Wireless Resource Management Using Machine Learning

Abstract

The wireless device (104) is configured to receive a set of sensor data (114) from one or more sensors of the wireless device, receive a set of wireless measurements (116) from the wireless interface of the wireless device, process the set of sensor data and the set of wireless measurements with the wireless resource management (RRM) neural network (112) of the wireless device to generate an output representing an RRM action (118), and then execute the RRM action. The wireless device is further configured to provide a representation of the sensor function (122) of the wireless device to the sensors for reception by infrastructure components (110) of a network infrastructure (102) wirelessly connected to the wireless device, and in response to providing the representation of the sensor function, receive a neural network architecture configuration (124) from the infrastructure components and be able to implement the neural network architecture configuration in the RRM neural network.

Description

Background Art

[0001] Cellular networks and other wireless networks often use radio resource management (RRM) to manage network capacity issues at a large scale (e.g., at the level of multiple users or multiple cells), rather than addressing point-to-point network capacity issues. Therefore, RRM utilizes various techniques to provide efficient throughput for the entire network while seeking efficient power consumption for some of the various networked components. These techniques can include, for example, techniques targeting power control, scheduling, cell search, cell reselection, handover, wireless link or connection monitoring, connection establishment / re-establishment, co-interference management, etc. User equipment (UE) within a wireless network often plays an important role in RRM by collecting various wireless measurement values and other system observations and reporting them to the network for use not only in the implementation of RRM actions on the network side but also in the implementation of specific procedures based on these measurements. In the conventional approach, a serving base station (BS) or other components of the network infrastructure instruct the UE to use a static or algorithmic configuration regarding the UE's role in the overall RRM process. For example, configuring the UE to use a static schedule for in-band, inter-band, or inter-RAT (radio access technology) scans of the serving cell or neighboring cells, or specifying a particular algorithm or fixed set of thresholds that the UE uses when making conditional handover decisions (CHO) for inter-cell handovers. This static approach to setting the UE's role in RRM often cannot take into account the specific situation of the UE and thus may result in non-optimal RRM behavior at the UE. This often leads to inefficient acquisition of RRM-related information at the UE and thus may impair the overall RRM decision-making process that utilizes such information. Also, a non-optimal RRM configuration at the UE may cause the UE to perform various RRM actions that are less relevant or untimely to the overall efficiency of the RRM process, resulting in unnecessary power consumption at the UE.

[0002] By referring to the accompanying drawings, the present disclosure can be better understood, and numerous features and advantages will become apparent to those skilled in the art. The use of the same reference symbols in different drawings indicates similar or identical items.

Brief Description of the Drawings

[0003]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Figure 12

[0004] The static configuration or algorithmic configuration of a UE for RRM operation, as seen in many conventional wireless systems, may result in low UE utilization efficiency for the overall RRM process, at the expense of excessive power consumption in the UE. Furthermore, configuring the UE to employ these static RRM configurations typically requires significant design, testing, and implementation efforts. As will be described below with reference to FIGS. 1-12, the static or algorithmic approach to RRM operation on the UE side can be replaced or complemented by a neural network (NN)-based approach that fuses sensor data from the UE's available sensors with wireless measurements made by the UE to arrive at one or more RRM actions used by the UE. Such actions can include, for example, setting the frequency or type of wireless measurements to be performed, determining specific conditional RRM actions (such as conditional handover (CHO) decisions or conditional primary cell changes (CPCs)), performing specific types of measurements or other RRM actions, and the like.

[0005] In at least one embodiment, a base station (BS) (or other infrastructure component) is provided access to a set of neural network architecture configurations trained using various training data sets reflecting various sensor functions, various sensor data, various RRM-related measurements performed by UEs or other wireless devices, etc. During or after establishment of a wireless connection between the BS and the UE, the UE supplies the BS with an expression of its sensor capabilities, or in particular, sensor capabilities related to the types of sensors 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 instructs the UE to implement the selected neural network architecture configuration in the UE's RRM neural network (e.g., a deep neural network (DNN)).

[0006] Configuring the UE in this way causes the UE to capture a series of one or more sensor measurements from a set of sensors and supply them to the RRM neural network as a set of sensor data. Similarly, the UE captures a series of one or more radio measurements (also called radio parameters or radio metrics), such as signal power measurements of the serving cell or one or both of one or more target cells or neighboring cells, via a radio interface separate from the set of sensors, and supplies the measurements to the RRM neural network as a set of radio measurements. Next, the RRM neural network uses these inputs to generate an output representing one or more RRM actions that the UE takes in response to these input data sets. The RRM actions specified by the output of the RRM neural network can include instructions to perform specific RRM actions, such as performing specified RRM-related measurements, or instructions to perform a CHO decision (e.g., to initiate a handover to a target cell or, conversely, to defer any handover), or instructions to refrain from executing a specific RRM action. The RRM actions can also, alternatively, include settings of aspects of the RRM actions, such as the frequency at which a specific RRM action is performed (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), or settings of parameters used in the RRM actions (e.g., the specification of a particular frequency band used for inter-frequency measurements).

[0007] Furthermore, the UE can report to the BS either or both of the sensor data or radio measurements used by the RRM neural network in determining the resulting RRM actions, and the BS can then use this feedback to retrain a model or copy of the RRM neural network and provide the updated version of the RRM neural network to the UE for later use. Alternatively, the UE can utilize this same data and other information to retrain or modify its local copy of the RRM neural network.

[0008] The incorporation or fusion of the UE's recent local sensor data with recent wireless measurements by an RRM neural network facilitates the local RRM process used by the UE to more easily adapt to the UE's current transmission environment or other current situations than what is provided via the UE's static / algorithmic RRM configuration. For example, as a result of training the implemented neural network architecture, sensor data indicating the proximity of an object, building, or other interference source may trigger the RRM neural network to instruct an RRM action that reduces (and thus conserves power) the frequency at which certain measurements are made using the UE's millimeter wave (mmW) antennas, or trigger the RRM neural network to use a lower threshold for signal power parameters measured via the mmW antennas before triggering a CHO, thereby potentially eliminating an unnecessary handover process. Further, by utilizing a neural network with an architecture configuration 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 substantial design, testing, and implementation efforts otherwise required for implementation of a conventional static / algorithmic RRM configuration.

[0009] The systems and techniques for neural network-based RRM detailed in this specification utilize coordination between infrastructure components and wireless devices of a wireless network. For ease of explanation, these systems and techniques are described in the exemplary context of a cellular network where a base station (BS) functions as the aforementioned infrastructure component and a UE functions as the aforementioned wireless device. However, these systems and techniques are not limited to this exemplary implementation. For example, in the same cellular context, the infrastructure component may be a server separate from the BS or other components, or may include multiple infrastructure components such as a cooperating server and the BS. As another example, in an implementation of a wireless local area network (WLAN), the aforementioned infrastructure component may be a wireless access point (AP), or other components “upstream” from the aforementioned wireless device connected to the wireless AP.

[0010] FIG. 1 shows an exemplary wireless communication network 100 that uses a neural network-based RRM scheme according to some embodiments. In the illustrated example, wireless communication network 100 is a cellular network that includes a network infrastructure 102 wirelessly connected to one or more wireless devices such as UE 104. Network infrastructure 102 includes a core network 106 coupled to one or more wide area networks (WANs) 108, or other packet data networks (PDNs) such as the Internet. 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 applicable RATs defined by one or more communication protocols or standards. Thus, each BS 110 operates as a wireless interface between one or more wireless devices and various networks and services provided by network infrastructure 102, such as packet switched (PS) data services, circuit switched (CS) services, and the like. Conventionally, signaling communication from BS 110 to UE 104 is referred to as "downlink" or "DL", while signaling communication from UE 104 to BS 110 is referred to as "uplink" or "UL".

[0011] Each BS110 can adopt any of various RATs, such as operating as a NodeB (or Base Transceiver Station (BTS)) of a Universal Mobile Telecommunications System (UMTS) RAT (also known as "3G"), an enhanced NodeB (eNodeB) of a 3rd Generation Partnership Project (3GPP (R)) Long Term Evolution (LTE) RAT, or a 5G NodeB ("gNB") of a 3GPP 5th Generation (5G) New Radio (NR) RAT. The UE104 then represents any of various electronic devices operable to communicate with the BS110 via an appropriate RAT, including, for example, a mobile phone, a cellular-enabled tablet computer or laptop computer, a desktop computer, a cellular-enabled video game system, a server, a cellular-enabled home appliance, a cellular-enabled vehicle communication system, a cellular-enabled smartwatch, or other wearable device.

[0012] For the following purposes, in FIG. 1, BS110-1 is currently wirelessly connected to UE104 and providing network services via the resulting wireless connection, and thus is referred to herein as the "serving" BS110-1 (also known as the "primary" BS in the art). In this example, BS110-2 is currently not providing network services to UE104 but is available for handover and initiation of network service provision, and thus is referred to as the "target" BS110-1 (or known as the "adjacent" BS or "secondary" BS in the art).

[0013] The exemplary wireless communication network 100 of FIG. 1 shows only a single UE 104 and two BSs 110 for ease of explanation, but in a real implementation, such a wireless system has a large number of BSs and UEs 104 operating in a nearby geographical area, and thus there are more opportunities for RF interference and contention to occur around 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. This RRM scheme involves the UE 104 by performing various measurements that can be used in various RRM processes and also executing some or all of these RRM processes. As described above, conventional RRM schemes rely on algorithms or static approaches that utilize the UE for RRM-related measurements and operations.

[0014] In contrast, in at least one embodiment, the wireless communication network 100 utilizes a neural network-based RRM scheme, in which the UE 104 is configured to adaptively and non-statically implement RRM actions using a trained RRM neural network (NN) 112 based on information regarding the current context of the UE 104. Specifically, in at least one embodiment, the RRM NN 112 receives local sensor data 114 and wireless measurement data 116 as separate inputs and provides an output indicating at least one RRM action 118 from at least these inputs.

[0015] The local sensor data 114 includes sensor data obtained from one or more sensors of the UE 104's sensor set (see FIG. 2), and can be provided as a single-point sampling of sensor data / sensor status from the participating sensors, or as a time series of sampling of sensor data from related sensors over a fixed or variable sliding time window. It should be recognized that the information captured by the sensors of the UE 104's sensor set can reflect the current operating state of the UE 104, both with respect to the physical local RF transmission environment of the UE 104 and with respect to the internal operating status of the UE 104. For example, the current situation involving the UE 104 that may impair the RF signaling between the UE 104 and the serving BS 110-1 (and the absence of such a situation) can be detected (or otherwise represented) from sensor data generated by an object detection sensor such as a radar, lidar, or imaging device (e.g., an imaging camera) that generates sensor data reflecting the presence or absence of interfering objects in the line-of-sight (LOS) propagation path between the serving BS 110-1 and the UE 104. Similarly, positioning data from a global positioning system (GPS) sensor, gyroscope, accelerometer, or camera-based visual odor sensor system, etc., can, for example, find one of the positions and / or movements of the UE 104 with respect to the serving BS 110-1, and thus represent the current RF signal propagation environment. As another example, an optical sensor, image sensor, or touch sensor can provide sensor data indicating the posture of the UE 104 with respect to the user's body, and thereby can function as an indicator of the RF signal propagation environment in which the UE 104 is likely to be present. Regarding the current internal operating environment of the UE 104, a battery power sensor indicates the remaining amount of battery power, and thus can indicate the function of the UE to continue to perform various tasks without impairing the overall operation of the UE 104. On the other hand, a thermal sensor of the UE 104 can indicate whether the UE 104 is close to thermal limits, and thus can indicate the extent to which the UE 104 can continue to perform various RRM actions that can contribute to the thermal output of the UE 104.Therefore, the input of the local sensor data 114 to the trained RRM NN112 can facilitate the decision-making process of the RRM NN112 in order to adapt the resulting input to reflect the current operating environment of the UE104.

[0016] Regarding the radio measurement data 116, this information can likewise be provided either as a single-point measurement sample, as time-series measurement samples over a fixed or variable sliding window, or as a combination thereof. The radio measurement data 116 reflects measurements of various parameters or metrics of either or both received RF signaling and transmitted RF signaling. Examples of such radio measurements can 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, thermal interference (IoT) ratio measurements.

[0017] As described above, the output of the RRM NN 112 in response to the input of the local sensor data 114 and the wireless measurement data 116 represents one or more RRM actions 118 performed by the UE 104. The RRM actions 118 can be classified into at least one of the following categories: (1) an instruction to perform or refrain from performing an RRM process at the UE 104, (2) an instruction to change or set parameters of an RRM process executed at the UE, or (3) a determination of a decision for a conditional RRM process of the UE 104. The first category of RRM actions represents the RRM actions 118 that respond either by the UE 104 performing an RRM process or refraining from performing an RRM process that it would otherwise perform. For example, based on local sensor data 114 indicating that the UE 104 is stationary and in a location with relatively few signal obstructions, the RRM NN 112 may output an RRM action 118 instructing the UE 104 to perform a single RRM measurement (e.g., a scan of a particular frequency band) and report the result to the serving BS 110-1. Conversely, based on local sensor data 114 indicating that the UE 104 is moving at a very high speed or that the UE 104 is indoors and surrounded by substantial RF-blocking objects, the RRM NN 112 may output an RRM action 118 instructing the UE 104 to skip the next scheduled RRM measurement (thus potentially conserving the UE's power and computational resources) in reflection of the potential futility of attempting the scheduled RRM measurement in the UE's current operating environment.

[0018] The second category of RRM actions represents RRM action 118 in which UE104 responds by changing an RRM process that UE104 has already scheduled or has been instructed to execute. For example, UE104 may be configured to perform a particular RRM measurement according to a particular timing, and RRM action 118 can represent a change to this timing. By way of illustration, the 3GPP 5th generation new radio (5G NR) specification provides for comparing the relative signal strength of the serving cell (e.g., provided by serving BS110-1) with that of an adjacent cell (e.g., provided by target BS110-2) through a cell signal measurement process using a synchronization signal (SS) / physical broadcast channel (PBCH) block, or SSB, and the timing, or periodicity, of each successive measurement using 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 provide an output representing an RRM action 118 that triggers the trained RRM NN112 to instruct UE104 to adjust the SMTC window, thereby increasing or decreasing (depending on the adjustment) the period of SSB-based cell measurements for the serving cell and adjacent cells.

[0019] The third category of RRM actions represents RRM action 118 that represents decisions made on one or more conditional RRM processes that can be performed by UE104. For example, the 3GPP 5G NR Release 16 specification defines a conditional handover (CHO) RRM process, in which the handover command is sent to the UE along with one or more conditions monitored by the UE in relation to 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 then initiates the previously received handover. Thus, for this example, RRM action 118 can mimic the CHO process without requiring that specific handover conditions be defined statically or algorithmically. Rather, since RRM NN112 is trained using a training data set with similar training sensor data and similar training radio measurement data to make decisions such as CHO, the RRM action 118 output by RRM NN112 in the field can include a CHO decision to either initiate a handover or refrain from initiating a handover based on the current network and UE context reflected in the input local sensor data 114 and radio measurement data 116. Other such conditional RRM process decisions that can be specified as RRM action 118 can include, for example, conditional primary cell change (CPC), where a change of the primary cell (Pcell) is performed by the UE when one or more specified conditions are met.

[0020] The RRM action 118 may also be a combination of two or more of the above three categories. That is, the RRM action 118 may be a hybrid RRM action. By way of example, the output of the RRM NN112 may represent an RRM action 118 that includes activating a previously deactivated RRM measurement process and setting one or more parameters for that RRM measurement process (e.g., activating an inter-RAT scan and setting the frequencies to be scanned).

[0021] RRM NN112 may also receive other inputs that assist in generating an output representing RRM action 118. For example, the 5G NR specification states that certain RRM measurements are specific to the combination of the UE's RRC state (IDLE, INACTIVE, CONNECTED) and whether the UE's 5G NR RAT is in stand-alone (SA) mode or non-stand-alone (NSA) mode. For example, if the UE is in the RRC IDLE state and in SA mode, the UE is permitted to perform cell selection or reselection, but if the UE is in NSA mode while in the RRC IDLE state, the 5G NR specification results in the intention that the UE refrain from cell selection or reselection. Similarly, if the UE is RRC CONNECTED, 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 results in the intention that the UE perform handover only when in SA mode while in this RRC state. Thus, UE104 can provide operating state data 120 as an input to RRM NN112, and this operating state data 120 represents specific operating state parameters of UE104 and / or UE104's RAT, such as the RRC state and SA / NSA mode of UE104's 5G NR RAT. In this example, RRM NN112 can be trained using training data to avoid providing an output that triggers an RRM action 118 representing an RRM process that conflicts with the RRC state and SA / NSA mode restrictions for a particular RRM measurement value.

[0022] For the UE 104 to utilize the RRM NN 112 for RRM actions, the UE 104 is first configured with the RRM NN 112. Further, in some embodiments, the architectural configuration underlying the RRM NN 112 can be updated either via local updates or remote updates via reporting of relevant information by the UE 104. View 121 of FIG. 1 shows an overall overview of the process of initial setup and update of the UE 104 using the RRM NN 112. In at least one embodiment, the RRM NN 112 used by the UE 104 is trained using training data similar to the data inputs expected to be provided by the UE 104. As described above, the data inputs to the RRM NN 112 include local sensor data 114 generated by a specific set of sensors available to the UE 104 and radio measurement data 116 representing specific types of radio measurements that can be performed by the UE 104. Thus, for the RRM NN 112 employed by a UE 104 trained using training data from similar sensors and for the NN architectural configuration trained for similar radio measurement types, it is advantageous to select. For example, adopting an NN architectural configuration of the RRM NN 112 trained using training data significantly influenced by radar sensor data will typically provide less effective results in a UE without a radar sensor. For that purpose, the serving BS 110-1 stores or otherwise has access to a set of various NN architectural configurations, each trained according to a specific 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 the serving BS 110-1 with an expression of its capabilities (UE capabilities message 122). The UE capabilities message 122 includes expressions of various capabilities of the UE 104, including one or both of an expression of the sensor capabilities of the UE 104 or an expression of the radio measurement capabilities of the UE 104.The representation of the sensor function of UE104 may include, for example, the representation of the type of sensors included in the sensor set of UE104, the functions or other parameters of each part or all of the sensors. For example, during the connection process, the serving BS110-1 may send a UE Capabilities Enquiry Radio Resource Control (RRC) message to UE104, and UE104 may use one or more fields of the UECapabilitiesInformation RRC message, which includes data or other information representing the type, quantity, and parameters of the sensors in the sensor set of UE104, to respond with the UECapabilitiesInformation RRC message.

[0023] As shown in FIG. 1, the serving BS110-1 (or another infrastructure component such as a server in the core network 106) then selects an appropriate UE104 RRM NN architecture configuration 124 for the indicated sensor function and / or radio measurement function using the sensors and / or radio measurement functions of UE104 shown in the UE function message 122. This selection can be performed using any of a variety of techniques, such as using one or more look-up tables (LUTs) indexed by the indicated functions through a weighted evaluation that compares the functions indicated with the sensor type / parameters and / or radio measurement types used to train corresponding candidate RRM NN architecture configurations, etc. Next, the serving BS110-1 signals to UE104 to use the selected RRM NN architecture configuration 124. For example, in some embodiments, the serving BS110-1 may send one or more messages to UE104 using data representing the actual RRM NN architecture configuration 124. In other embodiments, UE104 may pre-provide a set of stored RRM NN architecture configurations, and then the serving BS110-1 may send a message to UE104 that includes an index or other identifier of the specific RRM NN architecture configuration utilized from its stored set.

[0024] 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 initiates the operation of the RRM NN 112 as so configured. During the operation and in line with many RRM schemes, the UE 104 may provide and return various RRM reports 126 to the serving BS 110-1. This RRM report 126 can include reports of RRM measurements made by the UE 104, including those directed or configured by the RRM NN 112. The RRM report 126 can further include reports of decisions or settings reflected in the RRM actions 118 output by the RRM NN 112. For example, the UE 104 can report, via the RRM report 126, that as a result of the RRM action 118 output by the RRM NN 112, the UE has changed the frequency at which it performs inter-RAT or intra-RAT scans. Next, the serving BS 110-1 can take local actions in response to this RRM report 126, or transfer the RRM report 126 to the core network 106 for further consideration by the network infrastructure 102.

[0025] In some embodiments, the RRM NN architecture configuration used by RRM NN112 can be dynamically updated based on usage. In some embodiments, the NN update to RRM NN112 is provided by serving BS110-1 (or other infrastructure components). In this approach, UE104 can provide a copy 128 of local sensor data 114 and / or wireless measurement data 116 to serving BS110-1, and serving BS110-1 or other components can use this data, e.g., together with RRM reports 126, to retrain the NN architecture configuration 124 utilized by UE104 to generate an updated NN architecture configuration and provide this updated NN architecture configuration to UE104 for implementation as NN update 130. In other embodiments, the NN update is performed by UE104 itself. For example, UE104 can be configured to periodically provide a copy of the current state, or "snapshot," of the NN architecture configuration of RRM NN112 to serving BS110-1 as NN update 130. Next, serving BS110-1 can update its local or accessible copy of RRM NN architecture configuration 124 to reflect the supplied NN update 130, and further, redistribute this updated NN architecture configuration to other UEs that employ the same NN architecture configuration 124.

[0026] FIG. 2 shows an exemplary hardware configuration of UE104 (as a representative wireless) according to some embodiments. The shown hardware configuration represents the processing and communication components most directly related to the neural network-based processes described herein, and it should be noted that certain components well-known to be frequently implemented in such electronic devices as displays, non-sensor peripherals, external power supplies, etc. are omitted.

[0027] In the illustrated configuration, UE104 includes a RF front end 202 having one or more antennas 203 and a wireless interface 204 including one or more modems supporting one or more RATs. The RF front end 202 effectively operates as a physical (PHY) transceiver interface to conduct and process signaling between one or more processors 206 of UE104 and the antennas 203 to facilitate various types of wireless communication. The antennas 203 can be arranged in one or more arrays of a plurality of antennas configured to be similar or different from each other and can be tuned to one or more frequency bands associated with the corresponding RAT. The one or more processors 206 can 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). By way of example, the processor 206 can include an application processor (AP) utilized by UE104 to execute an operating system and various user-level software applications, as well as one or more processors utilized by a baseband processor of the modem or wireless interface 204.

[0028] UE104 further includes one or more computer-readable media 208 including any of various media used by an electronic device 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. For ease and brevity of explanation, the computer-readable media 208 is referred to herein as "memory 208" considering that system memory or other memory is frequently used to store data and instructions for execution by the processor 206, but references to "memory 208" are understood to equally apply to other types of storage media unless otherwise noted.

[0029] In at least one embodiment, UE 104 further includes a plurality of sensors that are separate from the wireless interface 204 and are collectively referred to herein as sensor set 210, at least some of which are utilized in the neural network-based scheme described herein. Generally, the sensors of sensor set 210 are sensors that sense some aspect of the external RF environment of UE 104 (i.e., sensors that can at least somewhat affect, or sense parameters that reflect, the RF propagation path of UE 104, or the RF transmission / reception performance), sensors that sense some aspects of the current operating status of UE 104 such as battery status, thermal status, operating mode, screen status, etc. Thus, the sensors of sensor set 210 can include one or more sensors 212 for object detection, such as radar sensors, lidar sensors, image sensors, structured light-based depth sensors, proximity sensors, etc. Sensor set 210 can also include one or more sensors 214 for determining the position, orientation, or velocity / speed of UE 104, such as satellite positioning sensors such as GPS sensors, global navigation satellite system (GNSS) sensors, inertial measurement unit (IMU) sensors, visual odor sensors, accelerometers, gyroscopes, barometers, altimeters, tilt sensors or other inclinometers, ultra-wideband (UWB)-based sensors, etc. Other examples of the types of sensors of sensor set 210 can include sensors 216 for determining the current operating status of UE 104, such as battery level sensors, temperature sensors, screen mode sensors, etc. Although not shown, it should be appreciated that UE 104 can further include one or more batteries or other portable power sources, one or more user interface (UI) components such as touchscreens, user-operable input / output devices (e.g., "buttons" or keyboards), or other touch / contact sensors, microphones, or other audio sensors for capturing audio content, etc.

[0030] One or more memories 208 of the UE 104 are used to store one or more sets of executable software instructions and associated data for operating one or more processors 206 and other components of the UE 104, perform various functions described herein and attributed to the UE 104. Examples of sets of executable software instructions include, for example, an operating system (OS) and various drivers (not shown), as well as various software applications. The set of executable software instructions further includes one or more of a neural network management module 218, a function management module 220, or an RRM module 222. The neural network management module 218 implements one or more neural networks in the UE 104, as will be described in detail below. The function management module 220 determines various functions of the UE 104 that may be related to neural network architecture configuration or selection, reports such functions to the serving BS 110-1 (e.g., in one or more UE Capabilities Information RRC messages), and monitors the UE 104 for changes to such functions, including changes to RF and processing capabilities, accessory availability, or function changes, and manages the reporting of such functions and function changes to the serving BS 110-1. The RRM module 222 operates to cause the UE 104 to execute an RRM process that includes RRM measurements and reporting, as well as RRM action(s) 118 specified by the output of the RRM NN 112.

[0031] To facilitate the operations of the UE 104 described herein, one or more memories 208 of the UE 104 can further store data related to these operations. This data can include, for example, one or more neural network architecture configurations 224 (RRM NN architecture configuration 124, embodiment of FIG. 1), as well as device data 226. The device data 226 can represent, for example, user data, multimedia data, beamforming codebooks, and software application configuration information. The device data 226 can further include functional information of the UE 104, such as sensor functional information regarding one or more sensors of the sensor set 210 including the presence or absence of a particular sensor or sensor type, and for the sensors that exist, one or more representations of the corresponding type and function, such as the range and resolution of a LIDAR or radar sensor, the image resolution and color depth of an imaging camera, etc. The functional information can further include information regarding, for example, the function or status of the battery, the function or status of the wireless interface 204 and antenna(s) 203 (e.g., frequency function, wireless measurement function, etc.).

[0032] Each neural network architecture configuration 224 includes one or more data structures containing data and other information representing corresponding architectures and / or parameter configurations used by neural network management module 218 to form the corresponding RRM neural network 112 of UE 104. The information included in neural network architecture configuration 224 includes, for example, fully connected layer neural network architecture, convolutional layer neural network architecture, recurrent neural network layer, number of connected hidden neural network layers, input layer architecture, output layer architecture, number of nodes used by the neural network, coefficients (such as weights and biases) used by the neural network, kernel parameters, number of filters used by the neural network, stride / pooling configuration used by the neural network, activation function for each neural network layer, interconnections between neural network layers, neural network layers to skip, etc. Thus, neural network architecture configuration 224 includes any combination of neural network formation components (e.g., architectures and / or parameter configurations) that can be used to create a neural network architecture configuration (e.g., a combination of one or more neural network formation components) that defines and / or forms a DNN or other neural network.

[0033] FIG. 3 shows an exemplary hardware configuration of BS110 (as a representative infrastructure component) according to some embodiments. The hardware configuration shown represents the processing and communication components most directly related to the neural network-based processes described herein, and it should be noted that certain components well-known to be frequently implemented in such electronic devices as displays, non-sensor peripherals, external power supplies, etc. are omitted. Further, the illustrated figure represents an embodiment of BS110 as a single network node (e.g., a 5G NR Node B, i.e., a "gNB"), but it should be noted that the functions, and thus the hardware components of BS110, may instead be distributed across multiple network nodes or devices and distributed in a manner that performs the functions described herein.

[0034] In the illustrated configuration, BS110 includes an RF front end 302 with one or more antennas 303 and a wireless interface 304 with one or more modems that support one or more RATs, and operates as a PHY transceiver interface that implements and processes signaling between one or more processors 306 of BS110 and the antennas 303 to facilitate various types of wireless communication. The antennas 303 can be arranged in one or more arrays of a plurality of antennas configured to be similar or different from each other and can be tuned to one or more frequency bands associated with the corresponding RAT. The one or more processors 306 can include, for example, one or more CPUs, GPUs, TPUs, or other ASICs. BS110 further includes one or more computer-readable media 308 including any of various media used by an electronic device to store data and / or executable instructions, such as RAM, ROM, cache, flash memory, SSD, or other mass storage devices. Similar to the memory 208 of UE104, for ease and brevity of explanation, the computer-readable media 308 is referred to herein as "memory 308" considering that system memory or other memory is frequently used to store data and instructions for execution by the processor 306, but references to "memory 308" are understood to equally apply to other types of storage media unless otherwise noted.

[0035] One or more memories 308 of BS110 are used to store one or more sets of executable software instructions and related data for operating one or more processors 306 of BS110 and other components, execute various functions described herein and attributed to BS110. Sets of executable software instructions include, for example, an OS and various drivers (not shown), as well as various software applications. The set of executable software instructions further includes one or more of neural network management module 310, training module 312, and RRM module 314. One or more memories 308 further store a set 322 of one or more candidate neural network architecture configurations 324 (embodiments of RRM NN architecture configuration 124, FIG. 1), as well as various information such as various BS data 326. BS data 326 represents, for example, a beamforming codebook, software application configuration information, RRM scheme information, etc. Neural network architecture configuration 324 represents a trained neural network architecture configuration that can be used in RRM NN112 of UE014 or other UEs. Thus, similar to the neural network architecture configuration 224 of FIG. 2, each candidate neural network architecture configuration 324 includes one or more data structures including data and other information representing the corresponding architecture and / or parameter configuration that the neural network management module 218 of a UE such as UE104 uses to form the corresponding RRM NN112 in the UE. Neural network management module 310 manages the training, retraining, selection, and distribution of neural network architecture configuration 324 to UE104 and other UEs. Training module 312 performs 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. RRM module 314 operates to execute various RRM processes performed by BS110.

[0036] Figure 4 shows an exemplary machine learning (ML) module 400 for implementing a neural network according to some embodiments. As described herein, UE 104 implements one or more DNNs or other neural networks as RRM NN 112 to configure or control the RRM process in UE 104. In this regard, BS 110 trains, retrains, or otherwise updates a copy of one or more DNNs or other neural networks used as RRM NN 112 in one or more UEs. Thus, ML module 400 represents an exemplary module for implementing one or more of these neural networks.

[0037] In the illustrated example, 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 various ways, such as a partial connection configuration where a first subset of the nodes in the first layer is connected to a second subset of the nodes in the second layer, or a full connection configuration where each node in the first layer is connected to each node in the second layer. Neurons process input data to generate 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 a desired category. A perceptron performs linear classification, such as binary classification, on the input data. Nodes, whether neurons or perceptrons, can use various algorithms to generate output information based on adaptive learning. Using DNN 402, ML module 400 performs various different types of analysis, including simple linear regression, multiple linear regression, logistic regression, stepwise regression, binary classification, multi-class classification, multivariate adaptive regression splines, local estimated scatterplot smoothing, and the like.

[0038] In some embodiments, the ML module 400 adaptively learns 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 actually learns how to map the input training data to the desired output, i.e., the RRM action. Specifically, during the training procedure, the ML module 400 uses labeled data or known data as input to the DNN 402. The DNN 402 analyzes the input using nodes and generates the corresponding output. The ML module 400 compares the corresponding output with the true data and adapts the algorithm implemented by the nodes to improve the accuracy of the output data. Thereafter, the DNN 402 applies the adapted algorithm to 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 correlate unknown inputs to statistically likely outputs within a threshold range or threshold. Thereby, the ML module 400 can receive complex inputs and identify the corresponding outputs. As described above, some embodiments train the ML module 400 regarding the characteristics of RRM decisions based on input sensor data and radio measurement data. Thereby, the trained ML module 400 is enabled to receive a set of sensor data (either sensor data from a single time slice or sensor data over a sequence of time slices), and a set of radio measurement data (either from a single time slice or over a sequence of time slices), and generate an output representing the RRM action(s) to be performed from those inputs.

[0039] In the described example, DNN 402 includes an input layer 404, an output layer 406, and one or more hidden layers 408 disposed between the input layer 404 and the output layer 406. Each layer can have any number of nodes, and the number of nodes between layers can be the same or different. That is, the input layer 404 can have the same number and / or a different number of nodes as the output layer 406, the 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.

[0040] Node 410 corresponds to one of several nodes included in the input layer 404, and the nodes perform separate independent computations. As will be further described, the nodes receive input data, process the input data using one or more algorithms to generate output data. Typically, the algorithms include weights and / or coefficients that change based on adaptive learning. Thus, the weights and / or coefficients reflect the information learned by the neural network. Each node can, in some cases, determine whether to pass the processed input data to one or more of the next nodes. By way of example, after processing the input data, node 410 can determine whether to pass the processed input data to one or both of nodes 412 and 414 in the hidden layer 408. Alternatively or additionally, node 410 passes the processed input data to nodes based on the layer connection architecture. This process can be repeated across multiple layers until DNN 402 generates an output using a node (e.g., node 416) in the output layer 406.

[0041] A neural network can also use various architectures to determine which nodes within the neural network to connect, how to advance and / or hold data within the neural network, what weights and coefficients to use to process input data, how to process the data, and so on. These various elements collectively describe a neural network architecture configuration, such as the neural network architecture configuration briefly described above. By way of example, recurrent neural networks such as long short-term memory (LSTM) neural networks form a cycle between node connections to hold information from previous parts of the input data sequence. Next, the recurrent neural network uses the held information for subsequent parts of the input data sequence. As another example, a feedforward neural network transfers information and advances connections without forming a cycle to hold information. Although described in the context of node connections, it should be recognized that the neural network architecture configuration can include various parameter configurations that affect how a DNN402 or other neural network processes input data.

[0042] The neural network architecture configuration of a neural network can be characterized by various architectures and / or parameter configurations. To illustrate, consider an example where DNN402 implements a convolutional neural network (CNN). Generally, a convolutional neural network corresponds to a type of DNN where layers use a convolution operation to process data and filter input data. Thus, a CNN architecture configuration can be characterized by, for example, pooling parameter(s), kernel parameter(s), weights, and / or layer parameter(s).

[0043] Pooling parameters correspond to parameters that specify a pooling layer within a convolutional neural network that reduces the dimensionality of input data. By way of example, a pooling layer can couple the output of nodes in a first layer to the input of nodes in a second layer. Alternatively or additionally, the pooling parameters specify how and where in a layer of data processing the neural network pools data. For example, a pooling parameter indicating "max pooling" configures the neural network to pool by selecting the maximum value from a group of data produced by nodes in a first layer and use that maximum value as an input to a single node in a second layer. A pooling parameter indicating "average pooling" configures the neural network to produce an average value from a group of data produced by nodes in a first layer and use that average value as an input to a single node in a second layer.

[0044] The kernel parameters indicate the filter size (e.g., width and height) used for processing the input data. Alternatively or additionally, the kernel parameters specify the type of kernel method used for filtering and processing the input data. A support vector machine corresponds to a kernel method that uses, for example, regression analysis to identify and / or classify data. Other types of kernel methods include Gaussian processes, canonical correlation analysis, and spectral clustering methods. Thus, the kernel parameters can indicate the filter size and / or the type of kernel method applied to the neural network. The weight parameters specify the weights and biases used by the algorithm within the node to classify the input data. In some embodiments, the weights and biases are settings of learned parameters, such as settings of parameters generated from training data. The layer parameters specify the type of connection of the layers and / or the type of layer, such as a fully connected layer type indicating that all nodes of a first layer (e.g., output layer 406) are connected to all nodes of a second layer (e.g., hidden layer 408), a type of partial connection layer indicating which nodes of the first layer are disconnected from the second layer, and a type of activation layer indicating which filters and / or layers are activated within the neural network. Alternatively or additionally, the layer parameters specify the type of node layer, such as a normalization layer type, a convolutional layer type, and a pooling layer type.

[0045] Although described in the context of pooling parameters, kernel parameters, weight parameters, and layer parameters, it will be appreciated that other parameter settings can be used to form a DNN that is consistent with the guidelines provided herein. Thus, the neural network architecture configuration can include any suitable type of configuration parameter applicable to the DNN that affects the way the DNN processes the input data to generate the output data.

[0046] In some embodiments, the configuration of the ML module 400 is further based on the sensor function and / or the radio measurement function of the UE implementing the ML module 400. Thus, the architectural configuration of the ML module 400 may also be based on the capabilities of the UE implementing the ML module 400. For example, the UE 104 may have an important imaging function, whereby the ML module 400 of the UE 104 may be trained based on image data as input, thereby, for example, making it easier for the ML module 400 to generate RRM actions that are quite suitable for an RF transmission environment that depends on the presence or absence of an object or other interference source represented in such image data. However, in the case of a UE that does not have an imaging function, using an ML module 400 configured via training based extensively on training image data will generally not be very suitable for generating effective RRM actions based on input sensor data without imaging data. Thus, 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 function, radio measurement function, or both. For example, the device may have access to one or more neural network architectural configurations for use when a radar sensor is available for use in the device and a different set of one or more neural network architectural configurations for use when the radar sensor is not available, although the radar sensor is available.

[0047] In some embodiments, the UE 104 implementing the ML module 400 locally stores some or all of a set of candidate neural network architecture configurations that can be used by the ML module 400. For example, the candidate neural network architecture configuration 224 can be indexed at the UE 104 by a look-up table (LUT) or other data structure that receives, as input, one or more sensor function parameters or one or more parameters such as a radio measurement function, and outputs an identifier associated with the corresponding locally stored candidate neural network architecture configuration suitable for operation considering the input parameter(s). In other embodiments, the UE 104 provides an expression of its capabilities, and the BS 110 or other infrastructure components select 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 components use 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 files storing the training data, and obtain real-world communication data. Next, the neural network management module 310 extracts and stores various learned neural network architecture configurations for subsequent use. Some embodiments store input characteristics with each neural network architecture configuration, whereby the input characteristics describe various sensor characteristics and / or radio measurement characteristics.

[0048] As described above, a wireless device such as UE104 can be configured to determine one or more RRM actions using one or more RRM DNNs, where each RRM DNN complements or replaces one or more functions that were conventionally implemented by one or more hard codings 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 to modify, or otherwise adapt, its operation to take into account the current RF signal propagation environment or the operating state of the wireless device as reflected in the sensor data. For this purpose, FIG. 5 shows an exemplary operating environment 500 for DNN implementation in UE104. In the illustrated example, the neural network management module 218 of UE104 implements an RRM processing module 502 for a general RRM decision-making process. Further, to illustrate a particular use case of neural network-based RRM management in a UE, an exemplary implementation of the neural network management module 218 is shown as additionally implementing a CHO processing module 504, which in this case is a particular exemplary implementation of the RRM processing module for a particular purpose of conditional RRM decision-making that is a CHO decision. Thus, the CHO processing module 504 can be understood as a separate RRM processing module dedicated to CHO decision-making, or alternatively as that portion of the RRM processing module 502 that results in RRM actions that affect CHO decision-making.

[0049] As an overview of the general operation of UE104 with respect to the operating environment 500 depicted in FIG. 5, the neural network management module 218 supplies the neural network architecture configuration 224 (FIG. 2) to each of the RRM processing module 502 and the CHO processing module 504 based on the sensor functions of the sensors of the sensor set 210 and / or the radio measurement functions of the radio interface 204. According to each schedule, the sensor set 210 provides a most recent local sensor data set 514 (an embodiment of the local sensor data 114, FIG. 1) obtained from one or more sensors of the sensor set 210, and the radio interface 204 provides a most recent radio measurement data set 516 (an embodiment of the radio measurement data 116, FIG. 1) to the neural network management module 218, which then provides these data sets as inputs to each of the RRM processing module 502 and the CHO processing module 504. Other inputs such as the current RRC state (IDLE, INACTIVE, CONNECTED) or RAT mode (e.g., standalone (SA) or non-standalone (NSA) of 5G NR RAT) may also be provided to the processing modules 502 and 504.

[0050] The RRM processing module 502 utilizes these inputs to generate at least one RRM action 518 (an embodiment of RRM action 118), which is provided to the RRM module 222 for implementation. As described above, the RRM action 518 can represent an RRM process to be executed, such as performing RRM measurements and reporting them to the serving BS 110-1, an RRM process to be skipped, such as skipping a scheduled cell reselection, a modification to one or more RRM processes, such as resetting the intra-RAT measurement period, or a combination thereof. Thus, the RRM module 222 controls one or more parameters of the radio IF 204 and other controls to implement the RRM action 518. For example, the periodicity of inter-RAT scanning can be implemented as a value stored in a register of the radio IF 204, and the RRM module 222 can implement an RRM action 518 that attempts to change this periodicity by overwriting the register with a new value. As another example, the radio interface 204 can provide an application programming interface (API) or other interface, and the RRM module 222 can trigger the radio interface 204 to perform the RRM measurements represented by the RRM action 518 by triggering the RRM measurements via the API or other interface.

[0051] Concurrently (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 similarly provided to the RRM module 222 for implementation. The CHO action 520 can include, for example, a CHO decision to initiate a handover based on handover conditions interpreted through the local sensor dataset 514 and the radio measurement dataset 516, or a decision to defer the initiation of a handover based on handover conditions interpreted through those inputs. Additionally or alternatively, the CHO action 520 can include modifications to the static CHO configuration. For example, the BS 110 may send a message to the UE 104 to implement a CHO where a handover is initiated when a specified set of monitoring conditions is met, and the CHO action 520 can include modifications to the thresholds or other triggers represented by this set of monitored conditions. For example, the BS 110 can specify that a handover is initiated when the measured RSRP is below a specified threshold. However, the CHO processing module 504 may have been trained using training data reflecting a situation where the UE has a low measured RSRP but a high speed, and the measured RSRP may change rapidly if the position of the UE relative to the serving cell changes rapidly. Thus, the results of the training can cause the CHO processing module 504 to be configured to function to modify the RSRP threshold before a handover is triggered if the input local sensor dataset 514 includes sensor data indicating that the UE 104 is moving rapidly and the radio measurement dataset 516 includes data indicating that the measured RSRP bounces near 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 starting a handover, holding off on starting a handover, or modifying one or more monitored parameters of the conditions being monitored to determine whether the radio interface 204 should start a handover.

[0052] FIG. 6 shows an exemplary operating environment 600 of a BS 110 (e.g., serving BS 110-1) for supporting neural network-based RRM at a UE 104 according to at least one embodiment. As described above, for neural network-based RRM processes, the BS 110, in some embodiments, operates to provide a neural network architecture configuration suitable for a particular sensor function and / or radio measurement function of the UE 104 to the UE 104, to train / re-train the neural network architecture configuration, or both in combination. For selection and provision of an initial DNN architecture configuration to the UE 104, the BS 110 either implements a neural network architecture configuration data store 602 or has networked access to the data store 602 via a server or other infrastructure components of the network infrastructure 102 (FIG. 1). The data store 602 stores one or more candidate neural network architecture configurations 324 (FIG. 3) that can be indexed or otherwise accessed based on corresponding sensor function attributes, radio measurement attributes, etc. Thus, when the UE 104 supplies UE function information 604 indicating one or both of its sensor function or radio measurement function to the BS 110, the RF front end 302 of the BS 110 passes the UE function information 604 to the RRM module 314 and uses the UE function information 604 to select, from the data store 602, a neural network architecture configuration 324 trained using training data having the same or similar sensor function and / or the same or similar radio measurement function. This selected neural network architecture configuration 324, or its identifier or other representation, is then transmitted to the UE 104 for implementation.

[0053] Further, in response to providing the neural network architecture configuration 324 to UE 104, in some embodiments, the neural network management module 310 of BS 110 may instruct the training module 312 to instantiate a training processing module 606 that implements the DNN or other neural network first configured with the selected neural network architecture configuration 324 provided to UE 104. That is, the training processing module 606 is actually a copy of the RRM NN 112 operating simultaneously at BS 110. Thereafter, 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 UE 104, which is provided in response to UE 104 performing various RRM processes as a result of the RRM actions 118 generated by the RRM NN 112 based on the local sensor data set 514 and radio measurement data set 516 received as inputs. Thus, this feedback from UE 104 represents the input provided to the RRM NN 112. Therefore, the training module 312 can use this feedback as 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 trigger event, a modified neural network architecture configuration 624 of the training processing module 606 is extracted and provided to the RRM module 314, which then stores the representation of the modified neural network architecture configuration 624 in the data store 602 and / or transmits the representation of the modified neural network architecture configuration 624 to UE 104 for use when UE 104 updates the RRM NN 112.

[0054] Referring now to FIGS. 7 and 8, a method 700 for implementing an RRM scheme in a wireless communication network 100 that utilizes a neural network-based RRM process at a UE 104 is described in accordance with at least one embodiment. The order of operations described with reference to method 700 is for illustrative purposes only, and operations may be performed in a different order, and furthermore, one or more operations may be omitted, or one or more additional operations may be included in the method as shown. For ease of understanding, method 700 is described with reference to an exemplary embodiment of method 700 reflected in the ladder diagram 900 of FIG. 9.

[0055] In an embodiment, the UE 104 may have any of various combinations of sensor functions and wireless measurement functions. For example, the optical sensor of the UE may be on at times, but at other times, the UE turns off its optical sensor to conserve power. As another example, some UEs 104 may have a satellite-based positioning sensor, while others may not. As yet another example, one UE 104 may have a radar or lidar function but may not have a camera function, while another UE 104 may have a camera function but may not have a radar or lidar function. Since the RRM neural network(s) implemented at the UE 104 utilize sensor data and / or wireless measurements as inputs to instruct their operations, in many cases, the particular neural network architecture configuration implemented at the UE 104 is based on the particular sensors available to provide sensor data as inputs and / or on the particular wireless measurements that the UE is capable of performing. That is, the particular neural network architecture configuration implemented in the RRM NN 112 reflects one or both of the type of sensor currently providing an input to the RRM NN 112 in combination with, and the type and parameters of the wireless measurements currently being provided as an input to the RRM NN 112.

[0056] Accordingly, method 700 begins, at block 702, to determine the sensors and radio measurement capabilities of one or more test UEs, which may include UE 104 or may utilize UEs other than UE 104. It should be understood that a "test UE" in this sense may not actually be a physical UE, but rather may be a software simulation of the behavior of a test UE for the purpose of neural network training. Thus, references to a "test UE" should be understood to include references to such simulations. With respect to 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 particular sensor configuration and / or radio measurement configuration to train a candidate neural network architecture of the RRM NN 112 of the test UE. In some embodiments, the training module can attempt to train all permutations of available sensors and all permutations of available radio measurement capabilities. However, in embodiments where non-test UEs have a relatively large number and variety of suitable sensors or radio measurement capabilities, this attempt may not be feasible. Thus, in other embodiments, the training module selects from only a limited representative set of potential sensors and sensor configurations and radio measurement capabilities. By way of example, rider information from different rider modules manufactured by the same company may be relatively consistent, and thus, for example, if UE 104 can implement any of several rider sensors from its manufacturer, the training module can choose to exclude some rider sensors from the sensor configuration during training. As another example, a very small subset of UEs 104 may have the ability to measure SINR over a wider range or with a higher resolution, but the training module may limit the training to a smaller SINR range or lower SINR resolution that still covers most of the embodiments of UE 104.In yet other embodiments, there may be a defined set of sensor configurations that can be selected by the training module for training, and thus, the training module selects a sensor configuration from this defined set (and also avoids selecting sensor configurations that rely on sensor functions not generally supported by the associated devices).

[0057] When a sensor / wireless measurement configuration is selected, the training module identifies one or more sets of training data to be used when training candidate neural network architecture configurations based on the selected configuration. That is, one or more training data sets include or represent sensor data that can be generated by equivalent sensors of the selected sensor configuration, and represent wireless measurement data that can be generated in accordance with the selected wireless measurement capabilities, in order to train candidate neural network architecture configurations to be suitable for operating with sensor data provided by a particular sensor represented by the selected sensor / wireless measurement configuration. The training data may further include training data related to the operating state of the test UE, for example, the RRC state or 5G NR SA or NSA mode. When one or more training sets are obtained, the training module starts training the RRM NN on the test UE. In this training, typically, various bias weights and coefficients of the RRM NN are initialized with initial values, which are generally selected pseudo-randomly, and then a training data set (e.g., representing known sensor data from sensors of the selected sensor configuration and known wireless measurement data that matches the selected wireless measurement function configuration) is input, the test RRM NN processes the input to generate an output, the error between the actual output and the expected output is determined, the error is backpropagated throughout the test RRM NN, and this process is repeated for the next input data set. This process is repeated until a specific number of training iterations are performed or a specific minimum error rate is achieved. As a result of the training of the RRM NN of the test UE, the resulting neural network has a specific neural network architecture configuration, or a DNN architecture configuration if the implemented neural network is a DNN, which characterizes the architecture and parameters of the corresponding RRM NN, such as the number of hidden layers, the number of nodes in each layer, the connections between layers, the weights, coefficients, and other bias values implemented at each node.Therefore, once the training of the RRM NN of the test UE for the selected sensor / wireless measurement configuration is completed, the current neural network architecture configuration of the trained RRM NN is extracted, and at block 704, a local copy of the candidate neural network architecture configuration 324 is stored in BS110 or the copy of the candidate neural network architecture configuration 324 is stored in a remote data store accessible by BS110, so that it can be used in one or more BS110s as the corresponding candidate neural network architecture configuration 324 (Figure 3). In at least one embodiment, the candidate neural network architecture configuration 324 can be generated at the end of training by extracting the architecture and parameters of the corresponding RRM NN, such as the number of hidden layers, the number of nodes, connections, coefficients, weights, and other bias values.

[0058] If there are one or more other sensor / wireless measurement configurations that have not been trained, the training process is repeated by selecting the next sensor / wireless measurement configuration to be trained. Otherwise, if the candidate RRM NNs of UE104 have been trained for all intended sensor / wireless measurement configurations, network 100 can proceed to support the neural network-based RRM process in UE104 using the trained DNN.

[0059] Thus, as part of or subsequent to the initialization of the wireless connection between UE104 and serving BS110-1, at block 706, UE104 notifies serving BS110-1 of its sensor function and / or wireless measurement function by transmitting sensor / measurement function message 902 (FIG. 9) to serving BS110-1. For example, during the connection procedure, serving BS110-1 may transmit a UECapabilitiesEnquiry RRC message to UE104, and in response, UE104 may transmit a UECapabilitiesInformation RRC message to serving BS110-1, which has one or more fields of a UECapabilitiesInformation RRC message (one embodiment of sensor / measurement function message 902) that includes data representing one or both of the sensor function or wireless measurement function of UE104.

[0060] At block 708, the neural network management module 310 of serving BS110-1 selects a candidate neural network architecture configuration 324 for implementation in the RRM NN112 of UE104 based on one or both of the sensor functions advertised by UE104 or the radio measurement functions advertised by UE104. For example, if UE104 has an available radar sensor and imaging camera and can perform RSRP measurements at a specific range and resolution, the neural network management module 310 can select the candidate neural network architecture configuration 324 of UE104 that has been trained for this specific sensor configuration and radio measurement function. Once an appropriate neural network architecture configuration is selected, at block 710, serving BS110-1 instructs UE104 to implement the selected neural network architecture configuration in RRM NN112 by sending an NN configuration message 904 to UE104. The NN configuration message 904 can include one or more data structures that implement the selected neural network architecture configuration itself, or can include an index or other identifier of the selected neural network architecture configuration, such that UE104 can access the selected neural network architecture configuration from a local data store or from a remote server. In other embodiments, the neural network management module 218 of UE104 selects the neural network architecture configuration to be implemented in RRM NN112 based on the current sensor / measurement configuration of UE104, regardless of the direction from serving BS110-1.

[0061] In response to the NN configuration message 904 or in response to the selection of its own neural network architecture configuration, at block 712, UE 104 implements the indicated neural network architecture configuration in the RRM NN 112 of UE 104. After the RRM NN 112 is thus initialized, UE 104 can continue RRM management at UE 104 based on the RRM action 118 indicated by the output of the RRM NN 112. Thus, for the iteration of the input / output loop using the RRM NN 112, at block 714, the sensor set 210 provides a sensor data set 914 (an embodiment of the local sensor data 114, FIG. 1) representing sensor data captured by one or more sensors as an input to the RRM NN 112 of UE 104. At the same time, at block 716, the RRM module 222 and the RF front end 202 of UE 104 together obtain one or more radio measurements and provide these radio measurements as a radio measurement data set 916 (an embodiment of the radio measurement data 116, FIG. 1) as another input to the RRM NN 112. In some embodiments, these data sets 914 and 916 represent samples from a single time slice. For example, UE 104 may provide separate sensor data sets 914 and radio measurement data sets 916 every 10 milliseconds (ms). In other embodiments, one or both of these data sets are a sequence of sampled data over a plurality of time slices of a sliding window, such as a sequence of 10 separate data sets acquired every 100 ms in the previous second.

[0062] In 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 an output representing one or more RRM actions 918 (an embodiment of RRM action 118, FIG. 1) based on the current neural network architecture configuration and the current state of the nodes of the RRM NN 112. In block 720, the RRM module 222 of the UE 104 initiates the implementation of the RRM action(s) 918 by triggering the execution of RRM measurements by the RF front end 202 or other RRM processes, instructing the RF front end 202 to omit one or more scheduled RRM measurements or other RRM processes, reconfiguring one or more parameters of the scheduled RRM processes, or a combination thereof.

[0063] Consistent with a typical RRM scheme, in at least one embodiment, UE 104 contributes to the RRM scheme of wireless network 100 by providing an RRM report to serving BS 110-1, and serving BS 110-1, core network 106, or other infrastructure components can utilize it when making RRM decisions for the entire wireless system. This RRM report can be executed in response to a conventional RRM process implemented at UE 104, such as an embodiment of a static RRM measurement scheme separate from RRM NN 112. This RRM report can also be executed in response to the execution of RRM action(s) 918 by UE 104. Thus, continuing with method 700 in FIG. 8, at block 722, UE 104 provides an RRM report 920 to serving BS 110-1 in response to one or more RRM processes being executed at UE 104 due to RRM action 918 generated by RRM NN 112 or due to a separate and parallel static RRM process RRM NN 112 implemented at UE 104. Typically, RRM report 920 includes an RRM measurement report that includes one or more RRM measurements performed by UE 104 during a previous period. For example, if the RRM action 918 generated at block 718 represents an instruction to cause UE 104 to conduct a License Assisted Access (LAA) survey, as part of this LAA survey, UE 104 can execute radio measurements of this subset of cells, such as by identifying a subset of neighboring cells and then measuring the RSRP or SINR of each cell within the subset. Thus, the RRM report 920 in this example can include an RRM measurement report that includes the measured RSRP or measured SINR for each cell within this identified subset.

[0064] In response to the information contained in the RRM report 920, or in response to another trigger such as the expiration of a specified period since the last update, the serving BS 110-1 may determine that it is an appropriate point to update the underlying architecture configuration used in the RRM NN 112. This update process typically depends on using the same real-world inputs provided to the RRM NN 112 in a test UE that is typically in the serving BS 110-1 (as described above), and then iteratively backpropagating any error between the actual results output by the test UE and the expected or intended results. Thus, at block 724, the serving BS 110-1 transmits a sensor / wireless request 922 (Figure 9) to the UE 104, requesting that the UE 104 provide a copy of some or all of the sensor data set 914 and the wireless measurement data set 916 to the RRM NN 112, for example, since the last update process. In response to the sensor / wireless request 922, at block 726, the UE 104 transmits the requested sensor / wireless measurement data 924 (Figure 9) to the serving BS 110-1. In another embodiment, rather than buffering the sensor data and the wireless measurement data for transmission to the serving BS 110-1 in response to the request, the UE 104 can transmit a copy of each sensor data set 914 and / or a copy of each wireless measurement data set 916 at the time it is generated, or otherwise provide these data sets as input to the RRM NN 112 in parallel with transmitting them to the serving BS 110-1, in a single burst.

[0065] At block 728, the training module 312 of serving BS110-1 utilizes the sensor data and / or radio measurement data received from UE104 to execute a neural network update process 926, providing this data as input to RRM NN112, and revising the neural network architecture configuration that forms the basis of the test copy by comparing the resulting output with the expected or intended output, thereby retraining or updating the test copy of RRM NN112 implemented as a test UE by the training module 312. This training can further include the evaluation of RRM reports 920 enumerated during a period associated with the retraining data, evaluating the accuracy of the operation of RRM NN112 in UE104, and as a result, adjusting or otherwise revising the neural network architecture configuration of the test copy. Thereafter, the neural network management module 310 of serving BS110-1 extracts the updated neural network architecture configuration from the retrained test copy and provides the resulting updated neural network architecture configuration 324 to the RRM module 314 to replace or complement the previous version in data store 602, and at block 730, can transmit an update message 928 to UE104 that includes the representation of the updated neural network architecture configuration 324. At block 732, UE104 updates RRM NN112 using the updated neural network architecture configuration 324 represented in the update message 928, either by replacing the "old" neural network architecture configuration of RRM NN112 with the updated version or by selectively modifying or extending the previous version of RRM NN112 using the updated version.

[0066] Referring to FIG. 10, an alternative embodiment of method 700, called method 700-1, is shown. Method 700-1 proceeds in the same manner as method 700 with respect to the processes of blocks 702 to 720 of FIG. 7 and block 722 of FIG. 8. However, following the RRM reporting process of block 722, in method 700-1, UE 104 executes an NN update process and, rather than the reverse, reports the updated neural network architecture configuration to serving BS 110-1. Thus, in method 700-1, following the RRM reporting of block 722, at block 1024, the neural network management module 218 of UE 104 instantiates a copy of RRM NN 112 as a training copy and then evaluates some or all of the output RRM actions 118 generated from the input set of sensor data and radio measurement data, compared to the set of RRM actions that would be expected or intended if the same input were given (these expected RRM actions are specified, for example, by training guidance provided by network infrastructure 102). Next, the neural network management module 218 updates the architecture configuration underlying the training copy of RRM NN 112, for example, by repeatedly backpropagating the error between the actual and expected RRM actions. Next, at block 1026, the neural network management module 218 extracts the modified neural network architecture configuration, obtained as a result of the test copy of RRM NN 112, as the updated neural network architecture configuration and uses the updated neural network architecture configuration to update the actual RRM NN 112. At block 1028, this update can be transmitted to serving BS 110-1 for storage as candidate neural network architecture configuration 324 and / or for propagation to one or more other UEs 104 using the same original neural network architecture configuration with their respective RRM neural networks.

[0067] As described above, the RRM NN 112 can provide control for implementing various conditional RRM procedures that direct the UE 104 to conditionally execute RRM-related processes according to the status of one or more conditions monitored by the UE 104 by the network infrastructure 102. CHO is an example of such a conditional RRM process, and the UE is directed to initiate a handover on the condition that one or more monitored conditions subsequently reach corresponding thresholds. In some embodiments, the conditional RRM process is represented by the same RRM NN 112 that provides other types of RRM decisions in the UE 104. In other embodiments, one or more conditional RRM processes can operate in parallel with the RRM NN 112 and be implemented as separate conditional RRM neural networks that utilize overlapping inputs. FIG. 11 shows an exemplary manner 1100 of operation of a conditional RRM neural network, either as the RRM NN 112 (or RRM processing module 502) or as a separate neural network such as a CHO processing module 504. The order of operations described with reference to manner 1100 is for illustrative purposes only, and operations in a different order may be performed, and furthermore, one or more operations may be omitted, or one or more additional operations may be included in the manner shown. For ease of understanding, manner 1100 is described with reference to an exemplary embodiment of manner 1100 reflected in the ladder diagram 1200 of FIG. 12.

[0068] Method 1100 begins following the training of a set of candidate CHO neural network architecture configurations using the above-described training process. Thus, at block 1102, UE 104 notifies serving BS 110-1 of its sensor function and / or radio measurement function by transmitting sensor / measurement function message 1202 (FIG. 12) to serving BS 110-1. This message 1202 may include, for example, a UECapabilitiesInformation RRC message provided in response to a UECapabilitiesEnquiry RRC message from serving BS 110-1. At block 1104, neural network management module 310 of serving BS 110-1 selects a candidate CHO neural network architecture configuration (e.g., an example of neural network architecture configuration 324) from set 322 based on either or both of the sensor function advertised by UE 104 or the radio measurement function advertised by UE 104. Once an appropriate CHO neural network architecture configuration is selected, at block 1106, serving BS 110-1 instructs UE 104 to implement the selected CHO neural network architecture configuration in CHO processing module 504 by transmitting CHO configuration message 1204 (FIG. 12) to UE 104. CHO configuration message 1204 may include one or more data structures implementing the selected neural network architecture configuration itself, or may include an index or other identifier of the selected neural network architecture configuration, such that UE 104 can access the selected neural network architecture configuration from a local data store or from a remote server. In other embodiments, neural network management module 218 of UE 104 selects the CHO neural network architecture configuration to implement in CHO processing module 504 based on the current sensor / measurement configuration of UE 104, independent of the direction from serving BS 110-1.

[0069] In response to the CHO configuration message 1204, or in response to the selection of its own CHO neural network architecture configuration, at block 1108, UE 104 implements the indicated CHO neural network architecture configuration in the CHO processing module 504, thus completing the initialization of the CHO processing module 504.

[0070] The UE 104 can then proceed with conditional RRM management at the UE 104 based on the CHO action 520 indicated by the output of the CHO processing module 504. Thus, in the case of iterating 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) (an embodiment of the local sensor data 114, FIG. 1) representing sensor data captured by one or more sensors as an input to the CHO processing module 504. As will be appreciated, the decision to initiate a handover as a result of the CHO instruction typically reflects the result of comparing the suitability of the current serving cell with one or more adjacent cells or target cells, and the suitability is indicated by the comparison of various radio measurement values of the serving cell and the adjacent cell(s). Thus, concurrently with the iterative provision of sensor data, at block 1112, the RRM module 222 and the 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 the RF front end 202 obtain one or more radio measurement values of each applicable neighboring BS 110 such as the target BS 110-2, and provide these radio measurement values as another input to the CHO processing module 504 as a radio measurement data set 1216 (FIG. 12) (an embodiment of the radio measurement data 116, FIG. 1). 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 sample data over a plurality of time slices within a sliding window.

[0071] In block 1116, the CHO processing module 504 processes these inputs and, in some embodiments, additional inputs such as the current RRC state or the current 5G NR mode, to generate an output representing one or more CHO actions 1220 (FIG. 12) (an embodiment of CHO action 520, FIG. 5) based on the current neural network architecture configuration and the current state of the nodes of the CHO processing module 504. Next, the RRM module 222 of the UE 104 initiates the execution of the CHO action(s) 1220. As described above, in some embodiments, the CHO action 1220 can include a process for adjusting some parameters of the CHO process, such as changing the thresholds used when making handover decisions. However, in other embodiments, the CHO action 1220 represents the handover decision itself. In such embodiments, in block 1118, the RRM module 222 determines the handover decision represented by the CHO action 1220. If the handover decision is to anticipate the start of a handover, the method 1100 can return to blocks 1110, 1112, and 1114 for another iteration of the CHO action generation process using a new set of sensor and measurement data inputs to the CHO processing module 504.

[0072] In other cases, when the handover decision is to initiate a handover, at block 1120, the RRM module 222 and the RF front end 202 are adjusted to proceed with a handover 1222 (FIG. 12) to the indicated neighboring cell (represented in this example by target BS110-1). After the handover is executed, it will be understood that the UE 104 is now connected to a different cell that may have different RRM parameters compared to the previous cell associated with BS110-1. Thus, in at least one embodiment, at least one configuration of the RRM NN 112 (e.g., as the RRM processing module 502) or the CHO processing module 504 is replaced with a new configuration provided by BS110-2 as the new serving BS for the UE 104. Thus, at block 1122, the same or similar neural network architecture configuration process described above with reference to blocks 1102-1108 may be executed again, but this time using the neural network architecture configuration(s) selected and provided by BS110-2 instead of BS110-1. Thus, the UE 104 can provide the representation of its sensors and / or radio measurement capabilities as a function message 1224 (FIG. 12) to BS110-2, and BS110-2 can select an appropriate neural network architecture configuration 1226 (FIG. 12) based on these capabilities and instruct the UE 104 to implement the selected neural network architecture configuration 1226 of the CHO processing module 504.

[0073] After that, UE104 can execute the neural network-based CHO decision-making process in the same way as above using the newly configured CHO processing module 504, which includes providing the sensor data set 1234 (FIG. 12) and the radio measurement data set 1236 (FIG. 12) obtained from the sensor set 210 as inputs to the CHO processing module 504. From these inputs, the CHO processing module 504 generates a CHO action 1240 suitable for the CHO-related RRM scheme implemented by BS110-2. Next, UE104 may perform one or more CHO processes represented by the CHO action 1240 described herein.

[0074] Embodiments of the present disclosure may also be better understood by considering the following non-limiting examples.

[0075] Example 1: A computer-implemented method, in a wireless device, receiving a first set of sensor data from one or more sensors of the wireless device, receiving a first set of radio measurements from a radio interface of the wireless device different from the one or more sensors, processing the first set of sensor data and the first set of radio measurements in a radio resource management (RRM) neural network of the wireless device to generate a first output representing a first RRM action, and executing the first RRM action on the wireless device, A computer-implemented method comprising.

[0076] Example 2: The wireless device is a user equipment, and the method is providing at least one of the representations of the sensor functions of the wireless device for reception by an infrastructure component of a network infrastructure wirelessly connected to the wireless device, In response to providing the representation of the sensor function, receiving a first neural network architecture configuration from the infrastructure component, and implementing the first neural network architecture configuration in the RRM neural network, the method according to Example 1.

[0077] Example 3: The method according to Example 2, wherein the representation of the sensor function includes one or more fields of a UECapabilitiesInformation radio resource control (RRC) message.

[0078] Example 4: The method further includes providing at least one of the set of first sensor data or the set of first radio measurements to the infrastructure component, in response to providing at least one of the set of first sensor data or the set of first radio measurements, further receiving a second neural network architecture configuration from the infrastructure component, the second neural network architecture configuration representing a modification of the first neural network architecture configuration based on at least one of the set of first sensor data or the set of first radio measurements, and the method implementing the second neural network architecture configuration in the RRM neural network, The method according to Example 2 or 3, further comprising.

[0079] Example 5: Modifying the first neural network architecture configuration based on at least one of the set of first sensor data or the set of first radio measurements to generate a second neural network architecture configuration, and implementing the second neural network architecture configuration in the RRM neural network, The method according to Example 2 or 3, further comprising.

[0080] Example 6: further comprising receiving a representation of the operating state of the wireless device Processing includes processing the set of first sensor data, the set of first radio measurements, and the operating state in the RRM neural network to generate the first output, according to the method described in any of Examples 1 to 5

[0081] Example 7: The method according to Example 6, wherein the operating state includes the radio resource control (RRC) state of the wireless device

[0082] Example 8: The method according to any of Examples 1 to 7, wherein the one or more sensors include at least one of a position sensor, an attitude sensor, an accelerometer, a pressure sensor, or a proximity sensor

[0083] Example 9: The method according to Example 8, wherein the position sensor includes at least one of a satellite-based positioning sensor or a visual remote sensing imaging sensor, and the attitude sensor includes at least one of a gyroscope or a visual remote sensing imaging sensor

[0084] Example 10: The method according to any of Examples 1 to 9, wherein the set of first radio measurements includes at least one signal power measurement of a serving cell or an adjacent cell

[0085] Example 11: The method according to Example 10, wherein the signal power measurement includes at least one of a signal-to-noise (SNR) measurement, a signal-to-interference plus noise (SINR) measurement, a reference signal received power (RSRP) measurement, or a received signal strength indication (RSSI) measurement

[0086] Example 12: The method according to any one of Examples 1 to 11, wherein the RRM action includes at least one of: performing RRM-related measurements by the wireless device; setting characteristics of the RRM-related measurements performed by the wireless device; or performing an RRM reporting process in the wireless device.

[0087] Example 13: The method according to 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.

[0088] Example 14: The method according to Example 12 or 13, wherein the characteristics of the RRM-related measurements include at least one of the frequency or timing of the RRM-related measurements, or the frequency band or channel of the RRM-related measurements.

[0089] Example 15: The method according to Example 1, wherein the RRM action includes performing a conditional handover (CHO) decision or a conditional PSCell change (CPC) decision.

[0090] Example 16: In response to connecting to a cell as a result of a conditional handover decision, providing a representation of the sensor function of the wireless device for reception by an infrastructure component of the cell; in response to providing the representation of the sensor function, receiving a first neural network architecture configuration from the infrastructure component of the cell, and replacing the second neural network architecture configuration of the RRM neural network with the first neural network architecture configuration. The method according to Example 15, further comprising the above steps.

[0091] Example 17: Receiving a second set of sensor data from the one or more sensors of the wireless device. Receiving a second set of wireless measurement values from the wireless interface of the wireless device; Processing the second set of sensor data and the second set of wireless measurement values in the RRM neural network of the wireless device to generate a second output representing a second RRM action; and Executing the second RRM action on the wireless device, The method according to embodiment 16, further comprising.

[0092] Embodiment 18: The method according to any one of embodiments 1 to 17, wherein the wireless device is a user equipment.

[0093] Embodiment 19: The method according to embodiment 1, wherein the wireless device is a base station.

[0094] Embodiment 20: The method according to embodiment 19, further comprising receiving an RRM report from a user equipment.

[0095] Embodiment 21: A wireless interface, At least one processor coupled to the wireless interface, and At least one memory, the at least one memory being coupled to the at least one processor and storing executable instructions configured to operate the at least one processor to execute the method according to any one of embodiments 1 to 20. A wireless device.

[0096] Embodiment 22: A computer-implemented method in an infrastructure component wirelessly connected to a wireless device, comprising: Receiving, from the wireless device, a representation of the sensor function of the wireless device; Determining a first neural network architecture configuration based on the representation of the sensor function; and To determine the RRM actions to be performed by the wireless device, for implementation of the wireless resource management (RRM) neural network of the wireless device, transmitting an expression of the first neural network architecture configuration to the wireless device, A computer-implemented method, including

[0097] Example 23: Receiving at least one of a set of sensor data or a set of wireless measurements from the wireless device, Modifying the first neural network architecture configuration based on the at least one of the set of sensor data or the set of wireless measurements to generate a second neural network architecture configuration, and For implementation in the RRM neural network of the wireless device, transmitting an expression of the second neural network architecture configuration to the wireless device, The method according to Example 22, further including

[0098] Example 24: The method according to Example 22 or 23, wherein the set of sensor data includes at least one of position data, attitude data, acceleration data, air pressure or altitude data, or proximity data.

[0099] Example 25: The method according to any one of Examples 22 to 24, wherein the set of wireless measurements includes at least one signal power measurement of a serving cell or an adjacent cell.

[0100] Example 26: The method according to any one of Examples 22 to 25, wherein the determined RRM action includes at least one of performing RRM-related measurements by the wireless device or setting characteristics of RRM-related measurements performed by the wireless device.

[0101] Example 27: The determined RRM action is the method according to any one of Examples 22 to 26, including the execution of a conditional handover decision.

[0102] Example 28: The method according to any one of Examples 22 to 26, wherein the infrastructure component includes a base station.

[0103] Example 29: The method according to any one of Examples 22 to 28, wherein the wireless device includes a user equipment.

[0104] Example 30: The method according to any one of Examples 22 to 29, wherein the representation of the sensor function includes one or more fields of a UECapabilitiesInformation radio resource control (RRC) message.

[0105] Example 31: A set of one or more sensors, A wireless interface different from the set of one or more sensors, At least one processor coupled to the wireless interface, and At least one memory, the at least one memory being coupled to the at least one processor and storing executable instructions configured to operate the at least one processor to execute the method according to any one of Examples 22 to 30. An infrastructure component.

[0106] In some embodiments, certain aspects of the techniques described above may be implemented by one or more processors of a processing system that executes software. The software includes one or more sets of executable instructions stored on a non-transitory computer-readable storage medium or otherwise tangibly embodied. The software can include instructions and certain data that, when executed by one or more processors, cause the one or more processors to operate so as to execute one or more aspects of the techniques described above. Non-transitory computer-readable storage media can include, for example, magnetic or optical disk storage devices, solid state storage devices such as flash memory, cache, random access memory (RAM), or other single or multiple non-volatile memory devices, and the like. The executable instructions stored on the non-transitory computer-readable storage medium can be in source code, assembly language code, object code, or another instruction format that is interpretable by one or more processors or otherwise executable.

[0107] A computer-readable storage medium can include any storage medium, or combination of storage media, that is accessible by a computer system during use to provide instructions and / or data to the computer system. Such storage media can include, but are not limited to, optical media (e.g., compact disc (CD), digital versatile disc (DVD), Blu-ray disc), magnetic media (e.g., floppy (registered trademark) disk, magnetic tape, or magnetic hard drive), volatile memory (e.g., random access memory (RAM) or cache), non-volatile memory (e.g., read-only memory (ROM) or flash memory), or microelectromechanical systems (MEMS)-based storage media. The computer-readable storage medium can be incorporated in a computing system (e.g., system RAM or ROM), fixedly attached to a computing system (e.g., magnetic hard drive), removably attached to a computing system (e.g., optical disc or universal serial bus (USB)-based flash memory), or coupled to the computer system via a wired or wireless network (e.g., network-accessible storage (NAS)).

[0108] In addition to the above, note that not all activities or elements described above in the general description are required, that some activities or parts of a particular activity or device may not be required, that one or more additional activities may be performed, or that one or more additional elements may be included. Further, the order in which activities are listed is not necessarily the order in which they are performed. Also, the concepts are described with reference to specific embodiments. However, those skilled in the art will understand that various changes and modifications can be made without departing from the scope of the present disclosure as set forth in the following claims. Accordingly, the specification and drawings are to be regarded in an illustrative rather than a limiting sense, and all such modifications are intended to be included within the scope of the present disclosure.

[0109] Advantages, other advantages, and solutions to problems have been described above with respect to specific embodiments. However, an advantage, advantage, or solution to a problem, and any feature(s) that may give rise to or make more prominent any advantage, advantage, or solution, should not be construed as a critical, essential, or indispensable feature of any or all of the claims. Furthermore, the disclosed subject matter of the invention may be modified and practiced in different but equivalent manners that will be apparent to those skilled in the art having the benefit of the teachings herein, and thus the specific embodiments disclosed above are merely illustrative. It is not intended to limit the details of the construction or design shown herein to anything other than as described in the following claims. Therefore, it is evident that the specific embodiments disclosed above may be changed or modified and that all such variations are contemplated as being within the scope of the disclosed subject matter of the invention. Thus, the protection sought herein is as set forth in the following claims.

Claims

1. In a wireless device, receiving a first set of sensor data from one or more sensors of the wireless device; receiving a first set of radio measurements from a radio interface of the wireless device different from the one or more sensors; processing the first set of sensor data and the first set of radio measurements with a radio resource management (RRM) neural network of the wireless device to generate a first output representing a first RRM action; and performing the first RRM action at the wireless device, A computer-implemented method comprising:

2. The wireless device is a user equipment, and the method further comprises: providing at least one of representations of sensor functions of the wireless device for the one or more sensors for reception by an infrastructure component of a network infrastructure wirelessly connected to the wireless device; in response to providing the representation of the sensor function, receiving a first neural network architecture configuration from the infrastructure component; and implementing the first neural network architecture configuration in the RRM neural network. The method according to claim 1, further comprising:

3. The method according to claim 2, wherein the representation of the sensor function comprises one or more fields of a UE Capabilities Information radio resource control (RRC) message.

4. The method further comprises: providing at least one of the first set of sensor data or the first set of radio measurements to the infrastructure component; in response to providing the at least one of the first set of sensor data or the first set of radio measurements, further receiving a second neural network architecture configuration from the infrastructure component, the second neural network architecture configuration representing a modification of the first neural network architecture configuration based on the at least one of the first set of sensor data or the first set of radio measurements, and the method further comprises: Implementing the second neural network architecture configuration in the RRM neural network; The method according to claim 2 or 3, further comprising.

5. Modifying the first neural network architecture configuration based on at least one of the set of the first sensor data or the set of the first radio measurements to generate a second neural network architecture configuration, and Implementing the second neural network architecture configuration in the RRM neural network; The method according to claim 2 or 3, further comprising.

6. Receiving a representation of the operating state of the wireless device, further comprising, Processing includes processing the set of the first sensor data, the set of the first radio measurements, and the operating state in the RRM neural network to generate the first output. The method according to any one of claims 1 to 5.

7. The method according to claim 6, wherein the operating state includes a radio resource control (RRC) state of the wireless device.

8. The method according to any one of claims 1 to 7, wherein the one or more sensors include at least one of a position sensor, an attitude sensor, an accelerometer, a pressure sensor, or a proximity sensor.

9. The method according to any one of claims 1 to 8, wherein the set of the first radio measurements includes signal power measurements of at least one of a serving cell or an adjacent cell.

10. The method according to any one of claims 1 to 9, wherein the RRM action includes at least one of performing RRM-related measurements by the wireless device, setting characteristics of the RRM-related measurements performed by the wireless device, or performing an RRM reporting process at the wireless device.

11. The method according to claim 10, wherein the characteristics of the RRM-related measurements include at least one of the frequency or timing of the RRM-related measurements, or the frequency band or channel of the RRM-related measurements.

12. The method according to any one of claims 1 to 11, wherein the RRM action includes performing a conditional handover (CHO) decision or a conditional PSCell change (CPC) decision.

13. In response to connecting to a cell as a result of a conditional handover decision, providing an indication of a sensor function of the wireless device for reception by an infrastructure component of the cell; in response to providing the indication of the sensor function, receiving a first neural network architecture configuration from the infrastructure component of the cell; and replacing the second neural network architecture configuration of the RRM neural network with the first neural network architecture configuration, The method according to claim 12, further comprising.

14. receiving a second set of sensor data from the one or more sensors of the wireless device; receiving a second set of wireless measurements from the wireless interface of the wireless device; processing the second set of sensor data and the second set of wireless measurements in the RRM neural network of the wireless device to generate a second output representing a second RRM action; and executing the second RRM action at the wireless device, The method according to claim 13, further comprising.

15. The method according to any one of claims 1 to 14, wherein the wireless device is a user equipment.

16. The method according to claim 1, wherein the wireless device is a base station.

17. The method according to claim 16, further comprising receiving an RRM report from a user equipment.

18. one or more sensors; a wireless interface different from the one or more sensors; at least one processor coupled to the wireless interface; and at least one memory coupled to the at least one processor and storing executable instructions configured to operate the at least one processor to perform the method according to any one of claims 1 to 17, Wireless device.

Images