency at which the UE performs inter-RAT or intra-RAT scans as a result of the RRM actions 118 output by the RRM NN 112. The serving BS 110 - 1 may then take local action in response to the RRM report 126 , or forward the RRM report 126 to the core network 106 for further consideration by the network infrastructure 102 .
[0035] In some embodiments, the RRM NN architecture configuration employed by the RRM NN 112 may be dynamically updated based on usage. In some embodiments, NN updates to the RRM NN 112 are provided by the serving BS 110-1 (or other infrastructure component). In this approach, the UE 104 may provide a copy 128 of the local sensor data 114 and / or radio measurement data 116 to the serving BS 110-1, and the serving BS 110-1 or other component may use the data along with, for example, the RRM reports 126 to retrain the NN architecture configuration 124 utilized by the UE 104 to generate an updated NN architecture configuration, and provide the updated NN architecture configuration to the UE 104 as a NN update 130 for implementation. In other embodiments, the NN updates are performed by the UE 104 itself. For example, the UE 104 may be configured to periodically provide a copy or "snapshot" of the current state of the NN architecture configuration of the RRM NN 112 to the serving BS 110-1 as a NN update 130. The serving BS 110 - 1 may then update its local or accessible copy of the RRM NN architecture configuration 124 to reflect the provisioned NN update 130 , and may further redistribute the updated NN architecture configuration to other UEs employing the same NN architecture configuration 124 .
[0036] Figure 2 An example hardware configuration of a UE 104 (as a representative wireless) according to some embodiments is shown. It should be noted that the depicted hardware configuration represents the processing components and communication components most directly related to the neural network-based processes described herein, and omits certain components that are well known to be frequently implemented in such electronic devices, such as displays, non-sensor peripherals, external power supplies, etc.
[0037] In the depicted configuration, the UE 104 includes an RF front end 202 having one or more antennas 203 and a radio interface 204 having one or more modems to support one or more RATs. The RF front end 202 actually operates as a physical (PHY) transceiver interface to conduct and process signaling between one or more processors 206 of the UE 104 and the antenna 203, thereby facilitating various types of wireless communications. The antenna 203 can be arranged as one or more arrays of multiple antennas configured similarly or differently to each other, and can be tuned to one or more frequency bands associated with the corresponding RAT. The one or more processors 206 may include, for example, one or more central processing units (CPUs), graphics processing units (GPUs), tensor processing units (TPUs), or other application specific integrated circuits (ASICs), etc. For illustration, the processor 206 may include an application processor (AP) utilized by the UE 104 to execute an operating system and various user-level software applications, as well as one or more processors utilized by the modem or a baseband processor of the radio interface 204.
[0038] The UE 104 further includes one or more computer-readable media 208, which include any of a variety of media used by electronic devices to store data and / or executable instructions, such as random access memory (RAM), read-only memory (ROM), cache, flash memory, solid-state drive (SSD) or other mass storage devices, etc. For ease of description and brevity, the computer-readable medium 208 is referred to herein as "memory 208", given that system memory or other memory is often used to store data and instructions for execution by the processor 206, but it should be understood that references to "memory 208" should also apply to other types of storage media unless otherwise specified.
[0039] In at least one embodiment, the UE 104 further includes a plurality of sensors distinct from the radio interface 204 and referred to herein as a sensor set 210, at least some of which are used in the neural network-based scheme described herein. Typically, the sensors of the sensor set 210 include those sensors that sense certain aspects of the external RF environment of the UE 104 (i.e., sensors that are likely to sense parameters that have at least some effect on or reflect the RF propagation path or RF transmit / receive performance of the UE 104), sensors that sense certain aspects of the current operating state of the UE 104 (such as battery status, thermal status, operating mode, screen status, etc.). As such, the sensors of the sensor set 210 may include one or more sensors 212 for object detection, such as radar sensors, lidar sensors, imaging sensors, structured light-based depth sensors, proximity sensors, etc. The sensor set 210 may also include one or more sensors 214 for determining the position, posture, or speed / velocity of the UE 104, such as a satellite positioning sensor (such as a GPS sensor), a global navigation satellite system (GNSS) sensor, an inertial measurement unit (IMU) sensor, a visual odometry sensor, an accelerometer, a gyroscope, an odometer, an altimeter, a tilt sensor or other inclinometer, an ultra-wideband (UWB) based sensor, etc. Other examples of the types of sensors of the sensor set 210 may include a sensor 216 for determining the current operating state of the UE 104, such as a battery charge sensor, a thermal sensor, a screen mode sensor, etc. Although not shown, it is understood that the UE 104 may further include one or more batteries or other portable power sources, one or more user interface (UI) components (such as a touch screen, a user-manipulable input / output device (e.g., a "button" or keyboard) or other touch / contact sensor, a microphone or other voice sensor for capturing audio content, etc.).
[0040] The one or more memories 208 of the UE 104 are used to store one or more executable software instruction sets and associated data that operate the one or more processors 206 and other components of the UE 104 to perform various functions described herein and attributed to the UE 104. The executable software instruction sets include, for example, an operating system (OS) and various drivers (not shown) and various software applications. The executable software instruction sets further include one or more of a neural network management module 218, a capability management module 220, or an RRM module 222. The neural network management module 218 implements one or more neural networks for the UE 104, as described in detail below. The capability management module 220 determines various capabilities of the UE 104 related to neural network configuration or selection and reports such capabilities to the serving BS 110-1 (e.g., in one or more UECapabilitiesInformation RRC messages), as well as monitoring the UE 104 for changes in such capabilities (including changes in RF and processing capabilities, changes in accessory availability or capabilities, etc.) and managing the reporting of such capabilities and changes in capabilities to the serving BS 110-1. The RRM module 222 operates to perform RRM procedures at the UE 104 , including RRM measurements and reporting and RRM actions 118 specified by the output of the RRM NN 112 .
[0041] To facilitate the operations of UE 104 as described herein, one or more memories 208 of UE 104 may further store data associated with these operations. The data may include, for example, one or more neural network architecture configurations 224 ( Figure 1 124 ), and device data 226. The device data 226 represents, for example, user data, multimedia data, beamforming codebooks, software application configuration information, etc. The device data 226 may further include capability information of the UE 104, such as sensor capability information about one or more sensors of the sensor set 210, including the presence or absence of specific sensors or sensor types, and for those sensors that are present, one or more representations of the corresponding types and capabilities of those sensors, such as the range and resolution of a lidar or radar sensor, the image resolution and color depth of an imaging camera, etc. The capability information may further include information about, for example, the capability or status of a battery, the capability or status of the radio interface 204 and antenna 203 (e.g., frequency capabilities, radio measurement capabilities, etc.), etc.
[0042] Each neural network architecture configuration 224 includes one or more data structures containing data and other information representing a corresponding architecture and / or parameter configuration used by the neural network management module 218 to form a corresponding RRM neural network 112 for the UE 104. The information included in the neural network architecture configuration 224 includes, for example, parameters specifying a fully connected layer neural network architecture, a convolutional layer neural network architecture, a recurrent neural network layer, a number of connected hidden neural network layers, an input layer architecture, an output layer architecture, a number of nodes utilized by the neural network, coefficients (e.g., weights and biases) utilized by the neural network, kernel parameters, a number of filters utilized by the neural network, a stride / pooling configuration utilized by the neural network, an activation function for each neural network layer, interconnections between neural network layers, neural network layers to be skipped, etc. Thus, the neural network architecture configuration 224 includes any combination of neural network formation configuration elements that can be used to create a neural network architecture configuration (e.g., a combination of one or more neural network formation configuration elements) that defines and / or forms a DNN or other neural network.
[0043] Figure 3 An example hardware configuration of BS 110 (as a representative infrastructure component) according to some embodiments is shown. It should be noted that the depicted hardware configuration represents the processing components and communication components most directly related to the neural network-based processes described herein, and omits certain components that are well known to be frequently implemented in such electronic devices, such as displays, non-sensor peripherals, external power supplies, etc. It should be further noted that although the illustrated diagram represents the implementation of BS 110 as a single network node (e.g., a 5G NR Node B or "gNB"), the functionality of BS 110 and therefore its hardware components may be distributed across multiple network nodes or devices and may be distributed in a manner to perform the functions described herein.
[0044] In the depicted configuration, the BS 110 includes an RF front end 302 having one or more antennas 303 and a radio interface 304 having one or more modems to support one or more RATs and operating as a PHY transceiver interface to conduct and process signaling between one or more processors 306 of the BS 110 and the antennas 303, thereby facilitating various types of wireless communications. The antennas 303 may be arranged as one or more arrays of multiple antennas configured similarly or differently to each other, and may be tuned to one or more frequency bands associated with the corresponding RAT. The one or more processors 306 may include, for example, one or more CPUs, GPUs, TPUs, or other ASICs, etc. The BS 110 further includes one or more computer-readable media 308, including any of a variety of media used by electronic devices to store data and / or executable instructions, such as RAM, ROM, cache, flash memory, SSD, or other mass storage devices, etc. As with the memory 208 of the UE 104, given that system memory or other memory is often used to store data and instructions for execution by the processor 306, the computer-readable medium 308 is referred to herein as "memory 308", but it should be understood that references to "memory 308" should also apply to other types of storage media unless otherwise specified.
[0045] The one or more memories 308 of BS 110 are used to store one or more executable software instruction sets and associated data that operate the one or more processors 306 and other components of BS 110 to perform various functions described herein and attributed to BS 110. The executable software instruction sets include, for example, an OS and various drivers (not shown) and various software applications. The executable software instruction sets further include one or more of a neural network management module 310, a training module 312, or an RRM module 314. The one or more memories 308 further store various information, such as one or more candidate neural network architecture configurations 324 ( Figure 1 322 of an embodiment of an RRM NN architecture configuration 124 of a UE 014, and various BS data 326. The BS data 326 represents, for example, a beamforming codebook, software application configuration information, RRM scheme information, etc. The neural network architecture configuration 324 represents a trained neural network architecture configuration that can be employed at the RRM NN 112 of the UE 014 or other UEs. Figure 2Like the candidate neural network architecture configuration 224, each candidate neural network architecture configuration 324 includes one or more data structures containing data and other information representing the corresponding architecture and / or parameter configuration used by the neural network management module 218 of the UE (such as UE 104) to form the corresponding RRM NN 112 at the UE. The neural network management module 310 manages the training, retraining, selection, and delivery of the neural network architecture configuration 324 to the UE 104 and other UEs. The training module 312 performs the actual training / retraining of the selected neural network architecture configuration 324 using sensor data, radio measurement data, and other feedback from one or more UEs. The RRM module 314 operates to perform various RRM processes to be performed by the BS 110.
[0046] Figure 4 An example machine learning (ML) module 400 for implementing a neural network according to some embodiments is shown. As described herein, the UE 104 implements one or more DNNs or other neural networks as RRM NN 112 for configuring or controlling the RRM process at the UE 104. Relatedly, the BS 110 trains, retrains, or otherwise updates a copy of one or more DNNs or other neural networks used as RRM NN 112 at one or more UEs. Therefore, the ML module 400 shows an example module for implementing one or more of these neural networks.
[0047] In the depicted example, the ML module 400 implements at least one deep neural network (DNN) 402 having a group of connected nodes (e.g., neurons and / or perceptrons) organized into three or more layers. The nodes between the layers can be configured in a variety of ways, such as a partially connected configuration (wherein a first subset of nodes in a first layer is connected to a second subset of nodes in a second layer), a fully connected configuration (wherein each node in a first layer is connected to each node in a second layer), etc. Neurons process input data to produce continuous output values, such as any real number between 0 and 1. In some cases, the output value indicates how close the input data is to the desired category. Perceptrons perform linear classification on the input data, such as binary classification. Whether neurons or perceptrons, nodes can use a variety of algorithms to generate output information based on adaptive learning. Using the DNN 402, the ML module 400 performs a variety of different types of analysis, including single linear regression, multivariate linear regression, logistic regression, stepwise regression, binary classification, multiclass classification, multivariate adaptive regression splines, local estimation scatter smoothing, etc.
[0048] In some implementations, the ML module 400 performs adaptive learning based on supervised learning. In supervised learning, the ML module 400 receives various types of input data as training data. The ML module 400 processes the training data to learn how to map the input to the desired output. As an example, the ML module 400 receives a sequence of training data in the form of training sensor data and training radio measurement data, and learns how to effectively map the input training data to the desired output, i.e., RRM action. In particular, during the training procedure, the ML module 400 uses labeled or known data as input to the DNN 402. The DNN 402 uses nodes to analyze the input and generate corresponding outputs. The ML module 400 compares the corresponding output with the real data and adapts the algorithm implemented by the node to improve the accuracy of the output data. Thereafter, the DNN 402 applies the adapted algorithm to the unlabeled input data to generate the corresponding output data. The ML module 400 uses one or both of statistical analysis and adaptive learning to map the input to the output. For example, the ML module 400 uses the characteristics learned from the training data to relate unknown inputs to outputs that are statistically likely to be within a threshold range or value. This allows the ML module 400 to receive complex inputs and identify corresponding outputs. As described above, some implementations train the ML module 400 based on the characteristics of RRM decisions based on the input sensor data and radio measurement data. This allows the trained ML module 400 to receive sensor data sets (e.g., sensor data from a single time slice or within a series of time slices) and radio measurement data sets (e.g., from a single time slice or within a series of time slices), and generate outputs from these inputs that represent RRM actions to be performed.
[0049] In the depicted example, DNN 402 includes an input layer 404, an output layer 406, and one or more hidden layers 408 located between the input layer 404 and the output layer 406. Each layer has an arbitrary number of nodes, where the number of nodes between layers can be the same or different. That is, input layer 404 can have the same number and / or a different number of nodes as output layer 406, output layer 406 can have the same number and / or a different number of nodes as one or more hidden layers 408, and so on.
[0050] Node 410 corresponds to one of the several nodes included in input layer 404, wherein these nodes perform separate, independent calculations. As further described, the node receives input data and processes the input data using one or more algorithms to generate output data. Typically, the algorithm includes weights and / or coefficients that change based on adaptive learning. Therefore, the weights and / or coefficients reflect the information learned by the neural network. In some cases, each node can determine whether to pass the processed input data to one or more next nodes. For illustration, after processing the input data, node 410 can determine whether to pass the processed input data to one or both of nodes 412 and nodes 414 of hidden layer 408. Alternatively or additionally, node 410 passes the processed input data to the node based on the layer connection architecture. This process can be repeated throughout multiple layers until DNN 402 generates output using nodes (e.g., nodes 416) of output layer 406.
[0051] Neural networks may also employ a variety of architectures that determine which nodes in the neural network are connected, how data is advanced and / or retained in the neural network, which weights and coefficients are used to process input data, how data is processed, and so on. These various factors collectively describe a neural network architectural configuration, such as the neural network architectural configuration briefly described above. To illustrate, a recurrent neural network (such as a long short-term memory (LSTM) neural network) forms loops between node connections to retain information from previous portions of an input data sequence. The recurrent neural network then uses the retained information for subsequent portions of the input data sequence. As another example, a feedforward neural network passes information to a forward connection without forming a loop to retain the information. Although described in the context of node connections, it should be understood that a neural network architectural configuration may include a variety of parameter configurations that affect how DNN 402 or other neural networks process input data.
[0052] The neural network architecture configuration of a neural network can be characterized by various architectures and / or parameter configurations. For illustration, consider an example where DNN 402 implements a convolutional neural network (CNN). Typically, a convolutional neural network corresponds to a type of DNN in which layers process data using convolution operations to filter input data. Thus, a CNN architecture configuration can be characterized by, for example, pooling parameters, kernel parameters, weights, and / or layer parameters.
[0053] Pooling parameters correspond to parameters of a pooling layer that specifies a dimensionality of input data within a convolutional neural network. For illustration, a pooling layer may combine the outputs of nodes at a first layer into the inputs of nodes at a second layer. Alternatively or additionally, pooling parameters specify how and where the neural network pools data in a data processing layer. (For example) a pooling parameter indicating "max pooling" configures the neural network to perform pooling by selecting the maximum value from a grouping of data generated by nodes of the first layer, and using the maximum value as an input to a single node of the second layer. A pooling parameter indicating "average pooling" configures the neural network to generate an average value from a grouping of data generated by nodes of the first layer, and using the average value as an input to a single node of the second layer.
[0054] Kernel parameters indicate the filter size (e.g., width and height) used to process input data. Alternatively or additionally, kernel parameters specify the type of kernel method used to filter and process input data. (For example) support vector machine corresponds to a kernel method that uses regression analysis to identify data and / or classify data. Other types of kernel methods include Gaussian process, canonical correlation analysis, spectral clustering methods, etc. Therefore, kernel parameters can indicate the type of filter size and / or kernel method to be applied in a neural network. Weight parameters specify weights and biases used by algorithms within nodes to classify input data. In some implementations, weights and biases are learned parameter configurations, such as parameter configurations generated from training data. Layer parameters specify layer connections and / or layer types, such as fully connected layer types indicating that each node in the first layer (e.g., output layer 406) is connected to each node in the second layer (e.g., hidden layer 408), partially connected layer types indicating which nodes in the first layer are disconnected from the second layer, activation layer types indicating which filters and / or layers are activated within the neural network, and the like. Alternatively or additionally, the layer parameter specifies the type of the node layer, such as a normalization layer type, a convolutional layer type, a pooling layer type, etc.
[0055] Although described in the context of pooling parameters, kernel parameters, weight parameters, and layer parameters, it should be understood that other parameter configurations may be used to form a DNN consistent with the guidance provided herein. Thus, a neural network architecture configuration may include any suitable type of configuration parameters that may be applied to a DNN that affect how the DNN processes input data to generate output data.
[0056] In some embodiments, the configuration of the ML module 400 is further based on the sensor capabilities and / or radio measurement capabilities of the UE in which the ML module 400 is to be implemented. As such, the architectural configuration of the ML module 400 may also be based on the capabilities of the UE in which the ML module 400 is implemented. For example, the UE 104 may have considerable imaging capabilities, and thus the ML module 400 of the UE 104 may be trained based on image data as input, so as to facilitate, for example, the ML module 400 to generate RRM actions that are well suited to the RF transmission environment, which RRM actions depend on the presence or absence of objects or other interference sources to be represented in such image data. However, for UEs without imaging capabilities, using the ML module 400 that is configured via extensive training based on training image data will generally be less suitable for generating effective RRM actions based on input sensor data that lacks image data. Therefore, in some embodiments, the device implementing the ML module 400 may be configured to implement different neural network architectural configurations for different combinations of sensor capabilities, radio measurement capabilities, or both. For example, when a radar sensor is available at the device, the device may access one or more neural network architecture configurations for use, and when the radar sensor is not available but the radar sensor is available, the device may access a different set of one or more neural network architecture configurations for use.
[0057] In at least some embodiments, the UE 104 implementing the ML module 400 locally stores some or all of the candidate neural network architecture configurations in the set of candidate neural network architecture configurations that can be used for the ML module 400. For example, the candidate neural network architecture configurations 224 can be indexed at the UE 104 by a lookup table (LUT) or other data structure that takes as input one or more parameters (such as one or more sensor capability parameters or radio measurement capabilities) and outputs an identifier associated with a corresponding locally stored candidate neural network architecture configuration suitable for operation in view of the input parameters. In other embodiments, the UE 104 provides an indication of its capabilities, and the BS 110 or other infrastructure component selects a neural network architecture configuration for implementation at the UE 104 based on these capabilities. To facilitate the process of selecting an appropriate neural network architecture configuration, in at least one embodiment, the BS 110 or other infrastructure component uses the neural network management module 310 and the training module 312 to train different versions of the ML module 400. For example, the training module 312 can mathematically generate training data, access a file storing training data, obtain real-world communication data, etc. The neural network management module 310 then extracts and stores the various learned neural network architecture configurations for subsequent use. Some implementations store input characteristics with each neural network architecture configuration, where the input characteristics describe various sensor characteristics and / or radio measurement characteristics.
[0058] As described above, a wireless device (such as UE 104) can be configured to determine one or more RRM actions using one or more RRM DNNs, wherein each RRM DNN supplements or replaces one or more functions conventionally implemented by one or more hard-coded or fixed design blocks. Further, each DNN can further incorporate current sensor data from one or more sensors of the wireless device's sensor set, and in effect, modify or otherwise adapt its operation to take into account the current RF signal propagation environment or operating state of the wireless device as reflected in the sensor data. To this end, Figure 5 An example operating environment 500 for implementing a DNN at a UE 104 is shown. In the depicted example, the neural network management module 218 of the UE 104 implements an RRM processing module 502 for general RRM decision making processes. Furthermore, to illustrate a specific use case of neural network-based RRM management at the UE, the example implementation of the neural network management module 218 is described as additionally implementing a CHO processing module 504, which is a specific example implementation of an RRM processing module for the specific purpose of conditional RRM decision making (in this case, CHO decision making). Thus, the CHO processing module 504 may be understood to be a separate RRM processing module that is specifically implemented for CHO decision making, or alternatively may be understood to be that portion of the RRM processing module 502 that results in RRM actions that affect the CHO decision making.
[0059] As about Figure 5 104, based on the sensor capabilities of the sensors in the sensor set 210 and / or the radio measurement capabilities of the radio interface 204, a neural network architecture configuration 224 is supplied to the neural network management module 218 for each of the RRM processing module 502 and the CHO processing module 504. Figure 2 ). According to the corresponding schedule, the sensor set 210 provides the neural network management module 218 with the most recent local sensor data set 514 obtained from one or more sensors of the sensor set 210 ( Figure 1 The radio interface 204 provides the most recent radio measurement data set 516 ( Figure 1The neural network management module may also provide the data sets as input to the RRM processing module 502 and the CHO processing module 504. Other inputs such as the current RRC state (IDLE, INACTIVE, CONNECTED) or the RAT mode (e.g., standalone (SA) or non-standalone (NSA) mode of the 5G NR RAT) may also be provided to the processing modules 502 and 504.
[0060] The RRM processing module 502 utilizes these inputs to generate at least one RRM action 518 (an embodiment of the RRM action 118), which is provided to the RRM module 222 for implementation. As explained above, the RRM action 518 may represent: an RRM process to be performed, such as making RRM measurements and reporting the RRM measurements to the serving BS 110-1; an RRM process to be skipped, such as skipping a scheduled cell reselection process; a modification to one or more RRM processes, such as reconfiguring the intra-RAT measurement periodicity; or a combination thereof. As such, the RRM module 222 controls one or more parameters and other controls of the radio IF 204 to implement the RRM action 518. For example, the periodicity of the inter-RAT scan may be implemented as a value stored in a register of the radio IF 204, and the RRM module 222 may implement the RRM action 518 that attempts to change the periodicity by overwriting the register with a new value. As another example, the radio interface 204 may provide an application programming interface (API) or other interface, and the RRM module 222 may trigger the radio interface 204 to perform the RRM measurement represented by the RRM action 518 by triggering the RRM measurement via the API or other interface.
[0061] At the same time (or as part of the RRM processing module 502), the CHO processing module 504 utilizes the input to generate an output representing a CHO action 520, which is also provided to the RRM module 222 for implementation. The CHO action 520 may include, for example, a CHO decision, such as a decision to initiate a handover based on a handover condition as interpreted by the local sensor data set 514 and the radio measurement data set 516, or a decision to abandon the decision to initiate a handover based on the handover condition interpreted by these inputs. Additionally or alternatively, the CHO action 520 may include a modification to a static CHO configuration. For example, the BS 110 may have sent a message to the UE 104 to implement CHO, where a handover is initiated when a specified set of monitored conditions is met, and the CHO action 520 may include a modification to a threshold or other trigger represented in the monitored set of conditions. For example, the BS 110 may specify that a handover will be initiated when the measured RSRP is below a specified threshold. However, the CHO processing module 504 may have been trained using training data that reflects a situation where the UE has a low measured RSRP but has a high speed, and the measured RSRP may change rapidly as the UE's position relative to the serving cell changes rapidly. As such, the results of the training can configure the CHO processing module 504 to act to modify the threshold for RSRP before triggering a handover when the input local sensor data set 514 includes sensor data indicating that the UE 104 is moving rapidly and the radio measurement data set 516 indicates that the measured RSRP fluctuates around the original specified threshold. In either approach, the RRM module 222 receives the CHO action 520 and coordinates with the radio interface 204 to implement the process represented by the CHO action 520, such as initiating a handover, abandoning initiating a handover, or modifying one or more monitored parameters of the conditions that the radio interface 204 is to monitor for deciding whether to initiate a handover.
[0062] Figure 6 An example operating environment 600 of a BS 110 (e.g., serving BS 110-1) for supporting neural network-based RRM at a UE 104 in accordance with at least one embodiment is shown. As described above, for a neural network-based RRM process, in some embodiments, BS 110 operates to: provide a neural network architecture configuration to UE 104 that is also suitable for the specific sensor capabilities and / or radio measurement capabilities of UE 104, train / retrain the neural network architecture configuration, or a combination of both. To select and provide an initial DNN architecture configuration to UE 104, BS 110 implements a neural network architecture configuration datastore 602, or has stored the configuration datastore via network infrastructure 102 ( Figure 1) server or other infrastructure component accesses the data repository 602 via a network. The data repository 602 stores one or more candidate neural network architecture configurations 324 ( Figure 3 ), the one or more candidate neural network architecture configurations may be indexed or otherwise accessed based on corresponding sensor capability attributes, radio measurement attributes, etc. Thus, when UE 104 supplies BS 110 with UE capability information 604 indicating one or both of its sensor capabilities or its radio measurement capabilities, the RF front end 302 of BS 110 passes the UE capability information 604 to the RRM module 314, which utilizes the UE capability information 604 to select from the data repository 602 a neural network architecture configuration 324 that has been trained using training data having the same or similar sensor capabilities and / or the same or similar radio measurement capabilities. The selected neural network architecture configuration 324, or an identifier or other representation thereof, is then sent to UE 104 for implementation.
[0063] In addition, in response to providing the neural network architecture configuration 324 to the UE 104, in some embodiments, the neural network management module 310 of the BS 110 can direct the training module 312 to instantiate a training processing module 606 that implements a DNN or other neural network that is initially configured with the selected neural network architecture configuration 324 supplied to the UE 104. That is, the training processing module 606 is actually a copy of the RRM NN 112 that is simultaneously operating at the BS 110. Thereafter, the UE 104 provides feedback in the form of a copy of the local sensor data set 514 input to the RRM NN 112, a copy of the radio measurement data set 516 input to the RRM NN 112, and / or an RRM report 608 from the UE 104, which is provided in response to the UE 104 performing various RRM processes as a result of the RRM actions 118 generated by the RRM NN 112 as a result of receiving the local sensor data set 514 and the radio measurement data set 516 as input. Thus, this feedback from the UE 104 represents an input provided to the RRM NN 112. Thus, the training module 312 can use this feedback as an input to the training processing module 606 to retrain or update the DNN or other neural network implemented therein. Periodically or in response to some other triggering event, the modified neural network architecture configuration 624 of the training processing module 606 is extracted and provided to the RRM module 314, which can then store the representation of the modified neural network architecture configuration 624 in the data repository 602 and / or send the representation of the modified neural network architecture configuration 624 to the UE 104 for use by the UE 104 in updating the RRM NN 112.
[0064] Now go to Figure 7 and Figure 8 , describes a method 700 for implementing an RRM scheme in a wireless communication network 100 according to at least one embodiment, the method utilizing a neural network-based RRM process at a UE 104. It should be noted that the order of operations described with reference to the method 700 is for illustration purposes only, and a different order of operations may be performed, and further, one or more operations may be omitted, or one or more additional operations may be included in the method shown. For ease of understanding, reference is made to Fig. 9 Method 700 is described with reference to an example implementation of method 700 reflected in ladder diagram 900 of FIG.
[0065] In an implementation, the UE 104 may have any of a variety of sensor capabilities and radio measurement capabilities. For example, sometimes the UE light sensor is turned on, while other times the UE turns off its light sensor to save power. As another example, some UEs 104 may have satellite-based positioning sensors, while other UEs 104 may not. As yet another example, one UE 104 may have radar or lidar capabilities but no camera capabilities, while another UE 104 may have camera capabilities but no radar or lidar capabilities. Because the RRM neural network implemented at the UE 104 uses sensor data and / or radio measurements as input to prescribe its operation, it should be understood that in many cases, the specific neural network architecture configuration implemented at the UE 104 is based on a specific sensor that can be used to provide sensor data as input, and / or is based on a specific radio measurement that the UE can obtain; that is, the specific neural network architecture configuration implemented at the RRM NN 112 reflects one or both of the types and combinations of sensors currently providing input to the RRM NN 112 and the types and parameters of radio measurements currently provided as input to the RRM NN 112.
[0066] Therefore, method 700 starts at box 702 to determine the sensor capabilities and radio measurement capabilities of one or more test Ue, which may include UE 104 or may utilize Ue other than UE 104. It should be understood that in this sense, the "test Ue" may not actually be a physical Ue, but a software simulation of the operation of the test Ue for the purpose of neural network training. As such, references to "test UE" should be understood to also include references to such simulations. For the training process, a training module (such as the training module 312 of BS 110, or a similar training module of a server or other infrastructure component) selects a specific sensor configuration and / or radio measurement configuration for training the candidate neural network architecture of the RRM NN 112 of the test UE. In some embodiments, the training module may attempt to train each arrangement of available sensors and each arrangement of available radio measurement capabilities. However, in an implementation in which non-test Ue has a relatively large number and variety of suitable sensor capabilities or radio measurement capabilities, such efforts may be impractical. Therefore, in other embodiments, the training module selects only from a limited representative set of potential sensors and sensor configurations and radio measurement capabilities. For illustration, the lidar information from different lidar modules manufactured by the same company may be relatively consistent, and therefore, if, for example, UE 104 can implement any of a variety of lidar sensors from the manufacturer, the training module can choose to eliminate several lidar sensors from the sensor configuration being trained. As another example, although a very small subset of UE 104 may have the capacity to measure SINR at a wider range or higher resolution, the training module can limit training to a smaller SINR range or lower SINR resolution that still covers most implementations of UE 104. In yet other embodiments, there may be a set of defined sensor configurations that the training module can select for training, and therefore the training module selects a sensor configuration from the defined set (and avoids selecting a sensor configuration that relies on sensor capabilities that are not generally supported by the associated device).
[0067] In the event that a sensor / radio measurement configuration is selected, the training module identifies one or more training data sets based on the selected configuration for use in training the candidate neural network architecture configuration. That is, the one or more training data sets include or represent sensor data that can be generated by comparable sensors of the selected sensor configuration, and represent radio measurement data that can be generated consistently with the selected radio measurement capacity, and are therefore suitable for training the candidate neural network architecture configuration to operate with sensor data provided by the specific sensors represented in the selected sensor / radio measurement configuration. The training data may further include, for example, training data related to an operating state of the test UE, such as an RRC state or a 5G NR SA or NSA mode. In the event that one or more training sets are obtained, the training module initiates training the RRM NN at the test UE. The training typically involves: initializing the bias weights and coefficients of the various RRM NNs to have initial values, which are typically pseudo-randomly selected; then inputting a training data set (which represents, for example, known sensor data from sensors in the selected sensor configuration and known radio measurement data consistent with the selected radio measurement capability configuration); processing the input at the test RRM NN to generate an output; determining the error between the actual output and the expected output; and back-propagating the error throughout the test RRM NN; and repeating the process for the next input data set. The process repeats until a certain number of training iterations have been performed, or until a certain minimum error rate has been reached. Because of the training of the RRM NN of the test UE, the resulting neural network has a specific neural network architecture configuration, or in the case where the implemented neural network is a DNN, has a specific DNN architecture configuration, which characterizes the architecture and parameters of the corresponding RRM NN, such as the number of hidden layers, the number of nodes at each layer, the connections between each layer, the weights, coefficients and other bias values implemented at each node, etc. Thus, when the training of the RRM NN of the test UE for the selected sensor / radio measurement configuration is completed, the current neural network architecture configuration of the trained RRM NN is extracted and, at block 704, the corresponding candidate neural network architecture configuration 324 ( Figure 3 ) is stored at BS 110, or by storing a copy of the candidate neural network architecture configuration 324 in a remote data repository accessible by BS 110, so that the current neural network architecture configuration is available for use as a candidate neural network architecture configuration 324 for one or more BSs 110. In at least one embodiment, the candidate neural network architecture configuration 324 can be generated by extracting the architecture and parameters of the corresponding RRMNN (such as the number of hidden layers, the number of nodes, connections, coefficients, weights and other bias values, etc.) at the end of training.
[0068] In the event that one or more other sensor / radio measurement configurations remain to be trained, the training process repeats with the selection of the next sensor / radio measurement configuration to be trained. Otherwise, if the candidate RRM NN for UE 104 has been trained for all expected sensor / radio measurement configurations, the network 100 may turn to using the trained DNN to support a neural network-based RRM process at UE 104.
[0069] Thus, as part of or subsequent to initialization of a wireless connection between UE 104 and serving BS 110-1, at block 706, UE 104 initiates the wireless connection by sending a sensor / measurement capability message 902 ( Fig. 9 ) to inform the serving BS 110-1 of its sensor capabilities and / or radio measurement capabilities. For example, during the attach procedure, the serving BS 110-1 may send a UECapabilitiesEnquiry RRC message to the UE 104, and in response, the UE 104 sends a UECapabilitiesInformation RRC message to the serving BS 110-1, wherein one or more fields of the UECapabilitiesInformationRRC message (an embodiment of the sensor / measurement capabilities message 902) contain data representing one or both of the sensor capabilities or the radio measurement capabilities of the UE 104.
[0070] At block 708, the neural network management module 310 of the serving BS 110-1 selects a candidate neural network architecture configuration 324 from the set 322 for implementation at the RRM NN 112 of the UE 104 based on one or both of the sensor capabilities advertised by the UE 104 or the radio measurement capabilities advertised by the UE 104. For example, if the UE 104 has a radar sensor and an imaging camera available for use and is capable of making RSRP measurements at a particular range and resolution, the neural network management module 310 may select a candidate neural network architecture configuration 324 that has been trained for the particular sensor configuration and radio measurement capabilities for the UE 104. In the event that a suitable neural network architecture configuration is selected, at block 710, the serving BS 110-1 directs the UE 104 to implement the selected neural network architecture configuration at the RRM NN 112 by sending a NN configuration message 904 to the UE 104. The NN configuration message 904 may contain one or more data structures that implement the selected neural network architecture configuration itself, or it may contain an index or other identifier of the selected neural network architecture configuration so that the UE 104 can access the selected neural network architecture configuration from a local data repository or from a remote server. In other embodiments, independent of guidance from the serving BS 110-1, the neural network management module 218 of the UE 104 selects a neural network architecture configuration to implement at the RRM NN 112 based on the current sensor / measurement configuration of the UE 104.
[0071] In response to the NN configuration message 904 or in response to the selection of the UE 104's own neural network architecture configuration, at block 712, the UE implements the indicated neural network architecture configuration at the RRM NN 112 of the UE 104. After the RRMNN 112 is so initialized, the UE 104 can continue RRM management at the UE 104 based on the RRM actions 118 indicated by the output of the RRM NN 112. Thus, for an iteration of the input / output cycle using the RRM NN 112, at block 714, the sensor set 210 provides a sensor data set 914 ( Figure 1 114) as an input to the RRM NN 112 of the UE 104, and simultaneously, at block 716, the RRM module 222 of the UE 104 and the RF front end 202 together obtain one or more radio measurements and generate a radio measurement data set 916 ( Figure 1The UE 104 may provide a separate sensor data set 914 and a radio measurement data set 916 every 10 milliseconds (ms). In other embodiments, one or both of these data sets are sequences of data sampled over multiple time slices in a sliding window, such as a sequence of 10 separate data sets obtained every 100 ms over the previous 1 second.
[0072] At block 718, the RRM NN 112 processes these inputs, and in some embodiments, additional inputs (such as the current RRC state or the current 5G NR mode) to generate a message representing one or more RRM actions 918 (based on the current neural network architecture configuration and current state of the nodes of the RRM NN 112). Figure 1 At block 720, the RRM module 222 of the UE 104 initiates implementation of the RRM action 918, such as by triggering the RF front end 202 to perform RRM measurements or other RRM procedures, directing the RF front end 202 to abandon one or more scheduled RRM measurements or other RRM procedures, reconfiguring one or more parameters of the scheduled RRM procedures, or a combination thereof.
[0073] Consistent with typical RRM schemes, in at least one embodiment, UE 104 contributes to the RRM scheme of wireless network 100 by providing RRM reports to serving BS 110-1, which can be utilized by serving BS 110-1, core network 106, or other infrastructure components to make RRM decisions for the entire wireless system. The RRM reporting can be performed in response to conventional RRM procedures implemented at UE 104, such as implementation of a static RRM measurement scheme separate from RRM NN 112. The RRM reporting can also be performed in response to execution of RRM action 918 by UE 104. Thus, in Figure 8Continuing with method 700, at block 722, in response to performing one or more RRM processes at UE 104 due to an RRM action 918 generated by RRM NN 112 or due to a separate and parallel static RRM process implemented at UE 104, UE 104 provides an RRM report 920 to serving BS 110-1. Typically, RRM report 920 includes an RRM measurement report that includes one or more RRM measurements performed by UE 104 in a previous period. For example, if RRM action 918 generated at block 718 indicates a directive for UE 104 to perform a licensed assisted access (LAA) survey, as part of the LAA survey, UE 104 may identify a subset of neighboring cells and then perform radio measurements on the subset of cells, such as by measuring RSRP or SINR for each cell in the subset. Thus, in this example, the RRM report 920 may include an RRM measurement report containing the measured RSRP or measured SINR for each cell in the identified subset.
[0074] In response to the information contained in the RRM report 920, or in response to another triggering factor (such as the expiration of a specified duration since the last update), the serving BS 110-1 can determine that it is an appropriate time to update the underlying architecture configuration employed in the RRM NN 112. This update process generally relies on utilizing the same real-world inputs provided to the RRM NN 112 at the test UE implemented at the serving BS 110-1 (as described above), and then iteratively back-propagating any errors between the actual results output by the test UE and the expected or projected results. Therefore, at block 724, the serving BS 110-1 sends a sensor / radio request 922 to the UE 104 ( Fig. 9 ) to request that UE 104 provide a copy of some or all of the sensor data sets 914 and radio measurement data sets 916 input to RRMNN 112 since, for example, the last update process. In response to sensor / radio request 922, at block 726, UE 104 provides the requested sensor / radio measurement data 924 ( Fig. 9 ) to the serving BS 110-1. In another embodiment, rather than buffering the sensor data and the radio measurement data to send to the serving BS 110-1 on request and in a single burst, the UE 104 may send a copy of each sensor data set 914 and / or each radio measurement data set 916 to the serving BS 110-1 as it is generated, or otherwise in parallel with providing these data sets as input to the RRM NN 112.
[0075] At block 728, the training module 312 of the serving BS 110-1 utilizes the sensor data and / or radio measurement data received from the UE 104 and performs a neural network update process 926 to retrain or otherwise update a test copy of the RRM NN 112 implemented by the training module 312 as a test UE by providing the data as input to the RRMNN 112 and revising the underlying neural network architecture configuration of the test copy based on a comparison of the resulting output with the expected or predicted output. The training may further include evaluating the RRM reports 920 narrated during a time period associated with the retraining data to evaluate the correctness of the operation of the RRM NN 112 at the UE 104 and adjusting or otherwise revising the neural network architecture configuration of the test copy accordingly. Thereafter, the neural network management module 310 of the serving BS 110-1 may extract the updated neural network architecture configuration from the retrained test copy and provide the resulting updated neural network architecture configuration 324 to the RRM module 314 to replace or supplement the previous version in the data repository 602, and at block 730, send an update message 928 containing a representation of the updated neural network architecture configuration 324 to the UE 104. At block 732, the UE 104 updates the RRM NN 112 using the updated neural network architecture configuration 324 represented in the update message 928 by replacing the “old” neural network architecture configuration of the RRM NN 112 with the updated version, or by selectively modifying or enhancing the previous version of the RRM NN 112 using the updated version.
[0076] refer to Fig.10 , depicts an alternative implementation of method 700, referred to as method 700-1. Method 700-1 Figure 7 Boxes 702 to 720 and Figure 8The process of block 722 of the method 700 is performed in the same manner as in the method 700. However, after the RRM reporting process of block 722, in the method 700-1, the UE 104 performs the NN update process and reports the updated neural network architecture configuration to the serving BS 110-1, rather than the reverse. Therefore, in the method 700-1, after the RRM reporting of block 722, at block 1024, the neural network management module 218 of the UE 104 instantiates a copy of the RRM NN 112 as a training copy, and then evaluates some or all of the output RRM actions 118 generated from the input sensor data set and the radio measurement data set relative to the expected or predicted set of RRM actions given the same input (these expected RRM actions are specified by, for example, training guidelines provided by the network infrastructure 102). The neural network management module 218 then updates the underlying architecture configuration of the training copy of the RRM NN 112 by, for example, iteratively back-propagating the error between the actual RRM actions and the expected RRM actions. At block 1026, the neural network management module 218 then extracts the resulting modified neural network architecture configuration of the test copy of the RRM NN 112 as an updated neural network architecture configuration, and uses the updated neural network architecture configuration to update the actual RRM NN 112. At block 1028, the update may also be sent to the serving BS 110-1 for storage as a candidate neural network architecture configuration 324, and / or for propagation to one or more other UEs 104 that are using the same original neural network architecture configuration in their respective RRM neural networks.
[0077] As explained above, the RRM NN 112 may provide controls for implementing various conditional RRM procedures, where the network infrastructure 102 directs the UE 104 to conditionally perform RRM-related processes based on the status of one or more conditions to be monitored by the UE 104. CHO is an example of such a conditional RRM process, as the UE will be directed to initiate a handover if one or more monitored conditions subsequently reach corresponding thresholds. In some embodiments, the conditional RRM process is represented in the same RRM NN 112 that provides other types of RRM decisions at the UE 104. In other embodiments, one or more conditional RRM processes may be implemented as a separate conditional RRM neural network that operates in parallel with the RRM NN 112 and utilizes overlapping inputs. Fig.11An example method 1100 of operation of a conditional RRM neural network as the RRM NN 112 (or RRM processing module 502) or as a separate neural network (such as the CHO processing module 504) is shown. It should be noted that the order of operations described with reference to the method 1100 is for illustration purposes only, and a different order of operations may be performed, and further, one or more operations may be omitted, or one or more additional operations may be included in the method shown. To facilitate understanding, reference is made to the description of the operation in the method 1100. Fig.12 Method 1100 is described with reference to an example implementation of method 1100 reflected in ladder diagram 1200 of FIG.
[0078] Method 1100 initiates subsequent training of a set of candidate CHO neural network architecture configurations using the training process described above. Thus, at block 1102, UE 104 sends a sensor / measurement capability message 1202 ( Fig.12 ) to inform the serving BS 110-1 of its sensor capabilities and / or radio measurement capabilities. The message 1202 may include, for example, a UECapabilitiesInformationRRC message provided in response to a UECapabilitiesEnquiry RRC message from the serving BS 110-1. At block 1104, the neural network management module 310 of the serving BS 110-1 selects a candidate CHO neural network architecture configuration (e.g., an example of a neural network architecture configuration 324) from the set 322 based on one or more of the sensor capabilities announced by the UE 104 or the radio measurement capabilities announced by the UE 104. In the event that a suitable CHO neural network architecture configuration is selected, at block 1106, the serving BS 110-1 sends a CHO configuration message 1204 ( Fig.12 ) to direct UE 104 to implement the selected CHO neural network architecture configuration at CHO processing module 504. CHO configuration message 1204 may contain one or more data structures that implement the selected neural network architecture configuration itself, or it may contain an index or other identifier of the selected neural network architecture configuration so that UE 104 can access the selected neural network architecture configuration from a local data repository or a remote server. In other embodiments, independent of the guidance from serving BS 110-1, neural network management module 218 of UE 104 selects a CHO neural network architecture configuration to implement at CHO processing module 504 based on the current sensor / measurement configuration of UE 104.
[0079] In response to the CHO configuration message 1204 or in response to the selection of its own CHO neural network architecture configuration, at box 1108, the UE 104 implements the indicated CHO neural network architecture configuration at the CHO processing module 504 and thus ends the initialization of the CHO processing module 504.
[0080] The UE 104 may then continue conditional RRM management at the UE 104 based on the CHO action 520 indicated by the output of the CHO processing module 504. Thus, for an iteration of the input / output loop using the CHO processing module 504, at block 1110, the sensor set 210 provides a sensor data set 1214 ( Fig.12 )( Figure 1 10-2) as an input to the CHO processing module 504. As should be appreciated, the decision to initiate a handover due to a CHO instruction typically reflects the result of a comparison of the suitability of the current serving cell relative to one or more neighboring or target cells, where the suitability is indicated by a comparison of various radio measurements of the serving cell and the neighboring cells. As such, concurrently with the iteration of providing sensor data, at block 1112, the RRM module 222 and RF front end 202 of the UE 104 together obtain one or more radio measurements of the serving BS 110-1, and at block 1114, the RRM module 22 and RF front end 202 obtain one or more radio measurements of each applicable neighboring BS 110 (such as the target BS 110-2) and provide them as a radio measurement data set 1216 ( Fig.12 )( Figure 1 The radio measurements 116 of the embodiment of the present invention are provided as another input to the CHO processing module 504. In some embodiments, these data sets 1214, 1216 represent samples from a single time slice, while in other embodiments, one or both of these data sets are sequences of data sampled over multiple time slices in a sliding window.
[0081] At block 1116, the CHO processing module 504 processes these inputs, and in some embodiments, additional inputs (such as a current RRC state or a current 5G NR mode) to generate a representation of one or more CHO actions 1220 based on the current neural network architecture configuration and current state of the nodes of the CHO processing module 504. Fig.12 )( Figure 5The RRM module 222 of the UE 104 then initiates the implementation of the CHO action 1220. As described above, in some embodiments, the CHO action 1220 may include a process for adjusting some parameters of the CHO process (such as, changing a threshold value for making a switching decision). However, in other embodiments, the CHO action 1220 represents the switching decision itself. In such embodiments, at box 1118, the RRM module 222 determines the switching decision represented by the CHO action 1220. If the switching decision is to avoid initiating a handover, the method 1100 may return to boxes 1110, 1112, and 1114 to perform another iteration of the CHO action generation process using a new set of sensor and measurement data inputs to the CHO processing module 504.
[0082] Otherwise, if the handover decision is to initiate a handover, then at block 1120, RRM module 222 and RF front end 202 coordinate to proceed with the handover 1222 to the indicated neighboring cell (represented by target BS 110-1 in this example). Fig.12 ). After performing the handover, it should be appreciated that UE 104 is now connected to a different cell, which may have different RRM parameters than the previous cell associated with BS 110-1. Therefore, in at least one embodiment, the configuration of at least one of RRM NN 112 (e.g., RRM processing module 502) or CHO processing module 504 is replaced with a new configuration supplied by BS 110-2 as the new serving BS for UE 104. Therefore, at block 1122, the same or similar neural network configuration process described above with reference to blocks 1102 to 1108 may be performed again, but this time with a neural network architecture configuration selected and provided by BS 110-2 rather than BS 110-1. As such, UE 104 may provide an indication of its sensor and / or radio measurement capabilities as a capability message 1224 ( Fig.12 ) is provided to BS 110-2, and BS 110-2 can select an appropriate neural network architecture configuration 1226 based on these capabilities ( Fig.12 ), and directs UE 104 to implement the selected neural network architecture configuration 1226 for the CHO processing module 504.
[0083] Thereafter, the UE 104 may utilize the newly configured CHO processing module 504 to perform a neural network-based CHO decision process in the same manner as described above, including converting the sensor data set 1234 ( Fig.12 ) and radio measurement dataset 1236 ( Fig.12) are provided as inputs to CHO processing module 504. Based on these inputs, CHO processing module 504 generates CHO actions 1240 suitable for the CHO-related RRM scheme implemented by BS 110-2. UE 104 may then implement one or more CHO procedures represented by CHO actions 1240 as described herein.
[0084] Embodiments of the present disclosure may also be better understood by considering the following non-limiting examples:
[0085] Example 1: A computer-implemented method in a wireless device, comprising: receiving a first sensor data set from one or more sensors of the wireless device; receiving a first set of radio measurements from a radio interface of the wireless device that is distinct from the one or more sensors; processing the first sensor data set and the first radio measurement set at a radio resource management (RRM) neural network of the wireless device to generate a first output representing a first RRM action; and The first RRM action is performed at the wireless device.
[0086] Example 2: The method of Example 1, wherein the wireless device is a user equipment, and the method further comprises: providing at least one of the representations of sensor capabilities of the wireless device for receipt by an infrastructure component of a network infrastructure wirelessly connected to the wireless device; In response to providing the representation of sensor capabilities, receiving a first neural network architecture configuration from the infrastructure component; and implementing the first neural network architecture configuration at the RRM neural network.
[0087] Example 3: The method of Example 2, wherein the representation of sensor capabilities comprises one or more fields of a UECapabilitiesInformation radio resource control (RRC) message.
[0088] Example 4: The method as described in Example 2 or 3 further comprises: providing at least one of the first sensor data set or the first radio measurement set to the infrastructure component; receiving, in response to providing the at least one of the first sensor data set or the first radio measurement set, a second neural network architecture configuration from the infrastructure component, the second neural network architecture configuration representing a modification to the first neural network architecture configuration based on the at least one of the first sensor data set or the first radio measurement set; and The second neural network architecture configuration is implemented at the RRM neural network.
[0089] Example 5: The method as described in Example 2 or 3 further comprises: modifying the first neural network architecture configuration based on at least one of the first sensor data set or the first radio measurement set to generate a second neural network architecture configuration; and The second neural network architecture configuration is implemented at the RRM neural network.
[0090] Example 6: The method as described in any one of Examples 1 to 5, further comprising: receiving an indication of an operational status of the wireless device; and Wherein the processing includes: processing the first sensor data set, the first radio measurement set, and the operating state at the RRM neural network to generate the first output.
[0091] Example 7: A method as described in Example 6, wherein the operating state includes a radio resource control (RRC) state of the wireless device.
[0092] Example 8: A method as described in any of Examples 1 to 7, wherein the one or more sensors include at least one of the following: a position sensor; a posture sensor; an accelerometer; a pressure sensor; or a proximity sensor.
[0093] Example 9: The method of Example 8, wherein the position sensor comprises at least one of a satellite-based positioning sensor or a visual telemetry imaging sensor, and the posture sensor comprises at least one of a gyroscope or a visual telemetry imaging sensor.
[0094] Example 10: A method as described in any of Examples 1 to 9, wherein the first radio measurement set includes at least one signal power measurement of a serving cell or a neighboring cell.
[0095] Example 11: A method as described in Example 10, wherein the signal power measurement includes at least one of the following: signal-to-noise ratio (SNR) measurement, signal-to-interference-plus-noise ratio (SINR) measurement, reference signal received power (RSRP) measurement, or received signal strength indicator (RSSI) measurement.
[0096] Example 12: A method as described in any of Examples 1 to 11, wherein the RRM action includes at least one of the following: performing RRM-related measurements by the wireless device; configuring characteristics of RRM-related measurements to be performed by the wireless device; or performing an RRM reporting process at the wireless device.
[0097] Example 13: The method of Example 12, wherein the RRM-related measurements include at least one of: inter-frequency measurements; intra-frequency measurements; or inter-radio access technology (RAT) measurements.
[0098] Example 14: The method of Example 12 or 13, wherein the characteristic of the RRM-related measurement includes at least one of: a frequency or timing of the RRM-related measurement, or a frequency band or channel of the RRM-related measurement.
[0099] Example 15: The method of Example 1, wherein the RRM action includes performing a conditional handover (CHO) decision or a conditional PSCell change (CPC) decision.
[0100] Example 16: The method of Example 15 further comprises: In response to connecting to a cell due to a conditional handover decision: providing a representation of the sensor capabilities of the wireless device for receipt by an infrastructure component of the cell; In response to providing the representation of sensor capabilities, receiving a first neural network architecture configuration from the infrastructure component of the cell; and The second neural network architecture configuration of the RRM neural network is replaced with the first neural network architecture configuration.
[0101] Example 17: The method of Example 16 further comprises: receiving a second sensor data set from the one or more sensors of the wireless device; receiving a second radio measurement set from the radio interface of the wireless device; processing the second sensor data set and the second radio measurement set at the RRM neural network of the wireless device to generate a second output representing a second RRM action; and The second RRM action is performed at the wireless device.
[0102] Example 18: A method as described in any of Examples 1 to 17, wherein the wireless device is a user equipment.
[0103] Example 19: The method of Example 1, wherein the wireless device is a base station.
[0104] Example 20: The method of Example 19 further comprises: An RRM report is received from the user equipment.
[0105] Example 21: A wireless device comprising: Radio interface; at least one processor coupled to the radio interface; and At least one memory, the at least one memory coupled to the at least one processor, the at least one memory storing executable instructions configured to manipulate the at least one processor to perform a method as described in any one of Examples 1 to 20.
[0106] Example 22: A computer-implemented method in an infrastructure component wirelessly connected to a wireless device, comprising: receiving, from the wireless device, an indication of sensor capabilities of the wireless device; determining a first neural network architecture configuration based on the representation of sensor capabilities; and A representation of the first neural network architecture configuration is sent to the wireless device for implementation at a radio resource management (RRM) neural network of the wireless device for use in determining an RRM action to be performed by the wireless device.
[0107] Example 23: The method of Example 22 further comprises: receiving at least one of a sensor data set or a radio measurement set from the wireless device; modifying the first neural network architecture configuration based on the at least one of the sensor data set or the radio measurement set to generate a second neural network architecture configuration; and A representation of the second neural network architecture configuration is sent to the wireless device for implementation at the RRM neural network of the wireless device.
[0108] Example 24: A method as described in Example 22 or 23, wherein the sensor data set includes at least one of the following: position data; posture data; acceleration data; atmospheric pressure or altitude data; or proximity data.
[0109] Example 25: A method as described in any of Examples 22 to 24, wherein the radio measurement set includes at least one signal power measurement of a serving cell or a neighboring cell.
[0110] Example 26: A method as described in any of Examples 22 to 25, wherein the determined RRM action includes at least one of the following: performing RRM-related measurements by the wireless device, or configuring characteristics of RRM-related measurements to be performed by the wireless device.
[0111] Example 27: The method of any one of Examples 22 to 26, wherein the determined RRM action comprises performing a conditional handover decision.
[0112] Example 28: A method as described in any of Examples 22 to 26, wherein the infrastructure component includes a base station.
[0113] Example 29: A method as described in any of Examples 22 to 28, wherein the wireless device includes a user equipment.
[0114] Example 30: The method of any of Examples 22 to 29, wherein the representation of the sensor capabilities comprises one or more fields of a UECapabilitiesInformation radio resource control (RRC) message.
[0115] Example 31: An infrastructure component comprising: a collection of one or more sensors; a radio interface distinct from the set of one or more sensors; at least one processor coupled to the radio interface; and At least one memory, the at least one memory coupled to the at least one processor, the at least one memory storing executable instructions configured to manipulate the at least one processor to perform a method as described in any one of Examples 22 to 30.
[0116] In some embodiments, certain aspects of the techniques described above may be implemented by one or more processors of a processing system executing software. The software includes one or more executable instruction sets stored or otherwise tangibly embodied on a non-transitory computer-readable storage medium. The software may include instructions and certain data that manipulate one or more processors to perform one or more aspects of the techniques described above when executed by one or more processors. A non-transitory computer-readable storage medium may include, for example, a magnetic or optical disk storage device, a solid-state storage device (such as a flash memory), a cache, a random access memory (RAM), or one or more other non-volatile memory devices, etc. The executable instructions stored on a non-transitory computer-readable storage medium may be source code, assembly language code, object code, or another instruction format that is interpreted or otherwise executable by one or more processors.
[0117] Computer-readable storage media may include any storage media or combination of storage media that can be accessed by a computer system during use to provide instructions and / or data to the computer system. Such storage media may include, but are not limited to, optical media (e.g., compact discs (CDs), digital versatile discs (DVDs), Blu-ray discs), magnetic media (e.g., floppy disks, tapes, or magnetic hard drives), volatile memory (e.g., random access memory (RAM) or cache), non-volatile memory (e.g., read-only memory (ROM) or flash memory), or micro-electromechanical system (MEMS)-based storage media. Computer-readable storage media may be embedded in a computing system (e.g., system RAM or ROM), fixedly attached to a computing system (e.g., a magnetic hard drive), removably attached to a computing system (e.g., an optical disc or a flash memory based on a universal serial bus (USB)), or coupled to a computer system via a wired or wireless network (e.g., a network accessible storage device (NAS)).
[0118] It should be noted that not all activities or elements described above in the general description are required, parts of specific activities or devices may not be required, and one or more further activities may be performed in addition to those described, or one or more further elements may be included. Further, the order in which the activities are listed is not necessarily the order in which the activities are performed. In addition, the concepts have been described with reference to specific embodiments. However, it should be understood by those of ordinary skill in the art that various modifications and changes may be made without departing from the scope of the present disclosure as set forth in the appended claims. Therefore, the specification and drawings will be regarded as having an illustrative rather than a restrictive meaning, and all such modifications are intended to be included within the scope of the present disclosure.
[0119] Benefits, other advantages, and solutions to problems have been described above with respect to specific embodiments. However, benefits, advantages, solutions to problems, and any features that may make any benefit, advantage, or solution appear or become more significant should not be interpreted as being key, required, or essential features of any or all claims. In addition, the specific embodiments disclosed above are merely illustrative, as the disclosed subject matter may be modified and practiced in different but equivalent ways that are obvious to those skilled in the art having the benefit of the teachings herein. Except as described in the appended claims, it is not intended to limit the details of the construction or design shown herein. Therefore, it is apparent that the specific embodiments disclosed above may be changed or modified, and all such variations are considered to be within the scope of the disclosed subject matter. Therefore, the protection sought herein is exactly what is set forth in the appended claims.
Claims
1. A computer-implemented method in a wireless device, comprising: receiving a first sensor data set from one or more sensors of the wireless device; receiving a first set of radio measurements from a radio interface of the wireless device that is distinct from the one or more sensors; processing the first sensor data set and the first radio measurement set at a radio resource management (RRM) neural network of the wireless device to generate a first output representing a first RRM action; as well as The first RRM action is performed at the wireless device.
2. The method of claim 1, wherein the wireless device is a user equipment, and the method further comprises: providing at least one of the representations of sensor capabilities of the wireless device for the one or more sensors for receipt by an infrastructure component of a network infrastructure wirelessly connected to the wireless device; receiving, in response to providing the representation of sensor capabilities, a first neural network architecture configuration from the infrastructure component; as well as The first neural network architecture configuration is implemented at the RRM neural network.
3. The method of claim 2, wherein the representation of sensor capabilities comprises one or more fields of a UECapabilitiesInformation Radio Resource Control (RRC) message.
4. The method according to claim 2 or 3, further comprising: providing at least one of the first sensor data set or the first radio measurement set to the infrastructure component; receiving, in response to providing said at least one of said first sensor data set or said first radio measurement set, from said infrastructure component a second neural network architecture configuration, said second neural network architecture configuration representing a modification of said first neural network architecture configuration based on said at least one of said first sensor data set or said first radio measurement set; and The second neural network architecture configuration is implemented at the RRM neural network.
5. The method according to claim 2 or 3, further comprising: modifying the first neural network architecture configuration to generate a second neural network architecture configuration based on at least one of the first sensor data set or the first radio measurement set; as well as The second neural network architecture configuration is implemented at the RRM neural network.
6. The method according to any one of claims 1 to 5, further comprising: receiving an indication of an operational status of the wireless device; and Wherein processing comprises: processing the first sensor data set, the first radio measurement set, and the operating state at the RRM neural network to generate the first output.
7. The method of claim 6, wherein the operating state comprises a radio resource control (RRC) state of the wireless device.
8. The method of any one of claims 1 to 7, wherein the one or more sensors include at least one of: a position sensor; a gesture sensor; an accelerometer; a pressure sensor; or a proximity sensor.
9. The method of any one of claims 1 to 8, wherein the first set of radio measurements comprises at least one signal power measurement of a serving cell or a neighboring cell.
10. The method of any one of claims 1 to 9, wherein the RRM action comprises at least one of: performing RRM-related measurements by the wireless device; configuring characteristics of RRM-related measurements to be performed by the wireless device; or performing an RRM reporting process at the wireless device.
11. The method of claim 10, wherein the characteristic of the RRM-related measurement comprises at least one of: a frequency or a timing of the RRM-related measurement, or a frequency band or a channel of the RRM-related measurement.
12. The method according to any one of claims 1 to 11, wherein the RRM action comprises executing a conditional handover (CHO) decision or a conditional PSCell change (CPC) decision.
13. The method of claim 12, further comprising: In response to connecting to a cell due to a conditional handover decision: providing a representation of sensor capabilities of the wireless device for receipt by an infrastructure component of the cell; receiving, in response to providing said representation of sensor capabilities, a first neural network architecture configuration from said infrastructure component of said cell; as well as The second neural network architecture configuration of the RRM neural network is replaced with the first neural network architecture configuration.
14. The method of claim 13, further comprising: receiving a second sensor data set from the one or more sensors of the wireless device; receiving a second radio measurement set from the radio interface of the wireless device; processing, at the RRM neural network of the wireless device, the second sensor data set and the second radio measurement set to generate a second output representing a second RRM action; as well as The second RRM action is performed at the wireless device.
15. The method of any one of claims 1 to 14, wherein the wireless device is a user equipment.
16. The method of claim 1, wherein the wireless device is a base station.
17. The method of claim 16, further comprising: An RRM report is received from the user equipment.
18. A wireless device comprising: one or more sensors; a radio interface distinct from the one or more sensors; at least one processor coupled to the radio interface; as well as At least one memory, the at least one memory is coupled to the at least one processor, the at least one memory stores executable instructions, the executable instructions are configured to operate the at least one processor to perform the method according to any one of claims 1 to 17.
