Abstract: Despite the usefulness of foundation models, they often return irrelevant or inaccurate responses to domain-specific queries, which often include technical jargon and patterns that are unique to the domain. Training data for specialized domains is not readily accessible or frequently discussed in online forums, so foundation models lack understanding in specialized domains. Moreover, while the embedding models underlying foundation models may be finetuned using annotated datasets, generating these datasets for niche and specialized domains is an extremely arduous, time-consuming, and expensive task. These issues are overcome by implementing a generative pseudo labeling (GPL) approach to creating labeled data for finetuning embedding models to recognize semantic similarities for specialized domains. In this way, the accuracy and relevance of foundation-model responses to domain-specific queries is improved.

Abstract: Described are examples for calculating and exposing network capacity and congestion to applications. A network entity such as a radio access network (RAN) intelligent controller (RIC) or virtual base station component receives measurements of a signal quality for a plurality of user devices connected to a RAN. The network entity estimates a deliverable throughput of a wireless link for a user device of the plurality of user devices based on at least the measurements. The network entity can consider other factors such as a number of competing users, queue sizes of the user device and of the competing users, or a scheduling policy. The network entity provides the deliverable throughput to an application server for an application of the user device communicating with the application server via the RAN. The application server can adapt a data rate for the application and the user device based on the deliverable throughput.

Abstract: Techniques are disclosed for dynamically adjusting associations between radio units (RUs) and a virtualized radio access network (vRAN) by a virtual translational layer running in a controller of the vRAN. Each of the RUs have a number of antennas and are configured to service a cell of a cellular communications network. Based on performance metrics, the virtual translational layer maps the RUs to a virtualized cell of the cellular communications network, the virtualized cell including a mapped selection of the RUs.

Abstract: The present optimization of guard bands repurposes some guard band spectrum for data transmission in a 5G network. This approach takes spectrum that is otherwise “wasted” for guard bands to enable overall spectrum utilization to be increased. To mitigate effects of inter-numerology interference (INI) with narrower guard band bandwidth, physical resource blocks (PRBs) for particular user equipment (UE) are allocated to BWPs that are modified with increased bandwidth that comes from narrowing the guard band bandwidth. These particular UEs have high signal strength, for example, as characterized by SINR (signal-to-interference plus noise ratio), relative to other UE. Allocating PRBs for high signal strength UE in BWPs near the edges of the narrower guard band increases the risk of INI, but the higher signal strength for these UE helps to lessen the INI impact and enable overall throughput for all users to be maximized in the network.

Abstract: A context analysis system receives a query from a user. The context analysis system generates one or multiple context profiles and generates a prompt for a foundation model for each of the context profiles. The context analysis system analyzes each of the context profiles and generates a relevancy score. The context analysis system selects one of the context profiles based on the relevancy score. In some examples, the context analysis system iteratively determines predicted latencies and relevancies of processing a query in conjunction with a generated context and, based on the predicted latencies and/or relevancies, processes the query using a foundation model, such as a large language model (LLM).

Abstract: Operators managing a cloud RAN collect vast amounts of data, e.g., node-level data, gNodeB level data, user level data, and flow-level data, which are utilized for network monitoring, evaluating key performance indicators (KPIs), and nodes management. Retrieving and visualizing information and values of different metrics is critical to managing network operation; however, data retrieval on large datasets is challenging. While foundation models perform poorly on large datasets, an accurate answer to a data query is generated by providing semantically similar metrics as context to a foundation model, thereby limiting the number of counters needed for processing the data query. The foundation model then generates a first output of metrics relevant to answering the data query and, based on the first output, generates a second output comprising query code (e.g., SQL or KQL) for computing the answer, thereby improving mathematical accuracy of the answer.

Abstract: The techniques disclosed herein manage computing environments associated with radio access networks using a natural language interface. This is achieved through utilizing natural language processing to analyze user generated inputs and generate robust large language model queries. In various examples, the queries can include radio access network documentation, diagnostic data, and past interactions to provide custom context to the large language model. Accordingly, the query can cause the large language model to generate an operation sequence comprising a plurality of commands to interface with a resource management tool and control computing resources and supporting components. In this way, the present techniques can alleviate the technical burden on end users and minimize the risk of errors.

Abstract: Techniques are disclosed for dynamically adjusting associations between radio units (RUs) and a virtualized radio access network (vRAN) by a virtual translational layer running in a controller of the vRAN. Each of the RUs have a number of antennas and are configured to service a cell of a cellular communications network. Based on performance metrics, the virtual translational layer maps the RUs to a virtualized cell of the cellular communications network, the virtualized cell including a mapped selection of the RUs.

Abstract: Systems and methods are provided for determining a set of control parameter data associated with a base station of a 5G multi-access edge computing and core network. In particular, the disclosed technology is directed to use a deep reinforcement-based learning (DRL) model to iteratively reinforce and improve the set of control parameter data at the base station. The DRL model determines the set of control parameter data as action based on a current set of network state data as state, according a set of target conditions used as rewards. A DRL server periodically receives network state data from the base station through a radio access network intelligent controller (RIC). Given the network state data, the DRL model determines control parameter data as output. The DRL server transmits the control parameter data to the base station via RIC. The periodic reinforcement-based learning dynamically improves a network performance of the base station.

Abstract: The present disclosure generally relates to improving energy efficiency of reality-based headsets by enabling discontinuous transmission (DRX) without significantly reducing the quality of experience (QoE) of a user of the headsets when consuming digital content presented thereon. The present disclosure includes a predictive content management system that obtains consumable content (including predictive content) to be consumed on a wearable UE. The system receives, from a radio access network (RAN), discontinuous reception (DRX) configuration information indicating active and inactive periods with which the UE and RAN may communicate. The system additionally facilitates stacking consumable content, delivering the consumable content based on the DRX configuration information, and performing certain latency masking features to avoid loss of quality of the presented content.

Abstract: Existing approaches to understanding, developing, and researching modern wireless communication technologies involve time intensive and arduous processes of sifting through numerous webpages and technical specification documents, gathering the required information and synthesizing it. The present disclosure describes a conversational artificial intelligence tool for information synthesis of wireless communication specifications. The system builds on recent advancements in foundation large language models (LLMs) and consists of three key additional components: a domain-specific database, a context extractor, and a feedback mechanism. The system appends user queries with concise contextual information extracted from a database of wireless technical specifications and incorporates tools for expert feedback and data contribution.

Abstract: A fifth generation (5G) mobile network radio access network (RAN) is virtualized for operations on edge computing platforms in a cloud-computing environment in which radio units (RUs) and radio frequency (RF) spectrum are shared among distributed units (DUs) to support use cases including: 1) live migration in which a DU is moved from one computing server to another without disruption to network traffic, and 2) RAN sharing in which two DUs share the same RU and spectrum.

Abstract: Aspects include a machine learning based resource block scheduler configured to meet service level requirements of applications. Aspects include receiving a plurality of scheduling requests each associated with a respective application of a plurality of applications on a plurality of wireless devices, identifying a plurality of current channel state information each associated with one of the plurality of wireless devices, and identifying a plurality of different types of service level requirements each associated with one of the plurality of applications.

Abstract: A method for controlling transmission power from one or more radio units is provided including monitoring channel state feedback for a signal communicated between a first radio unit of the one or more radio units and a user device in a transmitted frequency range, wherein the channel state feedback is based at least in part on a metric of quality of the communicated radiofrequency signal, determining that the channel state feedback satisfies a channel state condition, wherein the channel state condition includes a metric to evaluate performance of the one or more radio units relative to the user device based at least on the metric of quality of the communicated signal, and transmitting an instruction to adjust a transmission power in the transmitted frequency range of at least one of the one or more radio units based at least on the satisfaction of the channel state condition.

Abstract: Described are examples for machine learning based interference detection for tiered licensing deployments. A network entity in a general authorized access (GAA) deployment checks a local spectrum access database of GAA users to determine that a portion of shared use spectrum is free from known local users in a geographic area. The network entity receives samples of a wireless signal including at least a desired signal on the portion of shared use spectrum. The network entity determines whether the wireless signal includes multiple independent signals. The network entity identifies an interfering signal in response to determining that the wireless signal includes multiple independent signals.

Abstract: Described are examples for repurposing mobility management with virtual radio in software radio access networks. A virtual mobile network includes a first server configured to host a first mobile network distributed unit (DU) for providing a first virtual cell to a plurality of user devices via a radio unit. The virtual mobile network also includes a second server configured to host a second mobile network distributed unit providing a second virtual cell via the same radio unit. A radio access network (RAN) intelligent controller (RIC) is configured to control the first DU and the second DU to hand over the plurality of user devices from the first virtual cell to the second virtual cell. The first server may then be shut down for maintenance or updates without dropping service to the user devices.

Abstract: Described are examples for calculating and exposing network capacity and congestion to applications. A network entity such as a radio access network (RAN) intelligent controller (RIC) or virtual base station component receives measurements of a signal quality for a plurality of user devices connected to a RAN. The network entity estimates a deliverable throughput of a wireless link for a user device of the plurality of user devices based on at least the measurements. The network entity can consider other factors such as a number of competing users, queue sizes of the user device and of the competing users, or a scheduling policy. The network entity provides the deliverable throughput to an application server for an application of the user device communicating with the application server via the RAN. The application server can adapt a data rate for the application and the user device based on the deliverable throughput.

Abstract: Aspects of the present disclosure relate to determining reference symbol transmission times. In some examples, a method for determining reference symbol transmission times for cellular communications includes receiving signal feedback based on a wireless communication channel between a wireless communication device and a base station, identifying a periodic exchange of reference symbols that are used to adjust beamforming between the wireless communication device and the base station, generating a vector based on the signal feedback, and providing the vector as an input to a trained machine learning model. A training of the trained machine learning model includes calculating a plurality of rewards for a respective plurality of transmission time delays. The plurality of rewards are each calculated based on a function of downlink throughput and uplink overhead. The function of downlink throughput and uplink overhead are based upon a priority level of the wireless communication device.

Abstract: In a 5G network, a slice controller operating in a radio access network (RAN) is arranged to make predictions of channel state information (CSI) for user equipment (UE) on the network using a predictive propagation model. The slice controller uses the predicted CSI to schedule subcarriers and time slots associated with physical radio resources for data transmission on slices of the 5G network between a 5G radio unit (RU) and the UE to maximize network throughput on a slice for the radio spectrum that is utilized for a given time period. In view of the CSI predictions, the slice controller controls operations of the MAC (Medium Access Control) layer functions based on PHY (physical) layer radio resource subsets to schedule the subcarrier and time slots for data transmissions on a slice over the 5G air interface from RU to UE.

Abstract: Aspects of the present disclosure relate to allocating RAN resources among RAN slices using a machine learning model. In examples, the machine learning model may determine an optimal RAN resource configuration based on compute power needs. As a result, RAN resource allocation generation and compute power requirements may improve, even in instances with changing or unknown network conditions. In examples, a prediction engine may receive communication parameters and/or requirements associated with service-level agreements (SLAs) for applications executing at least partially at a device in communication with the RAN. The RAN may generate one or more RAN resource configuration for implementation among RAN slices. Upon a change in network conditions or SLA requirements, an optimal RAN configuration may be determined in terms of required compute power.