Automatic tuning of heterogeneous wireless infrastructure

Abstract

Embodiments for automatically tuning a heterogeneous wireless network are disclosed herein. In one example, performance data is received for a plurality of wireless networks. The wireless network is based on multiple wireless technologies and the performance data is based on multiple layers of a wireless technology protocol stack. One or more configuration settings to be adjusted for the one or more wireless networks are determined using the performance data. The determined configuration setting (s) is then adjusted.

Description

[0001] Cross Reference to Related Applications

[0002] This application claims priority to U.S. Patent Application No. 18 / 209,827, filed June 14, 2023, and titled “Automatic Tuning of Heterogeneous Wireless Infrastructure,” the entirety of which is incorporated herein by reference. BACKGROUND

[0003] Deployment of heterogeneous wireless infrastructure is emerging at the network edge for various use cases. In these deployments, the growing number and variety of wireless technologies present a variety of challenges related to anomalies and interference that result in subpar performance. For example, performance degradation can be caused by a variety of factors, such as interference between different radio technologies operating on the same or overlapping spectrum and needing to coexist, cross-technology interference caused by inadequate deployment planning, or unexpected or malicious interference caused by external sources. Furthermore, the impact on performance can vary due to the dynamic nature of the environment, such as work or business hours, personnel in the environment, physical obstructions, external factors, and the like. BRIEF DESCRIPTION OF DRAWINGS

[0004] Figure 1 A system for automatic tuning of wireless infrastructure is illustrated.

[0005] Figures 2A-2B A model development and deployment pipeline for supervised and unsupervised learning is illustrated.

[0006] Figure 3 A processing pipeline for automatic tuning of wireless infrastructure is illustrated.

[0007] Figure 4 A processing pipeline for continual learning is illustrated.

[0008] Figure 5 A decision tree model for anomaly detection using quality of service metrics is illustrated.

[0009] Figures 6A-6B An example of a recurrent neural network is illustrated.

[0010] Figure 7 An example of a protocol stack for network technologies is illustrated.

[0011] Figure 8 A flowchart for automatic tuning of wireless infrastructure is illustrated.

[0012] Figure 9 An overview of edge cloud configuration for edge computing is illustrated.

[0013] Figure 10 Operation layers between an endpoint, edge cloud, and cloud computing environments are illustrated.

[0014] Figures 11A-11B Example embodiments of computing devices and systems are illustrated. DETAILED DESCRIPTION

[0015] Heterogeneous wireless radio based technology deployments are increasingly arising at the network edge for various use cases, such as the Internet of Things (IoT). In these deployments, the growing number and variety of wireless technologies brings a variety of challenges related to anomalies and interference that result in poor performance. The root causes of performance degradation in a deployment can include a variety of factors, such as interference between different wireless technologies operating on the same or overlapping spectrum and needing to coexist, cross-technology interference due to inadequate deployment planning, and unexpected or malicious interference caused by external sources in the deployment infrastructure, such as weather conditions, neighboring wireless deployments, etc. The dynamic nature of the environment exacerbates these challenges, as the impact on performance can vary greatly based on changing environments, such as work schedules (e.g., business hours, production schedules, etc.), personnel in the environment, physical obstructions, external factors, etc.

[0016] As an example, portable “box cellular network” deployments are gaining interest as a deployment mode for standalone or private cellular networks, but such deployments are highly susceptible to signal interference and anomalies. A “box cellular network” deployment (e.g., a box 5G network) typically contains a portable container or enclosure that contains all of the equipment needed to deploy a standalone cellular network within it. In certain use cases, this deployment mode is used in environments where public cellular networks do not exist, are down, or provide unstable coverage to deploy cellular technologies such as 4G and 5G in a highly secure manner. This includes disaster recovery and humanitarian action scenarios where there is no known (or trusted) 4G or 5G infrastructure. In these use cases, the “box 5G” cellular deployment does not have knowledge of its deployment environment, including which radio frequencies are already in use, whether other wireless technologies are deployed, and whether there are any potential sources of interference. This can result in severe connectivity issues in these scenarios. Furthermore, in some cases, the box 5G deployment can be in a mobile state (e.g., within a vehicle), which greatly increases the likelihood of performance issues due to the changing deployment environment.

[0017] Accordingly, the present disclosure presents embodiments for automatically tuning heterogeneous wireless infrastructure. In some embodiments, a machine learning based solution is used to detect and classify sources of interference across a wide radio frequency range in a distributed and collaborative manner, helping to provide recommendations and pinpointing the source of errors. For example, this solution can detect and pinpoint sources of interference in heterogeneous wireless deployments in manufacturing plants, sub-manufacturing facilities (sub-factories) of semiconductor factories, or environments with standalone cellular networks, such as “box 5G” deployments, etc.

[0018] The solution uses trained artificial intelligence and machine learning (AI / ML) models to infer such anomalies by analyzing performance data for wireless technologies (covering multiple protocol layers of their respective network stacks) across layers, such as:

[0019] (i) scans of target radio frequency (RF) spectrum and energy;

[0020] (ii) signal quality data (e.g., signal strength, signal-to-noise ratio (SNR));

[0021] (iii) demodulation errors;

[0022] (iv) data transmission errors (e.g., bit error rate (BER) information before and after forward error correction (FEC) scheme application, block error rate (BLER), packet error rate (PER), Internet Protocol (IP) checksum errors); and

[0023] (v) traffic performance statistics at network and transport layers (e.g., packet collision / drop, retransmission statistics).

[0024] For example, the training phase uses cross-stack performance data from different protocol layers, such as performance metrics for a particular technology and RF spectrum scans obtained by using a set of distributed nodes in deployment for sensing, to determine what constitutes normal state in time and spatial domains for the respective wireless technology (e.g., normal performance and behavior based on time and location).

[0025] Subsequently, the trained model can analyze live performance data along with other input data sources to notify of external anomalies that can be sources of interference, as well as technology-specific optimizations (e.g., automatic corrections, steering of terminal and user equipment (UE)) to help meet overall goals related to quality of service (QoS) and service level objective (SLO) requirements for different traffic flows.

[0026] The solution provides various advantages. Network planning and troubleshooting for heterogeneous wireless infrastructure is a costly and dynamic problem, and the system provides suggestions and feedback on tuning and parameters for better and more efficient use of radio spectrum across protocols and technologies to meet the needs of various traffic flows in the infrastructure. Thus, the solution enables automation of costly network planning and troubleshooting, which provides significant cost savings in terms of operational expenditure (OPEX) and productivity.

[0027] Figure 1An example embodiment of a system 100 for automatically tuning wireless infrastructure is illustrated. In the illustrated embodiment, the wireless infrastructure includes four wireless networks 102a-d based on multiple heterogeneous wireless technologies, including WirelessHART 102a, Wi-Fi 102b, 5G cellular network 102c, and Bluetooth Low Energy (BLE) 102d.

[0028] Wireless deployments on the edge are becoming increasingly heterogeneous in terms of the various wireless standards and protocols in the infrastructure, and they are also susceptible to various potential external sources of interference (e.g., personal devices, ambient environmental factors such as physical obstructions or weather) that can adversely affect overall performance. These technologies are typically designed without coexistence concepts, such as cross-technology coordination to prevent interference and boost performance.

[0029] In the illustrated example, the system 100 addresses the problem from a system-level perspective, using artificial intelligence and machine learning (AI / ML) models to take a cross-protocol view across multiple technologies for recommendations, thus more effectively addressing multiple interference and performance degradation issues.

[0030] The solution uses network quality indicators from multiple layers of the wireless protocol stack, optionally combined with external RF transceivers for spectrum sensing, spatio-temporal data, and traffic flow targets, as input fed into an AI / ML recommendation engine to assist in accomplishing the following tasks: (i) tuning by wireless technology to reduce interference and strive to meet traffic demands of relevant traffic flows; and (ii) spatially pinpointing sources of interference, internal or external, that can require support from an administrator to continue to determine root causes.

[0031] In the illustrated embodiment, multiple wireless networks 102a-d based on heterogeneous wireless technologies (e.g., WirelessHART, Wi-Fi, 5G, BLE) are deployed in the environment, with at least some of the networks potentially operating on the same, adjacent, or overlapping frequency bands.

[0032] The system 100 includes a local feedback and tuning pipeline for each wireless network 102a-d that performs first-level filtering and analysis of input performance data for the respective wireless network 102a-d. In the illustrated embodiment, the local feedback and tuning pipeline includes ingestion / filtering 104, AI analysis / recommendation 106, and normalization 108 subsystems.

[0033] The ingestion and filtering subsystem 104 collects various quality of service (QoS) indicators across multiple layers of the protocol stack for the corresponding wireless network 102a-d. Examples of types of indicators that can be collected include, but are not limited to:

[0034] (i) signal quality data (e.g., received signal strength indication (RSSI), estimated signal-to-noise ratio (SNR));

[0035] (ii) general demodulation errors;

[0036] (iii) data transmission errors (e.g., bit error rate (BER), block error rate (BLER), packet collisions, packet error rate (PER), checksum errors at network / transport layer, before and after applying any forward error correction (FEC) algorithm);

[0037] (iv) traffic performance errors and statistics (e.g., jitter, latency, packet loss, packet drops, transmit (TX) / receive (RX) statistics, retransmission statistics, congestion window size, sliding window size).

[0038] Generally, these QoS metrics can encompass all relevant layers of a particular wireless technology protocol stack, including the physical layer or Layer 1 (LI) (e.g., signal quality data such as RSSI / SNR, demodulation error metrics), the data link layer (e.g., error detection / correction metrics, packet collisions), the network layer (e.g., Internet Protocol (IP) checksum errors, packet loss), and the transport layer (e.g., Transmission Control Protocol (TCP) / User Datagram Protocol (UDP) checksum errors, packet loss, retransmission statistics, congestion window size).

[0039] However, the specific QoS metrics collected for each wireless network 102a-d can differ from one wireless technology to another (e.g., based on how the technology works, availability of certain metrics and counters, etc.). In particular, certain metrics can be unique to a particular wireless technology, while other metrics can be common to some or all wireless technologies. For example, block error rate (BLER), which is defined as the ratio of blocks that are transmitted in error to the total blocks transmitted, is unique to 3GPP cellular technologies. In contrast, metrics for common network and transport layer protocols such as IP, TCP, and UDP can be common to any wireless technology that uses these protocols.

[0040] The QoS metrics are fed into the local AI analytics and recommendation subsystem 106, which analyzes the metrics and generates local feedback and tuning suggestions for the respective wireless networks 102a-d. These tuning suggestions are designed to reconfigure / re-calibrate how the applicable wireless networks 102a-d should operate in order to improve performance. Thus, these tuning suggestions are optionally fed back directly into the relevant wireless infrastructure to re-calibrate the respective wireless networks 102a-d.

[0041] Examples of local feedback and tuning suggestions include, but are not limited to, adjustments to any of the following parameters or settings:

[0042] (i) modulation settings;

[0043] (ii) transmission power settings (e.g., TX power adjustments);

[0044] (iii) operating channel settings (e.g., switching to an alternative operating channel);

[0045] (iv) radio resource management (RRM) settings; and

[0046] (v) QoS control settings (e.g., queuing policies and priorities for different traffic flows, traffic steering, bandwidth allocation, technology-specific QoS controls).

[0047] The specific tuning recommendations for each wireless network 102a-d can differ from wireless technology to wireless technology (e.g., based on how the technology works, availability of specific parameters, settings, or features). In particular, certain tuning parameters can be unique to a particular wireless technology, while other tuning parameters can be common to some or all wireless technologies.

[0048] As an example, tuning recommendations for Wi-Fi can recommend changing the operating channel to avoid interference, but this is not feasible for Bluetooth Low Energy (BLE) because BLE uses an explicit frequency hopping scheme that is determined based on connection time.

[0049] As another example, tuning recommendations for Wi-Fi and cellular technologies (e.g., 4G / 5G) can recommend selecting different Modulation Coding Scheme (MCS) indices to indirectly change the modulation scheme, coding scheme, and / or related parameters such as bandwidth, guard interval (e.g., amount of time between consecutive symbols), number of spatial streams (e.g., Multiple Input Multiple Output (MIMO)), and coding range (e.g., how much FEC to apply for the purpose of being able to detect and correct errors).

[0050] After first-level filtering and analysis on a per-wireless network basis, the QoS metrics collected for each wireless network 102a-d are normalized by the normalization subsystem 108 in the local pipeline for each wireless network 102a-d.

[0051] Subsequently, the normalized metrics for all wireless networks 102a-d are fed into the AI analysis engine 120, optionally in conjunction with other sources of information such as RF spectrum data, spatio-temporal data, and performance targets.

[0052] RF spectrum data provides information about the RF spectrum currently in use in the environment, such as frequency bands or channels in use and their associated power / energy levels. For example, in some embodiments, external RF transceivers 112a-b (e.g., independent of wireless infrastructures 102a-d) can be deployed within the entire environment or coverage area of the wireless infrastructure. These RF transceivers 112a-b perform spectrum sensing to detect RF activity in the environment (e.g., by measuring the amplitude and frequency of detected RF signals), and then the RF activity is analyzed by spectrum analysis subsystem 114 to detect potential sources of interference, which can assist in switching certain wireless networks 102a-d to a backup channel. Further, in some embodiments, demodulation subsystem 116 can perform first layer (LI) demodulation on select RF signals for known wireless technologies (e.g., signals associated with known wireless networks 102a-d) to extract more detailed information about the modulated signals, enabling more effective tuning recommendations. Subsequently, the RF spectrum data and / or demodulation data can be normalized by normalization subsystem 118 and fed into AI engine 120.

[0053] Spatiotemporal data 110 includes spatial and / or temporal data indicating how wireless network 102a-d performance can vary based on time and location. For example, network performance and interference conditions can vary across the wireless infrastructure coverage area over time and space based on various factors, including but not limited to network layout (e.g., location of nodes in the wireless infrastructure such as base stations or access points, resource capacity and coverage area, and relative distance between nodes), network load (e.g., amount of traffic, number / density / location of user equipment (UE) and other connected devices), scheduling (e.g., time of day, day of week, weekday vs. weekend, holidays, business hours, production schedule, rush hour, and other vehicle traffic), events (e.g., sporting events, concerts), personnel in the environment, physical obstructions, weather conditions (e.g., conditions such as fog, clouds, rain, lightning, etc. can reduce signal strength and / or cause signal interference), external factors, etc. Thus, spatiotemporal data 110 provides information about various factors that can have an impact on the performance of some or all of wireless networks 102a-d at a certain time and / or location, such as network layout, historical network load information, user equipment (UE) / device locations from location services, weather conditions from weather services, vehicle traffic patterns from navigation services, geographic terrain from mapping services, calendar / scheduling information about weekdays / personal days (e.g., day of week, weekday vs. weekend, holidays), business hours, business schedule, events, etc.

[0054] As an example, network nodes associated with the wireless networks 102a-d, such as access points, base stations, and endpoint devices connected to these nodes, are distributed in different locations throughout the environment covered by the wireless infrastructure. Based on the respective locations of the network nodes and endpoint devices, certain locations in the environment can be more susceptible to interference and / or poor signal quality than others. Thus, knowing the locations of the network nodes and endpoint devices and their relative distances from each other can be helpful in generating tuning recommendations to improve overall performance. Accordingly, in some cases, such information can be included in the spatiotemporal data 110.

[0055] As another example, certain external factors can be present and active at particular times that require adaptation, such as a large number of terminal devices in a dense area (e.g., workers' cell phones or other wireless devices during business hours or crowds of people at events / attractions), adverse weather conditions affecting signal quality, etc. In manufacturing scenarios, for example, different processes can be active at different times based on a schedule, which can activate different wireless terminal devices. In some cases, such enterprise scheduling information can be provided by fleet and field planning management systems. Similar types of scheduling information can be provided for wireless infrastructure deployed in crowded areas such as sports venues, event sites, shopping centers, etc. Accordingly, in some cases, such scheduling information can be included in the spatiotemporal data 110.

[0056] The performance goals 111 provide guidance in how to tune the wireless infrastructure for overall target functionality for all wireless networks 102a-d and related wireless technologies. For example, in some embodiments, the performance goals 111 can include traffic flow goals that prioritize various traffic flows across all wireless networks 102a-d.

[0057] In some cases, certain wireless networks 102a-d, wireless technologies, and / or traffic flows can have higher priority or be more important than others, e.g., from a quality of service perspective. As an example, WirelessHART 102a can be more critical than Wi-Fi 102b and can be driven by static or dynamic scheduling. Thus, such prioritization should be considered when generating tuning recommendations for the wireless networks 102a-d. In some cases, for example, overall priority across the wireless networks 102a-d in a deployment can influence physical layer or data link layer settings for certain networks 102a-d, such as modulation settings or transmit power (e.g., increasing transmission power for wireless networks 102a-d with high priority or high priority traffic flows). Accordingly, in some cases, these priorities can be specified as performance goals 111, which are considered when generating tuning recommendations across the wireless networks 102a-d.

[0058] Throughout this disclosure, various types of input data sources, such as QoS metrics, RF spectrum and demodulation data, spatio-temporal data, and performance targets, are referred to as “performance data.” This performance data is fed into the AI engine 120, which is trained to output technology-specific recommendations and parameter tuning. For example, in some embodiments, the AI engine 120 outputs network calibration and tuning recommendations 122 for each wireless network 102a-d based on the performance data, while identifying any spatial interference sources or other anomalies 124.

[0059] The types of tuning recommendations 122 output by the AI engine 120 can be similar to the tuning recommendation types identified earlier with respect to the local AI analysis / recommendation engines 106 for the individual wireless networks 102a-d. However, the tuning recommendations 122 output by the AI engine 120 can seek to balance overall priorities across all wireless networks 102a-d and wireless technologies in the deployment, rather than simply improving performance of each wireless network 102a-d in isolation.

[0060] For example, based on respective priorities across the networks 102a-d, the tuning recommendations 122 can define traffic classes that provide a higher degree of quality of service for certain traffic flows, wireless networks 102a-d, and / or wireless technologies, which can require adjustment of certain settings specific to the technology for QoS control. Examples of such QoS control include, but are not limited to, 5G network exposure functions (NEF) to provide quality of service to specific devices / flows, Wi-Fi 6+ QoS support, and QoS control at the converged IPv6 / IPv4 layer (e.g., through the Differentiated Services Code Point (DSCP) field in the IP header, bandwidth allocation, and queuing policies and priorities).

[0061] Further, the tuning recommendations 122 can be fed back to the wireless infrastructure to recalibrate or reconfigure the applicable wireless networks 102a-d. In some cases, the tuning recommendations can address, in whole or in part, any interference sources or anomalies 124 that have been identified. Alternatively or additionally, the interference sources or anomalies 124 can be provided to the management system 126 to enable a system administrator to perform further troubleshooting, identify root causes, and / or implement solutions.

[0062] In various embodiments, the AI engines 106, 120 can use any type and / or combination of artificial intelligence and / or machine learning techniques to generate tuning recommendations and identify potential sources of interference or anomalies, including but not limited to decision tree learning (e.g., decision trees, random forests, classification and regression trees (CART)), gradient boosting (e.g., extreme gradient boosting trees), clustering (e.g., k-nearest neighbors (kNN), Gaussian mixture models (gMM), k-means clustering), Bayesian networks, Naive Bayes, support vector machines (SVM), artificial neural networks (ANN), feedforward artificial neural networks, deep learning, deep neural networks, recurrent neural networks (RNN) (e.g., long short-term memory (LSTM) networks), Transformers, convolutional neural networks (CNN), moving average models, autoregressive moving average (ARMA) models, autoregressive integrated moving average (ARIMA) models, exponential smoothing models, regression analysis models (e.g., logistic regression), and / or ensembles of the above (e.g., models that combine predictions of multiple machine learning models to improve prediction accuracy), etc. Examples of various types of AI / ML used to implement the AI engines 106, 120 will be further described in FIGS. 2-6.

[0063] Artificial intelligence (AI) has been an important driver of the digital transformation process for service providers in the telecommunications industry, providing service providers with the insight and ability they need to become more agile and take a more software-centric approach to their role.

[0064] As a subset of AI, machine learning (ML) is a powerful tool that can be used for network anomaly detection via scientific analysis of traffic samples (this process varies by different types or classes of ML). The ability of ML to learn and analyze and then make decisions frees people from manually processing large amounts of data. Further, ML typically responds to abnormal behavior faster than a human, which is advantageous for early detection. ML can create different models using various algorithms, but the way these models are developed and trained also has a large impact on model performance.

[0065] Even if the same model is run to detect the same type of attack, the results can vary depending on the features considered for model training. In fact, in many cases, the most difficult step in ML model development is data preparation, as the output of the ML model is heavily dependent on the data that the algorithm learns from to distinguish between normal operation and abnormal behavior.

[0066] The present disclosure proposes various ML algorithms that can be used to implement ML models for anomaly detection in different network environments. Generally speaking, most types of machine learning can be divided into supervised learning or unsupervised learning. The typical model development and deployment process for supervised learning and unsupervised learning are as follows, respectively: Figure 2A and2B are shown.

[0067] For example, Figure 2A A model development and deployment pipeline 200 for supervised learning is shown. Supervised learning models learn from past experiences (e.g., represented by labeled training data) and use the learned knowledge as a reference to make decisions on future unknown (e.g., unlabeled) data. A labeled dataset is used to train the model, and then the model is validated using validation data to tune various hyperparameters. Supervised learning techniques are commonly applied to classification and regression problems, such as spam classification and weather prediction.

[0068] In the illustrated example, the supervised learning pipeline 200 includes data preparation 201 (e.g., preparing a training dataset, including feature selection, collecting data samples, and pre-processing / filtering data samples), algorithm selection 202 (e.g., selecting an appropriate supervised learning algorithm for a particular use case), model training 203 (e.g., training a model using the selected ML algorithm(s) and training dataset), model evaluation 204 (e.g., evaluating model performance using validation or test data and tuning any hyperparameters 206 to improve performance), and prediction / inference 205 (e.g., using the trained model to provide predictions or inferences based on unlabeled or live data).

[0069] Figure 2B A model development and deployment pipeline 210 for unsupervised learning is shown. Unsupervised learning models attempt to learn from unlabeled data by discovering hidden structures and patterns in the data samples to group them. This approach can be used for clustering or feature selection to identify relevant features in a dataset.

[0070] In the illustrated example, the unsupervised learning pipeline 210 includes data preparation 211 (e.g., feature selection, preparing unlabeled data samples), algorithm selection 212 (e.g., selecting an appropriate unsupervised learning algorithm for a particular use case), and processing and output 213 (e.g., using the model to provide predictions or inferences by identifying patterns, relevant features, and / or groupings of the unlabeled data samples).

[0071] Figure 3 A processing pipeline 300 for automatically tuning wireless infrastructure is shown. In some embodiments, for example, the processing pipeline 300 can be used to automatically tune wireless infrastructure having multiple wireless networks based on heterogeneous wireless technologies.

[0072] In the illustrated embodiment, the processing pipeline 300 receives various types of input or performance data 302 from different sources, such as quality of service (QoS) metrics for wireless networks deployed in the infrastructure (e.g., performance metrics across multiple layers of a protocol stack used by the wireless networks using respective wireless technologies), radio frequency (RF) data (e.g., frequency bands currently in use in the RF spectrum and / or their associated energy levels), spatio-temporal data (e.g., context information about the performance of certain wireless network(s) based on time and location), modulation and coding scheme (MCS) indices for various wireless technologies (e.g., MCS indices for Wi-Fi, cellular), and performance targets (e.g., traffic flow targets).

[0073] The input data 302 is initially collected for training purposes (e.g., to train AI / ML models to detect anomalies and / or provide tuning suggestions for wireless networks in the infrastructure). However, the same types of input data 302 are subsequently used for inference purposes (e.g., to use the trained model(s) to detect anomalies and / or provide tuning suggestions based on live or real-time input data 302).

[0074] In terms of the training pipeline, a pre-processing / feature extraction task 304 is performed to pre-process the input data 302 (also referred to as training data 302 in this context) and extract appropriate features to be used to train the model. For example, the collected input data 302 in its current form can not be suitable for training the model. Thus, various types of pre-processing can be performed, such as data deduplication, data balancing, normalization, adding timestamps, etc.

[0075] For example, duplicate data samples can be removed from the input data 302 to avoid bias in the trained model and to require data balancing across all output classes (e.g., anomaly, non-anomaly).

[0076] Individual features in the input data 302 can be normalized by converting them into a format that can be understood by the model. For example, since machine learning models typically require inputs to be represented in a numerical format, any features with values represented in a string format must be converted to a numerical representation.

[0077] Further, a timestamp can be added as a feature for each data sample in the input data 302. For example, in terms of network-related anomaly detection, scenarios that are true positives during the day can be false positives at night. Thus, time can be used as one of the features for training the model to reduce false positives and improve overall accuracy.

[0078] Further, the data samples in the input data 302 should only include features that are important for the purpose of training the model (e.g., highly relevant to the predictions / inferences the model will be trained to make). Thus, important features can be extracted from the input data 302, and the rest of the unimportant features can be discarded. In some embodiments, unsupervised learning can be used to identify relevant attributes for feature selection. For example, if two attributes are highly correlated with each other, they are linearly dependent on each other, and thus either one of these attributes can be selected as a feature while the other can be discarded.

[0079] Next, a model development and training task 306 is performed to train the model using the prepared training data 302 and appropriate ML algorithm(s). In some embodiments, since the input data 302 can include various features with different numerical ranges and different magnitudes, multiple models can be trained using different types of ML algorithms and different subsets of the input data 302.

[0080] For example, a quality of service (QoS) metric contains discrete performance metrics (e.g., signal strength, signal-to-noise ratio, packet loss, jitter, latency, etc.), each of which can be treated as a single feature. Thus, in some embodiments, a decision tree algorithm can be used to train a model based on the QoS metrics (e.g., as described below in connection with Figure 5

[0081] In contrast, for radio frequency (RF) data, the model can be trained using RF data captured over a period of time. For such time-series data, a recurrent neural network (RNN) supervised learning model, such as a long short-term memory (LSTM) network (e.g., as described below in connection with Figures 6A-6B

[0082] Next, a model evaluation task 308 is performed to validate and test the model. Validation is performed using a validation dataset, which contains data samples that were reserved during training. For example, the validation dataset can contain the same type of data as the training dataset 302, but the validation dataset can only include data samples that were omitted from the training dataset 302 and were not used to train the model. The validation dataset is used to evaluate the performance of the model while also tuning its hyperparameters to improve performance. For example, based on the prediction accuracy of the model in processing the validation dataset, certain hyperparameters of the model can be tuned, and the model can be retrained using the updated parameters.

[0083] The fully trained model is tested using a test dataset, which contains data samples that were reserved during training and validation and are thus completely new and unknown to the model. The test dataset is used to determine an unbiased estimate of the final model performance.

[0084] ​​Next, a model deployment task 310 is performed to deploy the model for providing real-time predictions or inferences using live input data 302.

[0085] Once the model is deployed, an anomaly detection task 312 is performed to detect anomalies 316 and / or recommend performance adjustments 314. For example, the anomaly detection task 312 uses the trained model to provide real-time predictions and inferences about the source of potential anomalies / interferences 316 and recommended performance adjustments 314 based on live input data 302 (e.g., real-time captured unlabeled / unserved input data).

[0086] In the illustrated embodiment, if no anomalies are detected, the anomaly detection task 312 can recommend performance adjustments 314 for one or more wireless networks to improve performance (e.g., suggestions for correct network calibration, configuration settings adjustments).

[0087] If anomalies are detected, the anomaly detection task 312 can identify a predicted spatial interference source 316 causing the anomaly. The predicted interference source 316 can then be reported to a human-monitored management system 318 to assist in troubleshooting the anomaly.

[0088] It is also beneficial for the model to continue learning so that it stays up-to-date and is aware of the latest traffic behavior. For example, in real-time scenarios, the model can encounter new data types / patterns that it did not know about or see during training, as it is not possible for the model to be trained with all possible data sets and scenarios the first time around. Therefore, the model preferably should be able to learn from new data and new scenarios it encounters during its real-time predictions to be up-to-date about the latest network behavior and patterns.

[0089] Accordingly, in some embodiments, continuous learning techniques can be applied to ensure that the model stays up-to-date. For example, the model parameters can be recomputed for any new input data that the model receives while providing real-time predictions. In this way, when the model receives new data that is different from the data on which it was trained, the model can learn from that data so that it makes better decisions when it encounters similar data in the future.

[0090] In the illustrated example, the input data 302, tuning suggestions 314, and identified interference sources / abnormalities 316 are used for continual learning. For example, with assistance from a management system 318 from human monitoring, a monitoring and labeling task 320 is performed to (i) monitor the real-time input data 302 and corresponding predictions 314, 316 generated by the model, and (ii) label the input data 302 with ground truth (e.g., abnormal or non-abnormal, performance adjustment) for retraining / updating the model for continual learning purposes. The newly labeled input data 302 is then added to the original training data 302, and the training portion of the pipeline 300 (e.g., 302-310) is repeated to retrain the model using the updated training data 302.

[0091] A variety of continual learning methods can be employed, such as brute-force update, periodic update, event-driven update, and online update.

[0092] Brute-force update provides the simplest solution, where the model parameters are simply recomputed based on the latest data window whenever a new data point arrives. However, this approach can not be feasible in some cases if the computational complexity of fitting the model to the data window is too high.

[0093] In the case of periodic update, the model parameters are cached for a specific time period (e.g., 24 hours) and the model is retrained based on the new data points at the end of each time period. However, if the behavior of the metric changes before each periodic update, there can be excessive false positive alerts.

[0094] In the case of event-driven update, the model parameters are recomputed when a high prediction error is detected for the latest set of data points. However, the timing of event-driven update can be unpredictable, which can cause operational challenges in the future.

[0095] In the case of online update, some machine learning algorithms can be reformulated to work in an online setting, continuously reading new data points and effectively updating the parameters with each data point.

[0096] While any of the above approaches can be employed, brute-force update and online update can be the most suitable for such application scenarios due to the drawbacks of periodic update and event-driven update.

[0097] Figure 4A processing pipeline 400 for continual learning is illustrated. In the illustrated example, the continual learning pipeline 400 includes data pre-processing 404 (e.g., pre-processing training data 402), model training and validation 406 (e.g., training a model using training data 602 and appropriate ML algorithm(s)), model deployment 408, prediction / inference 410 (e.g., using a trained / deployed model to provide predictions / inferences based on unlabeled / live data), monitoring 412 (e.g., monitoring input data samples and predictions generated by the model), and data cleaning and labeling 414 (e.g., cleaning / labeling new data samples to retrain / update the model). The newly cleaned and labeled data is then added to the training data 402, and the process 400 is repeated to retrain the model using the updated training data 402.

[0098] Figure 5 An example of a decision tree model 500 for anomaly detection using quality of service (QoS) metrics is illustrated. In some embodiments, for example, as described below, the decision tree model 500 can use Figure 3 The process 300 uses QoS metrics for training.

[0099] First, the QoS metrics are divided into three datasets: a training dataset, a validation dataset, and a testing dataset.

[0100] The decision tree 500 is initially trained with the training dataset. During the training process, the entire training dataset is initially considered at the root node. At each iteration of the algorithm, the entropy (H) and information gain (IG) are calculated for each attribute.

[0101] Entropy is a measure of the randomness of the information being processed. The higher the entropy, the more difficult it is to draw any conclusions from that information. Entropy can be calculated using the following formula:

[0102]

[0103] where S is the current state, p i is the probability of event i in state S or the percentage of class i in the node of state S.

[0104] Information gain (IG) is the reduction in entropy. IG is a statistical property that measures how well a particular attribute separates the training samples according to their target classification. IG can be calculated using the following formula:

[0105]

[0106] where “before” is the dataset before the split, k is the number of subsets generated by the split, and (j, after) is subset j after the split.

[0107] The attribute with the smallest entropy or the largest information gain is selected as the next node that splits the data set of its parent node into two or more subsets.

[0108] The process continues in this way, with each iteration considering only the remaining attributes that were not selected in previous iterations, until a stopping criterion is met. In the final tree, the number of leaf nodes is equal to the number of output classes, and each leaf node contains a subset of the training data corresponding to its respective output class.

[0109] In the illustrated example, the QoS metrics used to train the decision tree 500 for anomaly detection include bandwidth, jitter, packet loss, and delay, as shown in Table 1. In particular, each row of Table 1 represents a data sample in the training data set, and each data sample has values for the bandwidth, jitter, packet loss, and delay metrics, as well as a true value indicating whether an anomaly is present.

[0110]

[0111]

[0112] Table 1: QoS metrics data set for decision tree model training

[0113] Based on the data samples in Table 1, the information gain for jitter is higher than the other attributes, so jitter is selected as the root node. Similarly, for each value of jitter, the information gain is computed for the remaining attributes using the process described above until a stopping criterion is met. The decision tree 500 is then validated and tested using validation and test data sets.

[0114] Figure 5 The generated decision tree 500 is shown in FIG. 5. At the root node 502, jitter is evaluated.

[0115] If jitter is low at node 502, the tree branches to node 504, where packet loss is evaluated. If packet loss is high at node 504, the tree branches to leaf node 508, where no anomaly is detected. If packet loss is low at node 504, the tree branches to leaf node 510, where an anomaly is detected.

[0116] If jitter is medium at node 502, the tree branches to leaf node 510, where an anomaly is detected.

[0117] If jitter is high at node 502, the tree branches to node 506, where delay is evaluated. If delay is high at node 506, the tree branches to leaf node 508, where no anomaly is detected. If delay is low at node 506, the tree branches to leaf node 510, where an anomaly is detected.

[0118] In real-time scenarios where the decision tree model 500 is used to process live QoS data, the model 500 predicts whether there is an anomaly and, if there is, what the spatial source of the anomaly is. Information about the anomaly can be sent to a system administrator to perform further troubleshooting and corrective actions. If there is no anomaly, the model can suggest the correct network calibration.

[0119] Figures 6A-6B An example of a recurrent neural network (RNN) is illustrated, which can be used to detect interference and anomalies based on radio frequency (RF) spectrum data and / or provide tuning recommendations for wireless networks. In particular, Figure 6A A standard RNN 600 is illustrated, while Figure 6B A long short-term memory (LSTM) network 610 is shown.

[0120] The RNN 600 is a type of deep learning artificial neural network designed to process sequential data using the concept of “recurrence.” For example, RNNs capture and utilize context from previous inputs with the help of internal memory / state and feedback connections. This enables RNNs to capture dependencies and patterns in sequential data, which makes them suitable for tasks such as language modeling, speech recognition, machine translation, sentiment analysis, and time series prediction.

[0121] Unlike feedforward neural networks, which process data strictly in order from input to output, RNNs have feedback connections that allow information or context to be shared across multiple inputs. For example, while feedforward neural networks establish weighted connections between layers, RNNs establish weighted connections between layers and between neurons of the same layer (e.g., where the output of certain nodes serves as input to those same nodes), which enables RNNs to learn from time series data.

[0122] The basic building block of an RNN is a recurrent neuron, which receives an input as well as the output of the previous step. This enables the neuron to incorporate information from earlier steps into its current computation. The output of each recurrent neuron is then fed back into the network, creating a loop that enables the network to maintain and update its internal memory while processing sequential data.

[0123] The LSTM network 610 is a type of RNN that is capable of learning long-term dependencies in sequential data, which is particularly useful for sequence prediction problems. LSTM networks use memory cells and gating mechanisms to capture and retain information in long data sequences. For example, each LSTM layer is composed of multiple LSTM units, and each unit typically includes a memory cell, an input gate, a forget gate, and an output gate. The memory cell remembers a value over an arbitrary time interval, and the three gates regulate the flow of information into and out of the cell.

[0124] The input gate determines which information in the current input and previous hidden state should be stored in the memory cell. The forget gate determines which information should be discarded from the memory cell. The output gate determines which information in the memory cell should be output to the next hidden state.

[0125] These gates are learned during training, allowing the LSTM network to adaptively decide which information to keep or forget at each step. The presence of these gates and memory cells within the LSTM network makes them particularly effective at learning long-term dependencies in sequential data.

[0126] The following equations define the input gate (i), forget gate (f), output gate (o), candidate cell state (g) (e.g., new information to pass to the cell state), cell state (C), and cell output (h) for an LSTM unit, where σ is the logistic sigmoid function, W are weight matrices, b are bias vectors, x is the input vector, and t is the time:

[0127] i t = σ(W hi h t-1 + W xi x t + b i )

[0128] f t = σ(W hf h t-1 + W xf x t + b f )

[0129] o t = σ(W ho h t-1 + W xo x t + b o )

[0130] g t = tanh(W hC h t-1 + W xC x t + b C )

[0131] C t = f t C t-1 + (1 - f t )g t

[0132] h t = o t tanh(C t )

[0133] As described above, LSTM models can be used to process radio frequency (RF) data to detect interference and other anomalies and provide tuning recommendations for wireless networks to improve performance. In this use case, the input vector x contains radio frequency data samples, where the length of the vector is equal to the number of timestamps (t) or samples considered for training the model.

[0134] Figure 7 An example of a protocol stack 700 for network technology is illustrated. Protocols for network and communication technology are often depicted as“layers” 702 in the protocol stack 700 because each protocol layer 702 has a well-defined function (e.g., based on standardized communication protocols) and typically communicates or interacts with protocols in directly adjacent layers 702. For example, the lowest protocol layer (e.g., the physical layer) typically handles the transmission of raw bits or symbols across a physical transmission medium (e.g., radio waves, wires / cables), while the highest protocol layer (e.g., the application layer) typically provides application-level functionality.

[0135] In the illustrated example, the protocol stack 700 is based on the Open Systems Interconnection (OSI) model, which abstracts network and communication protocols into the following protocol layers 702: physical layer, data link layer, network layer, transport layer, session layer, presentation layer, and application layer. The respective layers 702 of the protocol stack 700 and the corresponding data units 704 are described in Table 2 below.

[0136] For wireless network and communication technology (e.g., cellular, Wi-Fi, Bluetooth, LR-WPAN), the actual wireless transmission is typically implemented in the lowest layers 702 of the protocol stack 700, such as the physical layer and the data link layer, using radio waves as the physical layer transmission medium.

[0137]

[0138] Table 2: Example Protocol Stack

[0139] It should be appreciated that the protocol stack 700 is just one example method for representing the respective protocols for network technology and their relationships. In other protocol stacks, certain layers can be omitted, added, merged, or split. For example, in the Internet Protocol Suite, the physical layer and the data link layer of the OSI model are merged into one link layer, while all layers above the transport layer in the OSI model (e.g., the session layer, the presentation layer, and the application layer) are merged into one application layer.

[0140] Figure 8A flowchart 800 for automatically tuning wireless infrastructure is illustrated in accordance with certain embodiments. The process flow can be performed using any type or combination of circuitry, including but not limited to processing circuitry (e.g., CPUs, GPUs, AI / ML accelerators, digital signal processors), interface circuitry, communication circuitry, and / or transceiver circuitry, which can be part of a single device or distributed across multiple devices. In some embodiments, the flowchart 800 can be performed using the example computing devices and systems described throughout this disclosure, such as the computing devices 1100, 1150 of the system 100, Figures 11A-11B

[0141] In the illustrated example, the process flow is for automatically tuning wireless infrastructure. The wireless infrastructure can include network nodes and devices (e.g., access points, base stations, network devices, and servers) associated with a plurality of wireless networks deployed in an environment. In some embodiments, the wireless networks can be based on various heterogeneous wireless technologies, including but not limited to cellular (e.g., 4G, 5G), Wi-Fi, Bluetooth, and / or IEEE 802.15.4 Low-Rate Wireless Personal Area Networks (LR-WPANs) (e.g., WirelessHART, Zigbee, ISA100.11a, MiWi, 6LoWPAN, Thread, SNAP). For example, in some cases, the infrastructure can include networks based on cellular technologies, Wi-Fi, Bluetooth, and / or IEEE 802.15.4 LR-WPAN technologies, or a subset or superset thereof. A wireless technology can refer to any technology used for communicating at least in part via wireless transmissions, such as radio frequency (RF) transmissions via certain frequency ranges or bands of the RF spectrum. Different wireless technologies can have respective different frequency spectrums, which can or can not overlap. A wireless network can refer to any communication mechanism that relies at least in part on a wireless technology.

[0142] The process flow begins at block 802 by receiving performance data for respective wireless networks deployed in an environment. In some embodiments, for example, the performance data can be received from various sources (e.g., via interface circuitry) and can include various types of information indicative of respective wireless network performance. Moreover, the performance data can be based on multiple protocol layers of respective wireless technologies. For example, the wireless technologies used to implement the various wireless networks can be based on corresponding protocol stacks, and the performance data can be based on multiple protocol layers from different levels of the protocol stacks, such as physical layers, data link layers, network layers, and / or transport layers.

[0143] ​As an example, for an infrastructure having cellular, Wi-Fi, and Bluetooth networks, performance data can be based on: the physical layer and data link layer for the cellular network; the physical layer, data link layer, network layer, and transport layer for the Wi-Fi network; and the physical layer for the Bluetooth network. It should be understood that this combination of wireless technologies and protocol layers is provided by way of example only, as performance data can be based on any combination of wireless technologies and protocol layers in an actual embodiment.

[0144] In some embodiments, for example, performance data can include quality of service (QoS) indicator(s) for some or all of the wireless networks, radio frequency (RF) spectrum data, and / or space-time performance data.

[0145] QoS indicators can indicate information such as signal quality (e.g., signal strength, signal-to-noise ratio), demodulation error, data transmission error, network layer performance, and / or transport layer performance.

[0146] RF spectrum data can indicate frequency bands in the RF spectrum that are currently in use in the environment and / or their associated energy levels (e.g., based on analysis of RF signals detected at various frequencies by one or more RF receivers deployed in the environment, which can be independent of the wireless networks).

[0147] Space-time performance data can include various types of contextual information that indicates performance of a wireless network(s) based on time and place, including but not limited to network layout, historical network load, device location and density, calendar / scheduling information (e.g., work / personal day, business hours, business schedule, events), vehicle traffic patterns, geographic terrain (e.g., mountains, buildings), and weather conditions. Such space-time data can be provided using any appropriate granularity of time and / or location. For example, in terms of time, certain space-time data can be provided based on time of day, day of week, weekday versus weekend, holiday, etc. In terms of location, certain space-time data can be provided based on government-defined boundaries (e.g., address, street, community, city, zip code, state, region, country), geographic boundaries (e.g., mountains, rivers, lakes, oceans), location coordinates (e.g., latitude / longitude), and / or coverage areas (e.g., coverage areas of particular networks, base stations, access points, etc.).

[0148] The process flow then proceeds to block 804 to determine recommended performance adjustments based on the performance data. In some embodiments, for example, the performance data can be used to determine one or more configuration settings to adjust for one or more of the wireless networks. For example, the one or more configuration settings can include one or more modulation settings, transmission power settings, operating channel settings, radio resource management settings, and / or quality of service settings.

[0149] In some embodiments, one or more configuration settings may be determined from performance data using a machine learning (ML) model trained to infer configuration adjustments based on past performance data and / or past configuration data for wireless networks and / or wireless technologies.

[0150] Furthermore, in some embodiments, one or more configuration settings may also be determined based on one or more performance objectives, such as traffic flow(s) objectives that indicate priority among various traffic flows on the wireless network.

[0151] The process then proceeds to box 806 to reconfigure at least some wireless networks based on recommended performance tweaks. In some embodiments, for example, configuration settings identified from or derived from the recommended performance tweaks may be adjusted for the applicable wireless networks.

[0152] The process then proceeds to block 808 to determine if an anomaly (e.g., significant signal interference) has been detected. In some embodiments, RF spectrum data and / or other performance data may be used to detect one or more sources of signal interference or anomalies targeting one or more wireless networks. For example, signal interference may be caused by transmissions from one or more devices (e.g., physically separate / independent and / or devices with physically separate / independent enclosures). In some embodiments, the source of signal interference or anomalies may be detected from RF spectrum data using an ML model trained to infer the source of interference and / or anomalies based on past RF spectrum data.

[0153] If no anomaly is detected, the process flow can complete. However, if an anomaly is detected, the process flow proceeds to box 810 to identify the anomaly(s) source and / or notify the system administrator. Further, in some embodiments, when an anomaly is detected, the process flow can return to box 804 to determine recommended performance adjustments based at least in part on the detected anomaly.

[0154] At this point, the process flow can be completed. In some embodiments, however, the process flow can be restarted and / or certain boxes can be repeated. For example, in some embodiments, the process flow can restart at box 802 to continue tuning the performance of the heterogeneous wireless infrastructure.

[0155] Calculation Example

[0156] The following provides examples of various computational embodiments that can be used to implement the wireless network tuning scheme described herein. In particular, various aspects of the scheme described above can be implemented using the embodiments described below. In some embodiments, for example, the wireless network tuning scheme can be implemented in… Figures 9-10 Edge computing environments and / or Figures 11A-11B Implemented in computing devices.

[0157] Edge computing

[0158] Figure 9 is a block diagram 900 illustrating an overview of a configuration for edge computing, which includes a processing tier referred to in many of the following examples as an "edge cloud." As shown, the edge cloud 910 is co-located at an edge location, such as an access point or base station 940, a local processing center 950, or a central office 920, and thus can include multiple entities, devices, and equipment instances. The edge cloud 910 is closer to the endpoint (consumer and producer) data sources 960 (e.g., autonomous vehicles 961, user equipment 962, commercial and industrial equipment 963, video capture devices 964, drones 965, smart city and building devices 966, sensors and loT devices 967, etc.) than the cloud data center 930. The computing, memory, and storage device resources provided at the edge in the edge cloud 910 are critical to providing ultra-low latency response times for services and functions used by the endpoint data sources 960, while also reducing network backhaul traffic from the edge cloud 910 to the cloud data center 930, resulting in benefits such as improved energy consumption and overall network usage.

[0159] Computing, memory, and storage devices are scarce resources, and are typically reduced depending on the edge location (e.g., less processing resources at a consumer endpoint device compared to at a base station, less processing resources at a base station compared to at a central office). However, the closer the edge location is to an endpoint (e.g., user equipment (UE)), the more constrained the space and power are typically. Thus, edge computing attempts to reduce the amount of resources needed for network services by distributing more resources that are geographically and network access time closer. In this way, edge computing attempts to bring the computing resources to the workload data, or the workload data to the computing resources, as appropriate.

[0160] Various aspects of an edge cloud architecture are described below, which encompasses multiple potential deployments and addresses certain limitations that can exist in a network operator's or service provider's own infrastructure. These limitations include: configuration changes based on the edge location (as the edge is at, for example, a base station level, there can be more constrained performance and functionality in a multi-tenant scenario); configuration based on the type of computing, memory, storage, fabric, acceleration, or similar resources available to the edge location, location tier, or location group; service, security, management, and orchestration capabilities; and related goals to achieve availability and performance of end services. These deployments can implement processing in the network tier, which can be considered a "near edge," "nearby edge," "local edge," "intermediate edge," or "far edge" tier depending on latency, distance, and timing characteristics.

[0161] Edge computing is an emerging paradigm in which computing is performed at or near the network "edge," typically using a computing platform (e.g., x86 or ARM computing hardware architecture) implemented at a base station, gateway, network router, or other device that is more proximate to the endpoint devices that are generating and consuming data. For example, an edge gateway server can be equipped with a pool of memory and storage resources to perform computations in real-time for low-latency use cases (e.g., autonomous driving or video surveillance) directed to connected customer devices. Or, as an example, a base station can leverage computing and acceleration resources to handle service workloads directly against connected user devices without further communication of data via a backhaul network. Or as another example, a central office network management hardware can be replaced with standardized computing hardware that performs virtual network functions and provides computing resources for execution of service and consumer functions directed to connected devices. Within an edge computing network, there can be scenarios in which computing resources are to be "moved" to the service of data, and scenarios in which data is to be "moved" to the service of computing resources. Or, as an example, base station computing, acceleration, and network resources can provide services to scale workloads on demand by activating dormant capacity (subscription, on-demand capacity) to manage extreme situations, emergencies, or to provide long-term longevity to deployed resources over a significantly longer implementation lifecycle.

[0162] Figure 10 The operational layers between endpoints, edge cloud, and cloud computing environments are illustrated. In particular, Figure 10 An example of a computing use case 1005 is depicted that leverages an edge cloud 910 between multiple illustrative layers of network computing. These layers begin at an endpoint (device and thing) layer 1000 that accesses the edge cloud 910 to perform data creation, analysis, and data consumption activities. The edge cloud 910 can span multiple network layers, such as an edge device layer 1010 with gateways, local servers, or network equipment (nodes 1015) located in physical proximity in edge systems; a network access layer 1020 that encompasses base stations, wireless processing units, network hubs, regional data centers (DCs), or local network equipment (equipment 1025); and any equipment, devices, or nodes in between (not shown in detail in layer 1012). Network communications within the edge cloud 910 and between layers can occur via any number of wired or wireless media, including via un-described connection architectures and technologies.

[0163] Examples of latency caused by network communication distance and processing time constraints can range from possibly less than one millisecond (ms) when within the endpoint tier 1000, less than 5 ms at the edge device tier 1010, to even possibly between 10 ms to 40 ms when communicating with nodes at the network access tier 1020. Beyond the edge cloud 910 are the core network 1030 and cloud data center 1040 tiers, each with increasing latency (e.g., 50-60 ms at the core network tier 1030, reaching 100 ms or more at the cloud data center tier). Thus, with at least 50 to 100 ms or more latency, operations at the core network data center 1035 or cloud data center 1045 would be unable to implement many of the time-critical functions of the use case 1005. Each of these latency values is provided for illustrative and comparative purposes; it should be understood that further reductions in latency are possible using other access network media and technologies. In some examples, various portions of the network can be categorized as being in a “near edge,” “local edge,” “nearby edge,” “intermediate edge,” or “far edge” tier relative to a network source and destination. For example, from the perspective of the core network data center 1035 or cloud data center 1045, a central office or content data network can be considered to be within a “nearby edge” tier (cloud “nearby” with a high latency value when communicating with devices and endpoints of the use case 1005), while an access point, base station, local server, or network gateway can be considered to be within a “far edge” tier (“far” from the cloud with a low latency value when communicating with devices and endpoints of the use case 1005). It should be understood that other categorizations of a particular network tier as being “near,” “local,” “nearby,” “intermediate,” or “far” edge can be based on latency, distance, network hops, or other measurable characteristics as measured from a source in any of the network tiers 1000-1040.

[0164] Due to the multiple services that utilize the edge cloud, various use cases 1005 can access resources under usage pressure of the incoming streams. To achieve results with low latency, services executing within the edge cloud 910 balance different requirements of: (a) priority (throughput or latency) and quality of service (QoS) (e.g., traffic for an autonomous vehicle can have a higher priority than a temperature sensor in terms of response time requirements; or, depending on the application, performance sensitivity / bottlenecks can exist at compute / accelerator, memory, storage devices, or network resources); (b) reliability and resiliency (e.g., some incoming streams need to be processed and traffic needs to be routed with mission-critical reliability, while some other incoming streams, depending on the application, can tolerate occasional failures); and (c) physical limitations (e.g., power, cooling, and form factor, among others).

[0165] The end-to-end service view for these use cases involves the concept of service flows and is associated with transactions. Transactions detail the overall service requirements for an entity consuming the service, and related services for resources, workloads, workflows, and business functions and business level requirements. Services performed in the described "clause" can be managed at each layer to ensure real-time and runtime contract compliance for the transaction over the service lifecycle. When a component in the transaction does not fulfill its agreed service level agreement (SLA), the overall system (components in the transaction) can be able to provide the following capabilities (1) understand the impact of the SLA breach, and (2) enhance other components in the system to restore the overall transaction SLA, and (3) implement remediation steps.

[0166] Accordingly, given these changes and service characteristics, edge computing within the edge cloud 910 can provide service capabilities and respond in real-time or near real-time for a variety of applications for the use cases 1005 (e.g., target tracking, video surveillance, connected cars, etc.) and meet ultra-low latency requirements for these variety of applications. These advantages enable a new class of applications (e.g., virtual network functions (VNFs), functions-as-a-service (FaaS), edge-as-a-service (EaaS), standard processes, etc.) that are not possible with traditional cloud computing due to latency or other limitations.

[0167] However, the advantages of edge computing come with the following limitations. Devices located at the edge are typically resource constrained, and thus there is pressure on the use of edge resources. Typically, this is addressed by pooling memory and storage resources for use by multiple users (tenants) and devices. The edge can be constrained in terms of power supply and cooling, and thus power usage by the most power consuming applications needs to be considered. There can be an inherent power-performance tradeoff in these pooled memory resources, as many of these resources can use emerging memory technologies where greater power requires greater memory bandwidth. Also, a root of trust for improved security and trusted functions of the hardware is necessary, as edge locations can be unattended and can even require licensed access (e.g., when stored at a third party location). Such issues can be more pronounced in a multi-tenant, multi-owner, or multi-access setting edge cloud 910, where services and applications are requested by many users, particularly when network usage fluctuates dynamically and the composition of multiple stakeholders, use cases, and services change.

[0168] At a more general level, an edge computing system can be described as encompassing any number of deployments of the previously discussed layers operating in the edge cloud 910 (network tiers 1000-1040), which provide coordination from client and distributed computing devices. One or more edge gateway nodes, one or more edge aggregation nodes, and one or more core data centers can be distributed across tiers of a network to provide implementations of an edge computing system by or on behalf of a telecommunications service provider (“telco” or “TSP”), an Internet of Things service provider, a cloud service provider (CSP), an enterprise entity, or any other number of entities. Various implementations and configurations of an edge computing system can be provided dynamically, such as when orchestrated to meet service objectives.

[0169] Consistent with examples provided herein, a client computing node can be embodied as any type of endpoint component, device, apparatus, or other thing capable of communicating as a data producer or consumer. Further, the term “node” or “device” used in the context of an edge computing system does not necessarily imply that such a node or device operates in a client or proxy / dependent / following role; rather, any node or device in an edge computing system refers to any entity, node, or subsystem comprising discrete or connected hardware or software configurations to facilitate or use the various entities of the edge cloud 910.

[0170] Accordingly, the edge cloud 910 is comprised of network components and functional features operated by and within the edge gateway nodes, edge aggregation nodes, or other edge computing nodes in the network tiers 1010-1030. As such, the edge cloud 910 can be embodied as any type of network that provides edge computing and / or storage resources for location proximate wireless access network (RAN) capable endpoint devices (e.g., mobile computing devices, IoT devices, smart devices, etc.), which are discussed herein. In other words, the edge cloud 910 can be conceived as an “edge” that connects endpoint devices and traditional network access points that act as entry points into a service provider core network (including mobile operator networks (e.g., Global System for Mobile Communications (GSM) networks, Long-Term Evolution (LTE) networks, 5G / 6G networks, etc.)) while also providing storage and / or computing capabilities. Other types and forms of network access (e.g., Wi-Fi, long-range wireless, wired networks including optical networks, etc.) can also be used instead of or in conjunction with such 3GPP operator networks.

[0171] The network components of edge cloud 910 can be servers, multi-tenant servers, device computing devices, and / or any other type of computing device. For example, edge cloud 910 can include a device computing device, which is a self-contained electronic device that includes a housing, chassis, enclosure, or case. In some cases, the housing can be sized for portability such that it can be carried and / or transported by a person. An example housing can include a material that forms one or more exterior surfaces that partially or fully protect the contents of the device, where protection can include weather protection, hazardous environment protection (e.g., electromagnetic interference (EMI), vibration, extreme temperatures, etc.), and / or enable submersion. An example housing can include power circuitry to provide power for fixed and / or portable implementations, such as alternating current (AC) power input, direct current (DC) power input, AC / DC converter(s), DC / AC converter(s), DC / DC converter(s), power regulators, transformers, charging circuitry, batteries, wired input, and / or wireless power input. An example housing and / or surface thereof can include or connect to mounting hardware to enable attachment to structures such as buildings, telecommunication structures (e.g., utility poles, antenna structures, etc.), and / or racks (e.g., server racks, blade installations, etc.). An example housing and / or surface thereof can support one or more sensors (e.g., temperature sensors, vibration sensors, light sensors, acoustic sensors, capacitive sensors, proximity sensors, infrared or other visual heat sensors, etc.). One or more such sensors can be contained in, carried by, or otherwise embedded in and / or mounted to a surface of the device. An example housing and / or surface thereof can support mechanical connections such as propulsion hardware (e.g., wheels, rotors (such as propellers, etc.)) and / or articulating hardware (e.g., robotic arms, rotatable appendages, etc.). In some cases, sensors can include any type of input device, such as user interface hardware (e.g., buttons, switches, dials, sliders, microphones, etc.). In some cases, an example housing includes output devices contained in, carried by, embedded in, and / or connected to the device. Output devices can include displays, touchscreens, lights, light emitting diodes (LEDs), speakers, input / output (I / O) ports (e.g., universal serial bus (USB)), etc. In some cases, an edge device is a device that is present in a network for a specific purpose (e.g., a traffic signal light) but can have processing and / or other capabilities that can be used for other purposes. Such edge devices can be independent of other network devices and can be provided with a housing that fits its primary purpose; however, can be used for other computing tasks that do not interfere with its primary task. Edge devices include Internet of Things devices. A device computing device can include hardware and software components for managing local issues such as device temperature, vibration, resource utilization, updates, power issues, physical and network security, etc. Example hardware for implementing a device computing device incorporates Figure 11BThe edge cloud 910 can also include one or more servers and / or one or more multi-tenant servers. Such servers can include an operating system and implement a virtual computing environment. The virtual computing environment can include a hypervisor that manages (e.g., spawns, deploys, debugs, destroys, deactivates, etc.) one or more virtual machines, one or more containers, and / or the like. Such virtual computing environments provide an execution environment in which one or more applications and / or other software, code, or scripts can execute in isolation from one or more other applications, software, code, or scripts.

[0172] Computing devices and systems

[0173] In further examples, any of the computing nodes or devices discussed herein with reference to current edge computing systems and environments can be implemented based on the components depicted in Figure 11A and Figure 11B The respective edge computing nodes can be embodied as a type of device, apparatus, computer, or other "thing" capable of communicating with other edge, network, or endpoint components. For example, an edge computing device can be embodied as a personal computer, a server, a smartphone, a mobile computing device, a smart appliance, a vehicle computing system (e.g., a navigation system), a self-contained device having a housing, chassis, or the like, or other device or system capable of performing the described functions.

[0174] In Figure 11A In the simplified example depicted, the edge computing node 1100 includes a compute engine (also referred to herein as "compute circuitry") 1102, an input / output (I / O) subsystem (also referred to herein as "I / O circuitry") 1108, data storage (also referred to herein as "data storage circuitry") 1110, a communication circuit subsystem 1112, and optionally one or more peripheral devices (also referred to herein as "peripheral circuitry") 1114. In other examples, the respective computing devices can include other or additional components, such as components typically found in a computer (e.g., a display, peripheral devices, etc.). Additionally, in some examples, one or more of the illustrative components can be incorporated in, or otherwise form part of, another component.

[0175] The computing node 1100 can be embodied as any type of engine, device, or collection of devices capable of performing various computing functions. In some examples, the computing node 1100 can be embodied as a single device, such as an integrated circuit, an embedded system, a field-programmable gate array (FPGA), a system on a chip (SoC), or other integrated system or device. In illustrative examples, the computing node 1100 includes or is embodied as a processor (also referred to herein as “processor circuitry”) 1104 and a memory (also referred to herein as “memory circuitry”) 1106. The processor 1104 can be embodied as any type(s) of processor capable of performing the functions described herein (e.g., executing applications). For example, the processor 1104 can be embodied as multi-core processor(s), a microcontroller, a processing unit, a special or special-purpose processing unit, or other processor or processing / control circuitry.

[0176] In some examples, the processor 1104 can be embodied as including or coupled to a field-programmable gate array (FPGA), an application-specific integrated circuit (ASIC), reconfigurable hardware, or hardware circuitry to facilitate performance of the functions described herein. Further, in some examples, the processor 1104 can be embodied as a special-purpose x-processing unit (xPU), also referred to as a data processing unit (DPU), an infrastructure processing unit (IPU), or a network processing unit (NPU). Such xPUs can be embodied as standalone circuitry or circuitry packages, integrated in a SOC, or integrated with network circuitry (e.g., in a SmartNIC or enhanced SmartNIC), acceleration circuitry, storage devices, storage disks, or AI hardware (e.g., GPUs, programmed FPGAs, or ASICs specifically designed to implement AI models such as neural networks). Such xPUs can be designed to receive, retrieve, and / or otherwise obtain programming to process one or more data streams and perform specific tasks and actions with respect to the data streams outside of a CPU or general-purpose processing hardware (e.g., host microservices, perform service management or orchestration, organize or manage server or data center hardware, manage service meshes, or collect and distribute telemetry). It will be understood, however, that xPUs, SOCs, CPUs, and other variants of the processor 1104 can work in coordination with one another to perform many types of operations and instructions within and on behalf of the computing node 1100.

[0177] Memory 1106 can be embodied as any type of volatile (e.g., dynamic random access memory (DRAM), etc.) or non-volatile storage or data storage device capable of performing the functions described herein. A volatile memory can be a storage medium that requires power to maintain the data state stored by the medium. Non-limiting examples of volatile memory can include various types of random access memory (RAM), such as DRAM or static random access memory (SRAM). One particular type of DRAM that can be used for the memory module is synchronous dynamic random access memory (SDRAM).

[0178] In examples, the memory device (e.g., memory circuit) is any number of block-addressable memory devices such as NAND or NOR technology based (e.g., single level cell (SLC), multi-level cell (MLC), quad-level cell (QLC), triple-level cell (TLC), or other NAND). In some examples, the memory device(s) include byte-addressable in-place write three-dimensional cross-point memory devices, or other byte-addressable in-place write non-volatile memory (NVM) devices such as single or multi-level phase change memory (PCM) or phase change memory with switching (PCMS), NVM devices using chalcogenide phase change materials (e.g., chalcogenide glass), resistive memory including metal-oxide based, oxygen-vacancy based, and conductive-bridge random access memory (CB-RAM), nanowire memory, ferroelectric transistor random access memory (FeTRAM), magnetoresistive random access memory (MRAM) including memristor technology, spin transfer torque (STT)-MRAM, spin-electron magnetic junction memory based devices, magnetic tunnel junction (MTJ) based devices, DW (domain wall) and SOT (spin orbit transfer) based devices, thyristor based storage devices, combinations of any of the above, or other suitable memory. The memory device can also refer to three-dimensional cross-point memory devices (e.g., 3D XPoint TM Memory) or other byte-addressable in-place write non-volatile memory devices. The memory device can refer to the die itself and / or to a packaged memory product. In some examples, the 3D cross-point memory (e.g., 3D XPoint TM The memory 1106 can include a transistor-less stackable cross-point architecture where memory cells are located at the intersection of word lines and bit lines and are individually addressable, and where bit storage is based on changes in bulk resistance. In some examples, all or portions of the memory 1106 can be integrated into the processor 1104. The memory 1106 can store various software and data used during operation of the device 1100 such as one or more applications, databases, and drivers.

[0179] In some examples, resistance- and / or transistor-less memory architectures include nanoscale phase change memory (PCM) devices in which a volume of phase change material is positioned between at least two electrodes. Portions of example phase change materials exhibit varying degrees of crystalline and amorphous phases, where varying degrees of electrical resistance between the at least two electrodes can be measured. In some examples, the phase change material is a chalcogenide-based glass material. Such resistance-based memory devices are sometimes referred to as memristor devices, which can remember a history of current previously flowing through them. Stored data is retrieved from example PCM devices by measuring resistance, where crystalline phases exhibit relatively lower resistance value(s) (e.g., a logical "0") when compared to amorphous phases having relatively higher resistance value(s) (e.g., a logical "1").

[0180] Example PCM devices can store data for long periods of time (e.g., about 10 years at room temperature). Write operations to example PCM devices (e.g., set to logical "0", set to logical "1", set to an intermediate resistance value) are achieved by applying one or more current pulses to the at least two electrodes, where the pulses have a particular current amplitude and duration. For example, applying a longer, low current pulse to the at least two electrodes (set) causes example PCM devices to be in a low resistance crystalline state, while applying a relatively shorter, high current pulse to the at least two electrodes (reset) causes example PCM devices to be in a high resistance amorphous state.

[0181] In some examples, implementation of PCM devices facilitates a non-von Neumann computing architecture that enables in-memory computing capabilities. Generally speaking, traditional computing architectures include a central processing unit (CPU) communicatively connected to one or more memory devices over a bus. As a result, transferring data between the CPU and the memory consumes a finite amount of energy and time, which is a known bottleneck of von Neumann computing architectures. However, PCM devices minimize, and in some cases eliminate, data transfer between the CPU and the memory by performing some computing operations in-memory. In other words, PCM devices both store information and perform computing tasks. Such non-von Neumann computing architectures can implement vectors having a relatively high dimensionality to facilitate high-dimensional computing, such as vectors having 10,000 bits. The relatively large bit-width vectors enable implementation of computing paradigms modeled after the human brain, which also processes information similar to wide bit vectors.

[0182] The compute circuitry 1102 is communicatively coupled with other components of the compute node 1100 through an I / O subsystem 1108, which can be embodied in circuitry and / or components to facilitate input / output operations with the compute circuitry 1102 (e.g., with the processor(s) 1104 and / or the main memory 1106) and other components of the compute circuitry 1102. For example, the I / O subsystem 1108 can be embodied as, or otherwise include, a memory controller hub, an input / output control hub, an integrated sensor hub, firmware devices, communication links (e.g., point-to-point links, bus links, wires, cables, light guides, printed circuit board traces, etc.) and / or other components and subsystems to facilitate the input / output operations. In some examples, the I / O subsystem 1108 can form a portion of a system on a chip (SoC) and be incorporated alongside the processor(s) 1104, the memory 1106, and other components of the compute circuitry 1102 in the compute circuitry 1102.

[0183] The one or more illustrative data storage devices / disks 1110 can be embodied as one or more any type(s) of physical device(s) configured for short- or long-term storage of data, such as memory devices, memory, circuitry, memory cards, flash drives, hard disk drives (HDDs), solid-state drives (SSDs), and / or other data storage devices / disks. The individual data storage devices / disks 1110 can include a system partition that stores data and firmware code for the data storage devices / disks 1110. The individual data storage devices / disks 1110 can also include one or more operating system partitions that store data files and executable files for an operating system, depending on the type of compute node 1100, for example.

[0184] The communication circuitry 1112 can be embodied as any communication circuitry, device, or collection thereof, which enables communication over a network between the compute circuitry 1102 and another computing device, such as an edge gateway implementing an edge computing system. The communication circuitry 1112 can be configured to enable such communication using any one or more communication technology (e.g., wired or wireless communications) and associated protocols (e.g., cellular network protocols such as 3GPP 4G or 5G standards, wireless local area network protocols such as IEEE 802.11 / a / b / g / n / ac / ad / ay / c, Ethernet, Bluetooth Low Energy, IoT protocols such as IEEE 802.15.4 or Zigbee®, Low-Power Wide Area Network (LPWAN) or Low-Power Wide (LPWA) protocols, etc.).

[0185] ​​The illustrative communication circuit 1112 includes a network interface controller (NIC) 1120, which can also be referred to as a host fabric interface (HFI). The NIC 1120 can be embodied as one or more add-in boards, daughter cards, network interface cards, controller chips, chipsets, or other devices used by the computing node 1100 to connect with another computing device, such as an edge gateway node. In some examples, the NIC 1120 can be embodied as part of a system on a chip (SoC) that includes one or more processors, or included in a multi-chip package that also contains one or more processors. In some examples, the NIC 1120 can include a local processor (not shown) and / or a local memory (not shown), which are both local to the NIC 1120. In such examples, the local processor of the NIC 1120 can be capable of performing one or more functions of the compute circuit 1102 described herein. Additionally or alternatively, in such examples, the local memory of the NIC 1120 can be integrated at a board level, a slot level, a chip level, and / or other level in one or more components of the client computing node.

[0186] Additionally, in some examples, the respective computing node 1100 can include one or more peripheral devices 1114. Such peripheral devices 1114 can include any type of peripheral device found in a computing device or server, such as audio input devices, displays, other input / output devices, interface devices, and / or other peripheral devices, depending on the particular type of computing node 1100. In further examples, the computing node 1100 can be embodied by a respective edge computing node in an edge computing system, whether a client, gateway, or aggregation node, or similar form of appliance, computer, subsystem, circuit, or other component.

[0187] In more detail examples, Figure 11B FIG. illustrates a block diagram of an example of components that can be present in an edge computing node 1150 for implementing the techniques (e.g., operations, processes, methods, and protocols) described herein. The edge computing node 1150 provides a more detailed view of respective components of the node 1100 when the node 1100 is implemented as or as part of a computing device, such as a mobile device, base station, server, gateway, etc. The edge computing node 1150 can include any combination of the hardware or logic components referenced herein, and it can include or be coupled to any device that can be used with an edge communication network or combination of such networks. The components can be implemented as integrated circuits (ICs), portions thereof, discrete electronic devices, or other modules, an instruction set of one or more processors, programmable logic, or algorithm(s), hardware, hardware / firmware / accelerator, software, firmware, or a combination thereof adapted in the edge computing node 1150, or otherwise incorporated within a chassis of a larger system as a component.

[0188] Edge computing device 1150 may include processing circuitry in the form of processor 1152, which may be a microprocessor, multi-core processor, multi-threaded processor, ultra-low voltage processor, embedded processor, xPU / DPU / IPU / NPU, special purpose processing unit, dedicated processing unit, or other known processing element. Processor 1152 may be part of a system-on-a-chip (SoC), in which processor 1152 and other components are formed as a single integrated circuit or a single package, such as the Edison from Intel Corporation. TM Or Galileo TM SoC board. As an example, processor 1152 may include a SoC-based... Architecture Core TM CPU processors, such as Quark TM Atom TM i3, i5, i7, i9 or MCU-level processors, or available from Other processors of this type are available. However, any number of other processors can be used, such as those from Advanced Micro Devices, Inc., Sunnyvale, California. The data was obtained from MIPS Technologies in Sunnyvale, California. Designed and licensed by ARM Holdings, Ltd. or its customers, or its licensors or adopters based on The processor can include units such as those from: Inc.'s A5-A13, from Snapdragon Technologies TM The processor may be an OMAP from Texas Instruments. TM Processor. The processor 1152 and its accompanying circuitry may be provided in a single-socket configuration, a multi-socket configuration, or various other formats, including in constrained hardware configurations or comprising fewer than Figure 11B The configuration of all components shown is provided.

[0189] The processor 1152 can communicate with the system memory 1154 over the interconnect 1156 (e.g., a bus). Any number of memory devices can be used to provide a given amount of system memory. As an example, the memory 1154 can be a random access memory (RAM) that conforms to a Joint Electron Devices

[0190] To provide persistent storage of information such as data, applications, operating systems, etc., a memory 1158 can also be coupled to the processor 1152 via the interconnect 1156. In one example, the storage device 1158 can be implemented via a solid state disk drive (SSDD). Other devices that can be used for the storage device 1158 include flash memory cards, such as Secure Digital (SD) cards, microSD cards, Extreme Digital (XD) Picture cards, and the like, as well as Universal Serial Bus (USB) flash drives. In examples, the memory device can be or can include a memory device using a chalcogenide glass, multi-level threshold NAND flash memory, NOR flash memory, single-level or multi-level phase change memory (PCM), resistive memory, nanowire memory, ferroelectric transistor random access memory (FeTRAM), anti-ferroelectric memory, magnetoresistive random access memory (MRAM) containing memristor technology, resistive memory including metal-oxide based, oxygen-vacancy based, and conductive-bridge random access memory (CB-RAM), or spin-transfer torque (STT)-MRAM, spintronic magnetic junction memory-based devices, magnetic tunnel junction (MTJ)-based devices, domain wall (DW) and SOT (spin orbit transfer)-based devices, thyristor-based memory devices, or any combination of the foregoing or other memory.

[0191] In low power implementations, the storage device 1158 can be an on-chip memory or register associated with the processor 1152. However, in some examples, the storage device 1158 can be implemented using a micro hard disk drive (HDD). Further, in addition to or in lieu of the technologies described, any number of new technologies can be used for the storage device 1158, such as resistive memory, phase change memory, holographic memory, or chemical memory, among others.

[0192] Components can communicate via the interconnect 1156. The interconnect 1156 can include any number of technologies, including industry standard architecture (ISA), extended ISA (EISA), peripheral component interconnect (PCI), peripheral component interconnect extended (PCIx), PCI Express (PCIe), or any number of other technologies. The interconnect 1156 can be a proprietary bus, such as used in a system on a chip (SoC) based system. Other bus systems can be included, such as an inter-integrated circuit (I2C) interface, serial peripheral interface (SPI) interface, point-to-point interfaces, power bus, and the like.

[0193] The interconnect 1156 can couple the processor 1152 to a transceiver 1166 for communication with the connected edge device 1162. The transceiver 1166 can use any number of frequencies and protocols, such as 2.4 gigahertz (GHz) transmissions under the IEEE 802.15.4 standard, using Bluetooth® or Bluetooth® Low Energy (BLE) protocols, ZigBee®, Z-Wave®, or any number of other protocols. Bluetooth® Low Energy (BLE) standards, such as defined by the Bluetooth® Special Interest Group, or Wi-Fi® standards, or other standards. Any number of radios configured for specific wireless communication protocols can be used for connectivity with connected edge devices 1162. For example, a wireless local area network (WLAN) unit can be used to implement Wi-Fi® communications according to the Institute of Electrical and Electronics Engineers (IEEE) 802.11 standard. Communications. Further, wireless wide area communications, for example according to cellular or other wireless wide area protocols, can be via a wireless wide area network (WWAN) unit.

[0194] The wireless network transceiver 1166 (or multiple transceivers) can use different standards or radios for communications at different ranges. For example, the edge computing node 1150 can use a local transceiver based on Bluetooth® Low Energy (BLE) or other low power radio to communicate with proximal devices (e.g., within about 10 meters) to conserve power. More distant connected edge devices 1162 (e.g., within about 50 meters) can reach by or other medium power radios. The two communication technologies can be at different power levels by a single radio, or can be by separate transceivers, such as a local transceiver using BLE and a separate mesh transceiver using .

[0195] The wireless network transceiver 1166 (e.g., radio transceiver) can be included to communicate with devices or services in the cloud (e.g., edge cloud 1195) via local or wide area network protocols. The wireless network transceiver 1166 can be a low power wide area (LPWA) transceiver following the IEEE 802.15.4 or IEEE 802.15.4g standards, among others. The edge computing node 1150 can use LoRaWAN® (Long Range Wide Area Network) developed by Semtech and the LoRa Alliance to communicate over wide areas. The technology described herein is not limited to these technologies, but can be used with any number of other cloud transceivers that enable long range, low bandwidth communications, such as Sigfox and other technologies. Further, other communication technologies can be used, such as the time-slotted channel hopping described in the IEEE 802.15.4e specification. TM

[0196] As described herein, any number of other radio communications and protocols can be used in addition to the systems for the wireless network transceiver 1166. For example, the transceiver 1166 can include a cellular transceiver that uses spread spectrum (SPA / SAS) communications for high speed communications. Further, any number of other protocols can be used, such as the ZigBee® for medium speed communications and providing network communications. ​Network. Transceiver 1166 can include a radio compatible with any number of 3GPP (Third Generation Partnership Project) specifications, such as Long-Term Evolution (LTE) and Fifth Generation (5G) communication systems, which are discussed in further detail at the end of this disclosure. A network interface controller (NIC) 1168 can be included to provide wired communication to nodes of the edge cloud 1195 or other devices, such as the connected endpoint devices 1162 (e.g., operating in a mesh fashion). The wired communication can provide an Ethernet connection, or can be based on other types of networks, such as Controller Area Network (CAN), Local Interconnect Network (LIN), DeviceNet, ControlNet, Data Highway+, PROFIBUS, or PROFINET, among others. Additional NICs 1168 can be included to enable connectivity with a second network, e.g., a first NIC 1168 provides communication over Ethernet to the cloud, and a second NIC 1168 provides communication with other devices over another type of network.

[0197] In view of the variety of applicable communication types from a device to another component or network, the applicable communication circuitry used by a device can be included or embodied as any one or more of the components 1164, 1166, 1168, or 1170. Thus, in various examples, the applicable means for communication (e.g., reception, transmission, etc.) can be embodied by such communication circuitry.

[0198] The edge computing node 1150 can include or be coupled to acceleration circuitry 1164, which can be embodied as one or more artificial intelligence (AI) accelerators, neural compute sticks, neuromorphic hardware, FPGAs, GPU arrays, xPU / DPU / IPU / NPU arrays, one or more SoCs, one or more CPUs, one or more digital signal processors, specialized ASICs, or other forms of specialized processors or circuitry designed to complete one or more specific tasks. These tasks can include AI processing (including machine learning, training, inferencing, and classification operations), visual data processing, network data processing, object detection, rule analysis, and the like. These tasks can also include the specific edge computing tasks for service management and service operations discussed elsewhere herein.

[0199] The interconnect 1156 can couple the processor 1152 to a sensor hub or external interface 1170 for connecting to additional devices or subsystems. The devices can include sensors 1172, such as accelerometers, liquid level sensors, flow sensors, optical light sensors, camera sensors, temperature sensors, global navigation system (e.g., GPS) sensors, pressure sensors, barometric pressure sensors, and the like. The hub or interface 1170 can also be used to connect the edge computing node 1150 to actuators 1174, such as power switches, valve actuators, sound generators, visual warning devices, and the like.

[0200] In some optional examples, various input / output (I / O) devices can be present within or connected to edge computing node 1150. For example, a display or other output device 1184 can be included to display information, such as sensor readings or actuator positions. An input device 1186, such as a touchscreen or keyboard, can be included to accept input. Output device 1184 can include any number of audio or visual display forms, including simple visual outputs such as binary status indicators (e.g., light emitting diodes (LEDs)) and multi-character visual outputs, or more complex outputs such as display screens (e.g., liquid crystal display (LCD) screens), where output of characters, graphics, multimedia objects, and the like are generated or produced from the operation of edge computing node 1150. In the present system context, a display or console hardware can be used to provide output and receive input for the edge computing system; to manage components or services of the edge computing system; to identify a status of edge computing components or services; or to perform any other number of management or administrative functions or service use cases.

[0201] A battery 1176 can power edge computing node 1150, however, in examples where edge computing node 1150 is installed in a fixed location, it can have a power supply connected to a grid, or the battery can be used as backup or for temporary capability. Battery 1176 can be a lithium ion battery or a metal-air battery, such as a zinc-air battery, an aluminum-air battery, a lithium-air battery, and the like.

[0202] A battery monitor / charger 1178 can be included in edge computing node 1150 to track the state of charge (SoCh) of battery 1176, if included. Battery monitor / charger 1178 can be used to monitor other parameters of battery 1176 to provide fault predictions, such as the state of health (SoH) and state of function (SoF) of battery 1176. Battery monitor / charger 1178 can include a battery monitor integrated circuit, such as the LTC4020 or LTC2990 from Linear Technologies, the ADT7488A from ON Semiconductor of Phoenix, Arizona, or the UCD90xxx series IC from Texas Instruments of Dallas, Texas. Battery monitor / charger 1178 can communicate information about battery 1176 to processor 1152 over interconnects 1156. Battery monitor / charger 1178 can also include an analog-to-digital (ADC) converter that enables processor 1152 to directly monitor the voltage of battery 1176 or the current from battery 1176. The battery parameters can be used to determine actions that edge computing node 1150 can perform, such as transmission frequency, mesh network operation, sensing frequency, and the like.

[0203] The power block 1180 or other power supply connected to the power grid can be coupled with the battery monitor / charger 1178 to charge the battery 1176. In some examples, the power block 1180 can be replaced with a wireless power receiver to receive power wirelessly, such as through a loop antenna in the edge computing node 1150. A wireless battery charging circuit, such as the LTC4020 chip from Linear Technologies of Milpitas, California, or the like, can be included in the battery monitor / charger 1178. The particular charging circuit can be selected according to the size of the battery 1176, and thus the current required. Charging can be performed using the Airfuel standard published by the Airfuel Alliance, the Qi wireless standard published by the Wireless Power Consortium, or the Rezence charging standard published by the Alliance for Wireless Power, or the like.

[0204] The storage device 1158 can include instructions 1182 in the form of software, firmware, or hardware commands to implement the techniques described herein. While such instructions 1182 are shown as code blocks included in the memory 1154 and the storage device 1158, it can be appreciated that any of the code blocks can be replaced with hardwired circuitry, such as built-in to an application specific integrated circuit (ASIC).

[0205] In examples, the instructions 1182 provided via the memory 1154, the storage device 1158, or the processor 1152 can be embodied as a non-transitory, machine-readable medium 1160 including code to direct the processor 1152 to perform electronic operations in the edge computing node 1150. The processor 1152 can access the non-transitory, machine-readable medium 1160 through the interconnect 1156. For example, the non-transitory, machine-readable medium 1160 can be embodied by a device described for the storage device 1158, or can include a particular storage unit, such as a storage device and / or storage disk including optical (e.g., digital versatile disk (DVD), compact disk (CD), CD-ROM, Blu-ray disk), flash drive, floppy disk, hard disk drive (e.g., SSD), or any number of other hardware devices, where information is stored for any duration (e.g., for extended time periods, permanently, for brief instances, temporarily hold, and / or cache). The non-transitory, machine-readable medium 1160 can include instructions to direct the processor 1152 to perform an action in a particular sequence or flow, such as described with respect to the flow diagram(s) and block diagram(s) of operations and functions depicted above. As used herein, the terms “machine-readable medium” and “computer-readable medium” are interchangeable. As used herein, the term “non-transitory computer-readable medium” is expressly defined to include any type of computer-readable storage device and / or storage disk, and excludes propagating signals and transitory media.

[0206] Furthermore, in a specific example, instructions 1182 on processor 1152 (alone or in combination with instructions 1182 on machine-readable medium 1160) can configure the execution or operation of trusted execution environment (TEE) 1190. In the example, TEE 1190 serves as a protected region accessible to processor 1152 for securely executing instructions and securely accessing data. Various implementations of TEE 1190, as well as accompanying secure regions in processor 1152 or memory 1154, can be provided, for example, by using... Software Protection Extensions (SGX) or Hardware security extensions Management Engine (ME) or Converged Security Management Engine (CSME). Security hardening, hardware root of trust, and other aspects of trusted or protected operation can be implemented in device 1150 via TEE 1190 and processor 1152.

[0207] although Figure 11A and Figure 11B The illustrations include example components for compute nodes and compute devices, respectively, but the examples disclosed herein are not limited to these. As used herein, "computer" can include various types of compute environments. Figure 11A and / or Figure 11B Some or all of the example components. Example computing environments include edge computing devices (e.g., edge computers) in a distributed network configuration, such that certain devices among the participating edge computing devices are heterogeneous or homogeneous devices. As used herein, "computer" can include personal computers, servers, user devices, accelerators, etc., including any combination thereof. In some examples, distributed networks and / or distributed computing include, for example, edge computing devices such as edge computers, ... Figure 11A and / or Figure 11B The illustrations depict any number of such edge computing devices, each of which may include different sub-components, different memory capabilities, I / O capabilities, etc. For example, the examples disclosed herein include those related to specific desired functionalities, as the implementation of certain distributed networks and / or distributed computing is associated with particular desired functionalities. Figure 11A and / or Figure 11B The illustrations depict different combinations of components to meet the functional objectives of distributed computing tasks. In some examples, the terms "compute node" or "computer" only include... Figure 11Aexample processor 1104, memory 1106, and I / O subsystem 1108. In some examples, the objective function of one or more of the distributed computing task(s) depends on one or more alternative devices / structures located in different parts of the edge network environment, such as devices for accommodating data storage (e.g., example data storage device 1110), input / output capabilities (e.g., example peripheral device(s) 1114), and / or network communication capabilities (e.g., example NIC 1120).

[0208] In some examples, computers running in a distributed computing and / or distributed network environment (e.g., an edge network) are structured in a manner that reduces computational waste to accommodate a particular objective function. For example, because the computers include a subset of the components disclosed in Figure 11A and Figure 11B , such computers satisfy execution of the objective function of the distributed computing task without including computing structures that would otherwise go unused and / or underutilized. Accordingly, the term “computer” as used herein includes any structural combination of Figure 11A and / or Figure 11B that is capable of satisfying and / or in addition to otherwise executing the objective function of a distributed computing task. In some examples, computers are structured in a manner commensurate with the corresponding distributed computing objective function so as to scale down or up in relation to dynamic demand. In some examples, different computers are invoked and / or in addition to otherwise instantiated in view of the ability of the different computers to process one or more tasks of the distributed computing request(s), such that any computer capable of satisfying a task continues with such computing activity.

[0209] In Figure 11A and Figure 11B illustrated examples, the computing device includes an operating system. As used herein, an “operating system” is software for controlling an example computing device, such as Figure 11A example edge computing node 1100 and / or Figure 11B example edge computing node 1150. Example operating systems include, but are not limited to, consumer-based operating systems (e.g., 10, OS, Example operating systems include, but are not limited to, Microsoft® Windows® (e.g., Windows® 10, Windows® 8, Windows® 7, Windows® XP, etc.), various versions of the Unix® or Linux® operating systems (e.g., the GNU system, Android®, iOS®, macOS®, etc.), and / or the like. Example operating systems also include, but are not limited to, industry-focused operating systems such as real-time operating systems, virtual machine monitors, and / or the like. Example operating systems on the first edge computing node can be the same as or different from example operating systems on the second edge computing node. In some examples, the operating system invokes replacement software to facilitate one or more functions and / or operations that are not native to the operating system, such as specific communication protocols and / or interpreters. In some examples, the operating system instantiates various functions that are not native to the operating system. In some examples, the operating systems include varying degrees of complexity and / or capability. For example, a first operating system corresponding to the first edge computing node includes a real-time operating system having specific performance expectations in response to dynamic input conditions, while a second operating system corresponding to the second edge computing node includes graphical user interface capabilities to facilitate end-user I / O.

[0210] Examples

[0211] The following provides illustrative examples of the technology described throughout this document. Embodiments of the technology can include any one or more, and any combination of, the examples described below. In some embodiments, at least one of the systems or components set forth in one or more of the preceding figures can be configured to perform one or more of the operations, techniques, processes, and / or methods set forth in the examples below.

[0212] Example 1 includes a system comprising: interface circuitry; and processing circuitry to: receive, via the interface circuitry, performance data indicative of performance of a plurality of wireless networks, wherein the wireless networks are based on a plurality of wireless technologies, and wherein the performance data is based on a plurality of protocol layers of the wireless technologies; determine, based on the performance data, one or more configuration settings to adjust for one or more of the wireless networks; and adjust the one or more configuration settings for the one or more of the wireless networks.

[0213] Example 2 includes the system of Example 1, wherein the performance data comprises: one or more quality of service (QoS) metrics for one or more of the wireless networks; and radio frequency (RF) spectrum data, wherein the RF spectrum data is based on an analysis of RF signals detected by one or more RF receivers at a plurality of frequencies, wherein the one or more RF receivers are independent of the plurality of wireless networks.

[0214] Example 3 includes the system of Example 2, wherein the one or more QoS metrics are indicative of one or more of: a demodulation error; a data transmission error; a signal quality; a network layer performance; or a transport layer performance.

[0215] Example 4 includes the system of any one of Examples 2-3, wherein the performance data further comprises spatiotemporal performance data, wherein the spatiotemporal performance data is indicative of performance of one or more of the wireless networks based on time and location.

[0216] Example 5 includes a system of any one of Examples 2 to 4, wherein the processing circuitry that determines one or more configuration settings to be adjusted for one or more wireless networks based on performance data is further configured to: detect one or more signal interference sources for one or more wireless networks based on RF spectrum data.

[0217] Example 6 includes a system of any one of Examples 1 to 5, wherein the processing circuitry for determining one or more configuration settings to be adjusted for one or more wireless networks based on performance data is further configured to: determine one or more configuration settings to be adjusted for one or more wireless networks based on performance data and one or more traffic flow targets, wherein the one or more traffic flow targets indicate the priority among multiple traffic flows on the wireless network.

[0218] Example 7 includes a system of any one of Examples 1 to 6, wherein one or more configuration settings include one or more of the following: modulation settings; transmission power settings; operating channel settings; radio frequency resource management settings; or quality of service settings.

[0219] Example 8 includes a system of any one of Examples 1 to 7, wherein the protocol layer includes a physical layer, a network layer, and a transport layer.

[0220] Example 9 includes a system of any one of Examples 1 to 8, wherein: the wireless technology is based on a corresponding protocol stack; and the performance data is based on protocol layers from multiple levels of the protocol stack.

[0221] Example 10 includes a system of any of Examples 1 to 9, wherein one or more configuration settings are determined based on a machine learning (ML) model, wherein the ML model is trained to infer configuration adjustments based on historical performance data for wireless technologies.

[0222] Example 11 includes a system of any one of Examples 1 to 10, wherein three or more wireless networks are based on three or more of the following wireless technologies: Bluetooth; Wi-Fi; cellular technology; or IEEE 802.15.4 low-rate wireless personal area network technology.

[0223] Example 12 includes at least one non-transitory machine-readable storage medium storing instructions thereon, wherein when the instructions are executed on processing circuitry, the processing circuitry causes the processing circuitry to: receive performance data indicating the performance of multiple wireless networks deployed in an environment, wherein the wireless networks are based on multiple wireless technologies, wherein the wireless technologies are based on corresponding protocol stacks, and wherein the performance data is based on multiple protocol layers from different levels of the protocol stacks; determine one or more configuration settings to be adjusted for one or more wireless networks based on the performance data; and adjust one or more configuration settings for one or more wireless networks.

[0224] Example 13 includes the storage medium of Example 12, wherein the performance data comprises one or more quality of service (QoS) metrics for the one or more wireless networks, wherein the one or more QoS metrics indicate one or more of: a demodulation error; a data transmission error; a signal quality; a network layer performance; or a transport layer performance.

[0225] Example 14 includes the storage medium of Example 13, wherein the performance data further comprises spatiotemporal performance data, and wherein the spatiotemporal performance data indicates a time and location based performance of the one or more wireless networks.

[0226] Example 15 includes the storage medium of any of Examples 13 to 14, wherein the performance data further comprises radio frequency (RF) spectrum data, wherein the RF spectrum data is based on an analysis of RF signals detected by one or more RF receivers at a plurality of frequencies, wherein the one or more RF receivers are independent of the plurality of wireless networks.

[0227] Example 16 includes the storage medium of Example 15, wherein the instructions that cause the processing circuit to determine, based on the performance data, one or more configuration settings to adjust for the one or more wireless networks, further cause the processing circuit to: detect, based on the RF spectrum data, one or more sources of signal interference for the one or more wireless networks.

[0228] Example 17 includes the storage medium of any of Examples 12 to 16, wherein the instructions that cause the processing circuit to determine, based on the performance data, one or more configuration settings to adjust for the one or more wireless networks, further cause the processing circuit to: determine, based on the performance data and one or more traffic flow objectives, the one or more configuration settings to adjust for the one or more wireless networks, wherein the one or more traffic flow objectives indicate a priority between a plurality of traffic flows on the wireless networks.

[0229] Example 18 includes the storage medium of any of Examples 12 to 17, wherein the one or more configuration settings comprise one or more of: a modulation setting; a transmission power setting; a control channel setting; a radio resource management setting; or a quality of service setting.

[0230] Example 19 includes the storage medium of any of Examples 12 to 18, wherein the one or more configuration settings are determined based on a machine learning (ML) model, wherein the ML is trained to infer configuration adjustments based on historical performance data for wireless technologies.

[0231] Example 20 includes the storage medium of any of Examples 12 to 19, wherein the protocol layer comprises a physical layer, a network layer, and a transport layer.

[0232] Example 21 includes the storage medium of any of Examples 12 to 20, wherein the three or more wireless networks are based on three or more of: Bluetooth; Wi-Fi; cellular technology; or IEEE 802.15.4 low-rate wireless personal area network technology.

[0233] Example 22 includes a method comprising: receiving performance data indicative of performance of a plurality of wireless networks deployed in an environment, wherein the wireless networks are based on a plurality of wireless technologies, wherein a wireless technology is based on a corresponding protocol stack, and wherein the performance data is based on a plurality of protocol layers from different levels of the protocol stack; determining, based on the performance data, one or more configuration settings to adjust for one or more of the wireless networks; and adjusting the one or more configuration settings for the one or more of the wireless networks.

[0234] Example 23 includes the method of Example 22, wherein the performance data comprises: one or more quality of service metrics for the one or more of the wireless networks; and radio frequency (RF) spectrum data, wherein the RF spectrum data is based on an analysis of RF signals detected at a plurality of frequencies by one or more RF receivers, wherein the one or more RF receivers are independent of the plurality of wireless networks.

[0235] Example 24 includes the method of any of Examples 22 to 23, wherein the one or more configuration settings are determined based on a machine learning (ML) model, wherein the ML model is trained to infer configuration adjustments based on historical performance data for the wireless technologies.

[0236] Example 25 includes the method of any of Examples 22 to 24, wherein the protocol layers comprise a physical layer, a network layer, and a transport layer.

[0237] Example 26 includes the system of Example 5, wherein the one or more signal interferers comprise transmissions from a plurality of devices, wherein each device has a physically independent housing.

[0238] While the concept of the present disclosure is susceptible to various modifications and alternative forms, specific embodiments thereof have been shown by way of example in the drawings and will herein be described in detail. It should be understood, however, that there is no intent to limit the concept of the present disclosure to the particular forms disclosed, but on the contrary, the intention is to cover all modifications, equivalents, and alternatives falling within the spirit and scope of the present disclosure and the appended claims.

[0239] References in the specification to "one embodiment," "an embodiment,” "an illustrative embodiment,” etc., indicate that the embodiment described can include a particular feature, structure, or characteristic, but every embodiment can not necessarily include the particular feature, structure, or characteristic. Moreover, such phrases are not necessarily referring to the same embodiment. Further, when a particular feature, structure, or characteristic is described in connection with an embodiment, it is submitted that it is within the knowledge of those skilled in the art to effect such feature, structure, or characteristic in connection with other embodiments whether or not explicitly described. Additionally, it should be understood that the inclusion of a listing of items in a "list of at least one of A, B, and C” does not mean that only A, B, or C can be included in the list, but rather, that A, B, or C can be included in the list, or any combination of A, B, and C can be included in the list. Similarly, the inclusion of a listing of items in a "list of at least one of A, B, or C” does not mean that only A, B, or C can be included in the list, but rather, that A, B, or C can be included in the list, or any combination of A, B, and C can be included in the list.

[0240] The disclosed embodiments can in some cases be implemented as software, firmware, hardware or any combination thereof. The disclosed embodiments can also be implemented as instructions carried by or stored on one or more transitory or non-transitory machine- readable (e.g., computer-readable) storage media, which can be read and executed by one or more processors. A machine- readable storage medium can be any storage medium for storing or transmitting information in a form readable by a machine (e.g., a volatile or non-volatile memory, a media disc or other media device).

[0241] In the drawings, some structural or methodological features can be shown in particular arrangements and / or orders. It should be appreciated, however, that such particular arrangements and / or orders are not essential. Rather, in some embodiments, such features can be arranged in a different manner and / or order than shown in the illustrative drawings. Moreover, inclusion of a structural or methodological feature in a particular drawing does not mean that the feature is essential in all embodiments, and in some embodiments, the feature can not be included or can be combined with other features.

Claims

1. A system comprising: Interface circuit; as well as Processing circuitry, used for: The interface circuit receives performance data indicating the performance of multiple wireless networks, wherein the wireless networks are based on multiple wireless technologies, each having its own spectrum, and wherein the performance data is based on multiple protocol layers of the wireless technologies. Based on the performance data, determine one or more configuration settings to be adjusted for one or more of the wireless networks; and Adjust the configuration settings of one or more of the wireless networks.

2. The system according to claim 1, characterized in that, The performance data includes: One or more Quality of Service (QoS) metrics for one or more of the aforementioned wireless networks; and Radio frequency (RF) spectrum data, wherein the RF spectrum data is based on the analysis of radio frequency signals detected by one or more radio frequency receivers at multiple frequencies, wherein the one or more radio frequency receivers are independent of the multiple wireless networks.

3. The system according to claim 2, characterized in that, The one or more QoS metrics indicate one or more of the following: Demodulation error; Data transmission error; Signal quality; Network layer performance; or Transport layer performance.

4. The system according to any one of claims 2 to 3, characterized in that, The performance data further includes spatiotemporal performance data, wherein the spatiotemporal performance data indicates the time- and location-based performance of one or more of the wireless networks.

5. The system according to any one of claims 2 to 4, characterized in that, The processing circuitry that determines one or more configuration settings to be adjusted for one or more of the wireless networks based on the performance data is further configured to: Based on the radio frequency spectrum data, one or more signal interference sources are detected targeting one or more of the wireless networks.

6. The system according to claim 5, characterized in that, The one or more sources of signal interference include transmissions from multiple devices, each of which has a physically independent housing.

7. The system according to any one of claims 1 to 6, characterized in that, The processing circuitry that determines one or more configuration settings to be adjusted for one or more of the wireless networks based on the performance data is further configured to: Based on the performance data and one or more traffic flow targets, determine one or more configuration settings to be adjusted for one or more of the wireless networks, wherein the one or more traffic flow targets indicate the priority among multiple traffic flows on the wireless network.

8. The system according to any one of claims 1 to 7, characterized in that, The one or more configuration settings include one or more of the following: Modulation settings; Transmission power settings; Working channel settings; Wireless resource management settings; or Service quality settings.

9. The system according to any one of claims 1 to 8, characterized in that, The protocol layer includes the physical layer, network layer, and transport layer.

10. The system according to any one of claims 1 to 9, wherein: The wireless technology is based on a corresponding protocol stack; and The performance data is based on protocol layers from multiple levels of the protocol stack.

11. The system according to any one of claims 1 to 10, characterized in that, The one or more configuration settings are determined based on a machine learning (ML) model, wherein the ML model is trained to infer configuration adjustments based on historical performance data for the wireless technology.

12. The system according to any one of claims 1 to 11, characterized in that, The three or more wireless networks are based on three or more of the following wireless technologies: Bluetooth; Wi-Fi; Cellular technology; or IEEE 802.15.4 low-speed wireless personal area network technology.

13. At least one non-transitory machine-readable storage medium storing instructions thereon, wherein when the instructions are executed on processing circuitry, the processing circuitry is caused to: Receive performance data indicating the performance of multiple wireless networks deployed in the environment, among which, The wireless network is based on a variety of wireless technologies, wherein the wireless technologies are based on corresponding protocol stacks, and wherein the performance data is based on multiple protocol layers from different levels of the protocol stack. Based on the performance data, determine one or more configuration settings to be adjusted for one or more of the wireless networks; as well as Adjust the configuration settings of one or more of the wireless networks.

14. The storage medium according to claim 13, wherein, The performance data includes one or more Quality of Service (QoS) metrics for one or more of the wireless networks, wherein the one or more QoS metrics indicate one or more of the following: Demodulation error; Data transmission error; Signal quality; Network layer performance; or Transport layer performance.

15. The storage medium according to claim 14, characterized in that, The performance data also includes spatiotemporal performance data, wherein the spatiotemporal performance data indicates the time- and location-based performance of one or more of the wireless networks.

16. The storage medium according to any one of claims 14 to 15, characterized in that, The performance data also includes radio frequency (RF) spectrum data, wherein the RF spectrum data is based on the analysis of radio frequency signals detected by one or more RF receivers at multiple frequencies, wherein the one or more RF receivers are independent of the multiple wireless networks.

17. The storage medium according to claim 16, characterized in that, The processing circuitry is prompted to determine, based on the performance data, one or more configuration settings to be adjusted for one or more of the wireless networks, further instructing the processing circuitry to: Based on the RF spectrum data, one or more signal interference sources are detected targeting one or more of the wireless networks.

18. The storage medium according to any one of claims 13 to 17, characterized in that, The processing circuitry is prompted to determine, based on the performance data, one or more configuration settings to be adjusted for one or more of the wireless networks, further instructing the processing circuitry to: Based on the performance data and one or more traffic flow targets, determine one or more configuration settings to be adjusted for one or more of the wireless networks, wherein the one or more traffic flow targets indicate the priority among multiple traffic flows on the wireless network.

19. The storage medium according to any one of claims 13 to 18, characterized in that, The one or more configuration settings include one or more of the following: Modulation settings; Transmission power settings; Working channel settings; Wireless resource management settings; or Service quality settings.

20. The storage medium according to any one of claims 13 to 19, characterized in that, The one or more configuration settings are determined based on a machine learning (ML) model, wherein the ML model is trained to infer configuration adjustments based on historical performance data for the wireless technology.

21. The storage medium according to any one of claims 13 to 20, characterized in that, The protocol layer includes the physical layer, network layer, and transport layer.

22. A method comprising: Receive performance data indicating the performance of multiple wireless networks deployed in an environment, wherein the wireless networks are based on multiple wireless technologies, wherein the wireless technologies are based on corresponding protocol stacks, and wherein the performance data is based on multiple protocol layers from different levels of the protocol stacks. Based on the performance data, determine one or more configuration settings to be adjusted for one or more of the wireless networks; as well as Adjust the configuration settings of one or more of the wireless networks.

23. The method according to claim 22, characterized in that, The performance data includes: One or more quality of service metrics for one or more of the aforementioned wireless networks; and Radio frequency (RF) spectrum data, wherein the RF spectrum data is based on the analysis of RF signals detected by one or more RF receivers at multiple frequencies, wherein the one or more RF receivers are independent of the multiple wireless networks.

24. The method according to any one of claims 22 to 23, characterized in that, The one or more configuration settings are determined based on a machine learning (ML) model, wherein the ML model is trained to infer configuration adjustments based on historical performance data for the wireless technology.

25. The method according to any one of claims 22 to 24, characterized in that, The protocol layer includes the physical layer, network layer, and transport layer.

Images