Disconnected spectrum vacancy sharing assignment system

Abstract

In one aspect, a method of assigning spectrum vacancies is provided herein. The method includes receiving uplink signal data from a base station to a preprocessing engine, processing the raw spectrum recordings into a stream of frames of past looking spectrograms, inputting the stream of frames of past looking spectrograms into a forecasting model trained on spectrum data that has vacancies labeled, inferring, using the forecasting model, future vacancies, charting, using the forecasting model, a path through the future vacancies for assignment, communicating the future vacancies to the base station, and assigning the future vacancies to the user devices.

Description

DISCONNECTED SPECTRUM VACANCY SHARING ASSIGNMENT SYSTEMCROSS REFERENCE TO RELATED APPLICATIONS

[0001] This application claims the benefit of U.S. Provisional Patent Application No. 63 / 667990 filled on July 5, 2024, which is incorporated by reference herein in its entirety.FIELD OF THE INVENTION

[0002] The present invention relates to a system and method for disconnected spectrum vacancy sharing assignment.BACKGROUND

[0003] Citizens Broadband Radio Service (CBRS) represents a shared spectrum model implemented within the 3.5 GHz band (3550-3700 MHz) in the United States, designed to enhance spectrum efficiency and provide greater access for various wireless communication applications. This model, established by the Federal Communications Commission (FCC), allows for dynamic spectrum sharing among different classes of users through a tiered access structure. The three-tiered CBRS system 110 includes Incumbent Access 104, Priority Access 106, and General Authorized Access 108, as shown in FIG. 1.

[0004] Incumbent Access 104 is reserved for existing users such as the U.S. Navy and satellite ground stations, which are afforded protection from interference by lower-tier users. Priority Access 106 Licenses (PALs) are the second tier and are assigned through competitive bidding, granting licensees interference protection from General Authorized Access 108 (GAA) users. The third tier, GAA, permits the use of available spectrum by any entity, subject to coordination to avoid interference with higher-tier users.

[0005] Typically, this framework is managed by a Spectrum Access System (SAS) 102. The SAS 102 dynamically allocates frequencies based on real-time usage and protects incumbents and PALs from harmful interference. CBRS enables a diverse array of applications, including private LTE networks, broadband services in rural areas, and enhanced in-building wireless communications, ultimately promoting innovation and efficient use of the spectrum.

[0006] The Spectrum Access System 102 is an integral component within the framework of the Citizens Broadband Radio Service (CBRS), designed to dynamically allocate frequencies in the 3.5 GHz band based on real-time usage while ensuring interference protection for higher-tier users. The SAS operates as an automated frequency coordinator, employing advanced algorithms and comprehensive databases to manage spectrum access and uphold the integrity of the CBRS three-tiered hierarchy.

[0007] The SAS is typically responsible for several critical functions to maintain efficient spectrum utilization and protect incumbent operations. Firstly, incumbent protection is paramount, wherein the SAS continuously monitors the activities of high-priority users such as the U.S. Navy and satellite ground stations. Upon detecting the presence of incumbents, the SAS mandates that other users, including Priority Access License (PAL) and General Authorized Access (GAA) users, vacate the frequencies in the affected geographic area, thereby preventing interference. This protective mechanism is facilitated through sensing networks and direct information sharing from incumbents.

[0008] In addition to incumbent protection, the SAS is tasked with dynamic spectrum allocation for PAL and GAA users. PAL users, having acquired licenses through competitive bidding, receive prioritized spectrum access within their licensed areas. The SAS dynamically assigns frequencies to these users and ensures that their operations are free from interference by coordinating frequency assignments with GAA users. GAA users are permitted to access the available spectrum on an opportunistic basis, provided they do not interfere with higher- tier users.

[0009] the SAS is also responsible for interference management by utilizing algorithms, the SAS evaluates the radio environment, considering factors such as signal strength, geographic location, and operational parameters of CBRS devices. The SAS dynamically adjusts frequency assignments and power levels to optimize spectrum utilization while minimizing interference potential. This ensures a balanced and efficient spectrum-sharing environment.

[0010] Furthermore, the SAS ensures coordination and compliance with CBRS regulations. It verifies that devices are authorized and operating within designated parameters, coordinating with multiple SAS administrators to provide a seamless and cohesive management system across different regions. The SAS relies on Environmental Sensing Capability (ESC) networks, which comprise sensor arrays that detect incumbent activities and report to the SAS, triggering necessary protective measures. Additionally, the SAS communicates in real-time with CBRS devices, continuously adapting frequency assignments based on the current spectrum environment.

[0011] In summary, the SAS plays a pivotal role in the CBRS framework by dynamically allocating frequencies based on real-time usage, protecting incumbent operations, managing interference, and ensuring compliance with regulatory standards. This advanced system enables flexible and efficient spectrum sharing, supporting a wide range of wireless applications and fostering innovation within the telecommunications sector.

[0012] The Spectrum Access System 102, while crucial for the dynamic management and allocation of the CBRS spectrum, does exhibit certain problems and disadvantages that impactits overall efficiency and reliability. These issues stem from inherent system limitations, operational challenges, and the complexity of real-time spectrum management.

[0013] One significant problem associated with the SAS is its inherent latency. The system's need to continuously monitor and respond to real-time spectrum usage involves complex computations and decision-making processes. This can introduce delays in frequency allocation and reallocation, particularly in environments with high user density or rapid changes in spectrum demand. Such latency can hinder the timely provision of spectrum resources to users, leading to suboptimal performance and potential interruptions in service.

[0014] Another disadvantage is the potential for the SAS to assign incorrect frequencies to users. Despite the advanced algorithms employed, the system's reliance on real-time data and environmental sensing data can result in erroneous frequency assignments. Incorrect assignments can arise from inaccuracies in sensing data, outdated information, or misinterpretations of the spectrum environment. These errors can lead to inefficient spectrum utilization, increased interference, and degraded service quality for both PAL and GAA users.

[0015] Additionally, the SAS can inadvertently cause frequency jamming, particularly in bands above 5 GHz. This issue arises when the system's algorithms fail to adequately separate frequencies used by different services or users, leading to overlapping assignments. Frequency jamming not only disrupts the intended communications but also poses a significant challenge in maintaining reliable and interference-free operations. This can be particularly problematic in densely populated areas or scenarios where multiple high- bandwidth applications are competing for spectrum access.

[0016] The SAS's control node is responsible for enforcing spectrum access policies also contributes to inefficiencies. The control node operates based on predefined policies and rules, which may not always align with the dynamic and varied conditions of the actual spectrum environment. This rigidity can result in suboptimal spectrum management, where the policies enforced by the control node do not fully accommodate real-time variations in user demand, interference patterns, or environmental factors. Consequently, the system may fail to adapt quickly to changing conditions, leading to further inefficiencies and reduced overall effectiveness.

[0017] In conclusion, while the SAS is a pivotal component in managing the CBRS spectrum, it is not without its problems and disadvantages. The issues of latency, potential for incorrect frequency assignments, risk of frequency jamming, and inefficiencies arising from rigid control node policies highlight the challenges in achieving optimal real-time spectrum management. Addressing these limitations requires ongoing advancements in sensing technologies,algorithmic improvements, and adaptive policy frameworks to enhance the system's responsiveness and accuracy in dynamic spectrum environments.SUMMARY

[0018] In one aspect, a spectrum assignment system includes: a plurality of user devices; a base station in communication with the plurality of user devices; a spectrum forecasting system, includes a spectrum receiver to monitor a radio spectrum and collect live signal data in the form of raw spectrum recordings, a preprocessing engine communicatively coupled to the spectrum receiver to receive the raw spectrum recordings and generate a stream of frames of past looking spectrograms from the raw spectrum recordings, and a forecasting model, trained on spectrogram data that has vacancies labelled, the forecasting model communicatively coupled to the preprocessing engine to receive the live signal data in the form of the stream of frames of past looking spectrograms from the preprocessing engine as input data, infer the best possible next resource grid for assignment, where the base station receives and transmits uplink signal data to the preprocessing engine and the forecasting model to infer the future vacancies, and where the future vacancies are communicated back to the base station and the user devices.

[0019] The spectrum forecasting system is adapted to ingest the live signal data which preferably includes the last 20 milliseconds, and the path or best possible next resource grid includes the next 2 milliseconds. The reader will appreciate that the time frames referred to herein are predetermined, but can take on a number of values such as 100ms, 50ms, 10ms, etc. The reader will further appreciate that the expression “next resource grid” or “future vacancy” is to be interpreted with the appropriate context. As it will be apparent further below, inferring a resource grid or vacancy is not instantaneous, and so the “next” or “future” is actually frame 2 (which is the next frame / slot), not frame 1 which is the instant frame / slot. The machine learning model of the spectrum forecasting system may also include at least one of: time series forecasting, spectrogram, and generative models.

[0020] In one aspect, a method of assigning spectrum vacancies, includes receiving uplink signal data from a base station to a preprocessing engine, processing the raw spectrum recordings into a stream of frames of past looking spectrograms, inputting the stream of frames of past looking spectrograms into a forecasting model trained on spectrum data that has vacancies labeled, inferring, using the forecasting model, future vacancies, charting, using the forecasting model, a path through the future vacancies for assignment, communicating the future vacancies to the base station, and assigning the future vacancies to the user devices.

[0021] The spectrum forecasting method may also include where the live signal data includes the last 20 milliseconds, and the path includes the next 2 milliseconds. The spectrumforecasting method may also include where the live signal data includes the last 200 milliseconds, and the path includes the next 20 milliseconds. Other technical features may be readily apparent to one skilled in the art from the following figures, descriptions, and claims. The spectrum forecasting system may also include where the live signal data includes the last 200 milliseconds, and the path includes the next 20 milliseconds. The spectrum forecasting system may also include where the base station includes at least one of: a router, general node base station, and a standalone access point.

[0022] The spectrum forecasting system may also include the forecasting model includes a machine learning model. The spectrum forecasting system may also include where the spectrum assignment system is deployed to an FPGA. The spectrum forecasting system may also include where the preprocessing engine and forecasting model are located inside the basestation.

[0023] Other technical features may be readily apparent to one skilled in the art from the following figures, descriptions, and claims.TERMS AND DEFINITIONS

[0024] 5G - Fifth generation telecommunications system.

[0025] Open RAN / ORAN - open radio access networks.

[0026] RAN - Radio Access Network.

[0027] gnodeB - basestation, generational node basestation, also includes routers and other access points.

[0028] Secondary user, secondary user node - used interchangeably to describe a secondary user which can be a single device, or a basestation that serves other multiple devices.BRIEF DESCRIPTION OF THE DRAWINGS

[0029] To easily identify the discussion of any particular element or act, the most significant digit or digits in a reference number refer to the figure number in which that element is first introduced.

[0030] FIG. 1 illustrates a diagram of the CBRS system, showcasing the tiered structure managed by the SAS for spectrum allocation among Incumbent, Priority Access, and General Authorized Access users.

[0031] FIG. 2 illustrates a block diagram of the spectrum forecasting system, including the preprocessing engine and forecasting model for predicting future spectrum vacancies.

[0032] FIG. 3 illustrates a schematic representation of the SAS managing the tiered CBRS spectrum access framework.

[0033] FIG. 4 illustrates a system diagram of a base station configured with commodity hardware or an open-source stack, interfacing with user devices and central network components.

[0034] FIG. 5 is a graph illustrating the predicted and recommended resource grids for spectrum sharing based on interference management and vacancy forecasting.

[0035] FIG. 6 illustrates a block diagram of the prediction application deployment architecture, showcasing the integration of SDR / RF receivers, preprocessing, inference, and control modules within various deployment infrastructures.

[0036] FIG. 7 illustrates a block diagram of the multi-threaded inference pipeline for spectrum forecasting and resource assignment in the spectrum assignment system.

[0037] FIG. 8 illustrates a process diagram for spectrum analysis, showcasing the transformation of IQ data into spectrograms and subsequent vacancy forecasting for transmission pathway recommendations.

[0038] FIG. 9 is a graph illustrating the comparison between predicted grids and target grids, along with recommended paths for spectrum vacancy assignment.

[0039] FIG. 10 illustrates a system diagram of a gNB and a central server integrated with Open5GS, RIA docker, srsRAN docker, and connected UE devices.

[0040] FIG. 11 illustrates a network diagram depicting resource allocation between a base station and mobile devices using a resource grid.

[0041] FIG. 12 illustrates a schematic representation of spectrum vacancy forecasting and resource block assignment within a wireless communication system.

[0042] FIG. 13 illustrates a diagram of spectrum sharing and coexistence within the CBRS framework, highlighting the tiered access structure managed by the SAS.

[0043] FIG. 14 illustrates schematically how signal data is used to train and test a model, and then refine the model to identify and predict future vacancies.

[0044] FIGS. 15A, 15B, 15C and 15D illustrate schematically the lookback period, inference period and forecast period (15A and 15B); per frame assignment (15C) and per slot assignment (15D). The reader will appreciate that 5G scheduling is better adapted to leverage per frame assignment.Detailed Description

[0045] It will be appreciated that numerous specific details are set forth in order to provide a thorough understanding of the example embodiments described herein. However, it will be understood by those of ordinary skill in the art that the embodiments described herein may be practiced without these specific details. In other instances, well-known methods, procedures and components have not been described in detail so as not to obscure the embodiments described herein. Furthermore, this description and the drawings are not to be considered as limiting the scope of the embodiments described herein in any way, but rather as merely describing the implementation of the various embodiments described herein.

[0046] It should be noted that terms of degree such as "substantially", "about" and "approximately" when used herein mean a reasonable amount of deviation of the modified term such that the end result is not significantly changed. These terms of degree should be construed as including a deviation of the modified term if this deviation would not negate the meaning of the term it modifies.

[0047] In addition, as used herein, the wording “and / or” is intended to represent an inclusive- or. That is, “X and / or Y” is intended to mean X or Y or both, for example. As a further example, “X, Y, and / or Z” is intended to mean X or Y or Z or any combination thereof.

[0048] It should be noted that the term “coupled” used herein indicates that two elements can be directly coupled to one another or coupled to one another through one or more intermediate elements.

[0049] The embodiments of the systems and methods described herein may be implemented in hardware or software, or a combination of both. These embodiments may be implemented in computer programs executing on programmable computers, each computer including at least one processor, a data storage system (including volatile memory or nonvolatile memory or other data storage elements or a combination thereof), and at least one communication interface. For example, and without limitation, the programmable computers (referred to below as computing devices) may be a server, network appliance, embedded device, computer expansion module, a personal computer, laptop, personal data assistant, cellular telephone, smart-phone device, tablet computer, a wireless device or any other computing device capable of being configured to carry out the methods described herein.

[0050] In some embodiments, the communication interface may be a network communication interface. In embodiments in which elements are combined, the communication interface may be a software communication interface, such as those for interprocess communication (IPC). In still other embodiments, there may be a combination of communication interfaces implemented as hardware, software, and combination thereof.

[0051] Program code may be applied to input data to perform the functions described herein and to generate output information. The output information is applied to one or more output devices, in known fashion.

[0052] Each program may be implemented in a high level procedural or object oriented programming and / or scripting language, or both, to communicate with a computer system. However, the programs may be implemented in assembly or machine language, if desired. In any case, the language may be a compiled or interpreted language. Each such computer program may be stored on a storage media or a device (e.g., ROM, magnetic disk, optical disc) readable by a general or special purpose programmable computer, for configuring and operating the computer when the storage media or device is read by the computer to perform the procedures described herein. Embodiments of the system may also be considered to be implemented as a non-transitory computer-readable storage medium, configured with a computer program, where the storage medium so configured causes a computer to operate in a specific and predefined manner to perform the functions described herein.

[0053] Furthermore, the system, processes and methods of the described embodiments are capable of being distributed in a computer program product comprising a computer readable medium that bears computer usable instructions for one or more processors. The medium may be provided in various forms, including one or more diskettes, compact disks, tapes, chips, wireline transmissions, satellite transmissions, internet transmission or downloading, magnetic and electronic storage media, digital and analog signals, and the like. The computer useable instructions may also be in various forms, including compiled and non-compiled code.

[0054] The various embodiments described herein generally relate to methods and associated systems configured to implement the methods for assigning vacant frequencies to a user device in a spectrum assignment system. The method presents a pathway for shared spectrum schemes that do not rely on a central access network, allowing independent nodes to assign resource usage independently of other users.

[0055] The description uses the expression “chart a path through vacancies”, which includes “define the next resource grid that accounts for anticipated occupancies”. In some cases, these expressions are used interchangeably, and refer to the advantages of the instant system and method to leverage machine-learning models trained on signal data to identify, based on live data, resource grids in the next frame or frames that are anticipated to be occupied, and permit the system and method to allow user devices to take advantage of anticipated vacancies, resulting in a more efficient use of the available spectrum, subject to other constraints described herein.

[0056] As mentioned above, the reader will appreciate that the inference process does take time. Thus, referring now to FIGURES 15A-15D, it will be apparent that the resource grid of the current frame to has already been assigned during the inference period, and thus the forecast period is actually the frame at ti . Thus, the reader will appreciate that the expression “next” in the context of predicting future vacancies is actually the next frame after the current frame. FIG. 15B illustrates the lookback period as it increases, but in all cases the forecast period is at ti. The reader will also appreciate that this limitation is due to current hardware and software implementations. It is possible that future hardware and software capabilities will further reduce the prediction time, and it may be possible to eventually predict the frame starting at to, or very shortly thereafter. For example, it will be possible to make predictions at the slot level or subframe, depending on the 5G configuration, using 10 or more subframes or 10 or more ms in total, and obtaining a prediction and the resulting assignment in 1 ms. Examples of such short timelines are shown in FIGS. 15C and 15D.

[0057] Advantageously, the spectrum assignment system can be used by anyone authorized within a spectrum sharing system that is not the incumbent. Examples include users with a Priority Access license (PAL), generally authorized access users (GAA), or secondary users. Collectively, these users are referred to as secondary users.

[0058] The invention leverages a robust data acquisition and labeling methodology to enable accurate spectrum vacancy forecasting and channel state prediction. Datasets are generated using a combination of synthetic, testbed-generated, and ambiently captured radio frequency (RF) data. The data collection process is structured to capture a wide variety of scenarios, including different sample rates, center frequencies, antenna configurations, gain settings, and environmental conditions (e.g., indoor, outdoor, varying weather).

[0059] For the vacancy forecasting model, time series or spectrogram data are labeled at the resource grid or pixel level, with each time-frequency element marked as “vacant” or “occupied” based on energy detection thresholds above the noise floor. This labeling enables the model to learn to identify and forecast spectrum vacancies with high temporal and frequency resolution.

[0060] For the channel state forecasting model, the system is trained to predict the optimal configuration for each resource block, based on observed channel effects. In one embodiment, the prediction is embodied in the form of Modulation and Coding Scheme (MCS) code, which has modulation, error correction and other similar signals. Data is generated by transmitting known OFDM signals with varying MCS codes and gain settings, passing them through real or synthetic channels, and measuring the resulting biterror rates and other channel metrics. The optimal configuration for each resource block is determined by comparing the received signal to the ground truth, and the bestperforming settings are used as labels for model training.

[0061] The invention is designed to operate at the granularity of the 5G resource grid, which consists of a matrix of resource blocks (RBs) defined in both time (slots, subframes, frames) and frequency (subcarriers). The system can forecast and assign resources at the frame or slot level, with each resource block potentially assigned a unique MCS code, gain, and other transmission parameters.

[0062] The machine learning inference system is preferably integrated with the gNodeB scheduler, either by replacing or constraining the default resource grid with the model’s predicted grid. The scheduler uses the model’s output to assign uplink and downlink resources, ensuring that transmissions are scheduled in time-frequency blocks predicted to be vacant or optimal, thereby minimizing interference and maximizing throughput.

[0063] The system supports both Time Division Duplex (TDD) and Frequency Division Duplex (FDD) operation, with the ability to process and forecast for both uplink and downlink channels, potentially using separate models or ensembles for each. It will be apparent to a person skilled in the art that FDD preferably uses two channels and two separate inferrer systems, but it will also be apparent that there is no limite to the number of observation channels and inference systems, as long as proper orchestration is arranged.

[0064] The spectrum assignment system is implemented as a modular architecture, comprising the following components:Spectrum Receiver: Typically a software-defined radio (SDR) capable of capturing wideband IQ data from the spectrum of interest.Preprocessing Engine: Processes raw IQ data into spectrograms or time series suitable for model input, applies masking to blank out self-transmissions, and prepares data batches for inference.Forecasting Model(s): Deep learning models (e.g., ConvLSTM, CNN, generative models) trained to forecast future spectrum occupancy and / or channel state at the resource block level.Inference Application: Deployed on edge hardware (e.g., Jetson Orin, Hailo-8, Raspberry Pi) or in the cloud, capable of real-time inference with sub-millisecond latency,Scheduler Integration: Interfaces with the gNodeB or other base station software (e.g., srsRAN), providing resource block masks or full resource grids for scheduling. In a transmission configuration, the scheduler integration will also be adapted to generate MCS codes. Advantageously, edge accelerator hardware can include one or more of a GPU, an inference ASIC, an FPGA, or other edge hardware.

[0065] The reader will appreciate that when the spectrum assignment system described herein is implemented on a gNodeB, the SRD acts as the receiver and transmitter,

[0066] The system supports deployment in a variety of configurations, including:Embedded within the base station (e.g., inside the gNodeB stack).As a standalone inference container communicating with the base station via network protocols (e.g., UDP, ZeroMQ).Distributed across multiple receivers and inference nodes for wide-area or multiband monitoring.

[0067] The invention contemplates the use of multiple models operating in concert, such as:A vacancy forecasting model to predict interference and spectrum availability.A channel state forecasting model to predict optimal transmission parameters (MCS, gain, FEC).Optional beamforming and angle-of-arrival estimation models for spatial interference mitigation.

[0068] The system includes algorithms for path recommendation, which process the model’s predicted vacancy or channel state grids to generate a recommended sequence of resource blocks for assignment. These algorithms may optimize for criteria such as:Maximizing contiguous block allocation.Preferring spectrum near the center of the band.Minimizing probability of collision or interference.Adhering to dynamic consumption limits (e.g., as set by a Spectrum Access System or local policy).

[0069] The recommended path is communicated to the base station scheduler, which assigns the corresponding resources to user devices for uplink or downlink transmission. In other words, the RAN scheduler controls the PUSCH and the PDSCH, to send the appropriate signalling commands to the user devices.

[0070] The system is designed for scalability, supporting:Multiple receivers for wideband or geographically distributed monitoring.Interoperability with third-party communication systems via open standards.Low-cost, low-power edge deployment using commercially available SDRs and Al accelerators.

[0071] The architecture supports both fully autonomous operation (disconnected from a central Spectrum Access System) and operation with limited SAS input (e.g., dynamic consumption limits), enabling flexible deployment in a variety of regulatory and operational environments.

[0072] The invention achieves ultra-low latency operation, with inference and scheduling cycles in the order of less than 10 milliseconds, enabling real-time adaptation to rapidly changing spectrum environments. This is a significant improvement over traditional rulebased or SAS-coordinated systems, which may have latencies of 100 ms or more.

[0073] The system is applicable to a wide range of use cases, including:Direct-to-device (D2D) satellite communications, enabling coexistence between terrestrial and satellite networks in shared bands.Dynamic spectrum sharing in congested, contested, or border regions with uneven regulatory environments.Interference mitigation in dense urban, tactical, or rural deployments.Integration with beamforming and null steering for spatial interference suppression across regional or international borders where licenses are assigned and managed differently.

[0074] The spectrum assignment system consists of a secondary user's 5G base station, which has been allowed to operate on the shared spectrum of an incumbent (such as another operator or a radar like in CBRS) and to consume spectrum and assign resources in a manner that is independent and disconnected from any spectrum assignment systems, without interfering with the incumbent nor getting interfered with by the incumbent.

[0075] The spectrum assignment system operates by forecasting in real-time future vacancies in time and in frequency on spectrum in the order future milliseconds, then assigns resource blocks in time and in frequency to its userbase or network for their own consumption. The spectrum assignment system aims to minimize the risk of collisions, while allowing for some tolerance for collisions from the incumbent to the secondary user.

[0076] The RF spectrum is increasingly constrained and limited at a time where demand in all areas of spectrum is increasing. Militaries and a number of government based incumbent technologies control large swaths of spectrum for critical applications. Many of theseincumbents do not use the spectrum 100% of the time, leaving significant amounts of spectrum bandwidth available for other uses. In an effort to enable spectrum efficiency and unlock more spectrum for more users, spectrum sharing technologies have been developed and explored. Many of these spectrum sharing technologies create a system for non-incumbent, secondary users to operate on the same bands as an incumbent by managing scheduling of spectrum utilization. These technologies are often controlled by a "Spectrum Access Service" or database which complete a series of tasks such as detection of the incumbent (via environmental data monitoring), ranking and scheduling by user priority, coordinating amongst users and access points. The integration of these technologies and solutions presents a complex challenge, stemming from coordinating multiple disparate users in real-time in an evolving environment. Interference management is the top challenge in coordinating the realtime sensing with predictive analytics and mechanisms for assigning allocations to secondary users of different priorities. Methods and systems for managing interference need to balance rapid decisions for frequency selection, power levels, and time-frequency coding while ensuring fairness among users and prevention of interference. There is a significant reliance on environmental sensing at the Spectrum Access System itself which then disseminates allocations to secondary users. Scalability and coordinating multiple users becomes increasingly intractable the more users are required. Another challenge is the presence of multiple SAS administrators operating in the same region. Coordinating their respective assignments is critical. For a Spectrum Access System to work, the mechanisms for sending commands to secondary users (and their own users) must be robust and free off interruption

[0077] SAS systems are complex, and it follows that their implementations are similarly complex. In addition, they are prone to failures resulting in poor service quality, interference with the incumbent, regulatory penalties and spectrum opportunity cost (from operational efficiencies).

[0078] A spectrum assignment system is proposed herein. The spectrum assignment system is a means for spectrum sharing that reduces or eliminates the need for a complex Spectrum Access Systems, allowing secondary users or networks of secondary users to make assignments themselves either fully autonomously, or as part of a real-time management policy.

[0079] Turning now to FIG. 2, a spectrum assignment system 210 is shown therein. The spectrum assignment system 210 comprises a base station 202, at least one user device 204, a preprocessing engine 206, and a forecasting model 208.

[0080] The base station 202 is providing service to cellular data user devices 204. The base station 202 continuously receives and transmits uplink signal data from user devices 204. Anuplink signal refers to the transmission of data from a user device 204, such as a mobile phone or laptop, to a base station 202 or network infrastructure. This is a fundamental aspect of wireless communication systems, including cellular networks like 4G LTE and 5G, as it enables devices to send information to the network for processing, storage, or further transmission. In the context of a cellular network, the base station 202, also known as a cell tower or eNodeB in 4G LTE, or gnodeB in 5G, plays a pivotal role in managing uplink signals. The base station 202 is equipped with antennae and radio equipment that receive and process uplink signals from multiple user devices 204 within its coverage area.

[0081] Uplink signals are generated by user devices 204 when they need to send data to the network. This data can include voice calls, text messages, internet browsing requests, or any other type of communication. The user device 204 modulates this data onto a radio frequency carrier and transmits it through its antenna. The base station 202, which typically consists of multiple antennae, receives the uplink signals transmitted by user devices 204. These antennae are designed to capture radio frequency signals over a broad range of frequencies and from various directions. Upon receiving the uplink signals, the base station 202 processes these signals to extract the transmitted data. This involves several steps, including demodulation, decoding, and error correction. The base station's radio equipment converts the radio frequency signals into digital data that can be further processed by the network.

[0082] The base station 202 dynamically manages and allocates radio resources to ensure efficient and reliable communication. This includes assigning specific frequency bands, time slots, or codes to different user devices 204 to avoid interference and optimize the use of available spectrum. Techniques such as Time Division Multiple Access (TDMA), Frequency Division Multiple Access (FDMA), and Code Division Multiple Access (CDMA) are employed to manage multiple uplink signals simultaneously. Once the base station 202 has processed the uplink signals and extracted the data, it forwards this data to the core network for further handling. The core network is responsible for routing the data to its final destination, whether it's another user device 204, an internet server, or a service platform.

[0083] The base station also manages the Quality of Service (QoS) to ensure that different types of data receive the appropriate level of service. For example, voice calls may require low latency and high reliability, while web browsing can tolerate some delays. Additionally, the base station controls the transmission power of user devices to maintain a balance between minimizing interference and ensuring sufficient signal strength for reliable communication. This seamless interaction between user devices and base stations is fundamental to the functioning of modern wireless networks.

[0084] The uplink signal is transmitted to the preprocessing engine 206 to identify spectrum vacancies. In some examples, spectrum vacancies, which may also be referred to as “whitespace” are forecasted, so they can be assigned to a transmitter or to enable a transmitter to avoid interference. The transmitter can be friendly, adjacent, local, etc. In some examples, this avoidance of interference helps to overcome congested and / or contested spectrum environments (e.g., in tactical or civilian settings). In some examples, the spectrum assignment system 210 includes a preprocessing engine 206 and a spectrum forecasting model 208. In some examples, a spectrum forecasting model 208 is trained on extensive spectrogram data that has vacancies (areas with no signal) labeled. In some examples, the data comes from raw IQ., The spectrum forecasting model 208 receives a stream of frames of past-looking spectrograms as the input data (e.g., the last 10 milliseconds or 20 milliseconds or the last 10-20 frames), infers / predicts future vacancies (e.g., next frame, next 2 milliseconds or next 5 milliseconds), and charts a path of potentially vacant areas (e.g., those areas that are predicted with a high probability to be vacant) for assignment. For example, the objective may be to have the “false negative” rate (the rate in which a vacancy was inferred but it was actually an occupancy) to be better that the error correction threshold of the signal. A person skilled in the art will recognize that some error correction schemes require 25% of the signal to be redundant or coded for error correction. If the false negative rate is less than the error correction scheme requirement, it follows that the rate of collisions will be below the error correction scheme requirement. Thus, the error correction scheme is able to reconstruct the waveform even if there is a collision. In another example, high probability can also mean the use of a statistical method to measure the probability of vacancy. The probability will in this case also depend in part on the error correction scheme of the underlying signal.

[0085] In some examples, the input data includes data from the last 5-30 milliseconds. In some examples, the input data includes data from the last 50-200 milliseconds. In some examples, the prediction includes the next 1-5 milliseconds. In some examples, the prediction includes the next 5-30 milliseconds. In some examples, the prediction includes the next 100 milliseconds. In some examples, the prediction includes the next one or two frames (e.g., 5G frames or LTE). In some examples, the input includes 1 to 15 times the prediction (e.g., 1 to 15 times 5 milliseconds if 5 milliseconds is the prediction or 1 to 15 frames if 1 frame is the prediction). In some examples, the input includes about 1 time the prediction. In some examples, the input includes about 2 times the prediction. In some examples, the input includes about 10 times the prediction. In some examples, the input includes only (i.e., is limited to) about 1 or 2 or 10 times the prediction (e.g., 1 or 2 or 10 times 5 milliseconds if 5 milliseconds is the prediction or 1 or 2 or 10 frames if 1 frame is the prediction). In someexamples, radio spectrum monitoring is based on a fixed center frequency, as wide band as the SDR and host system can allow (e.g., as opposed to sweeping through multiple bands).

[0086] In some examples, identifying future vacancies right above a certain threshold can lead to usage for other transmitters to consume spectrum more efficiently. In some examples, identifying a vacancy identifies an unoccupied (i.e., not occupied but with availability) area. In some examples, the system or method identifies or recommends a path of vacancies in time and in frequency in which the probability of collisions is very low. The reader is invited to review the above description relating to probabilities of collisions within the discussion of error correction schemes.

[0087] In some examples, the system or method relies on reducing collisions to the point that the robustness systems of a given communications protocol (e.g., forward error correction) overcome the impact of any false negatives (forecasted vacancies that are not actually vacant). In some examples, the system or method results in a path identification or recommendation in time and in frequency to reduce collisions below a threshold where error correction always succeeds. In some examples, the system or method enables loosening specifications such that the lowest overhead for error correction can be applied, at the highest symbol rates (highest bandwidths).

[0088] In some examples, the system is implemented using a low cost embedded device and a low cost (or mid cost) commercially available SDR, using a GPU enabled embedded device and a low cost (or mid cost) commercially available SDR, and / or using a workstation or laptop and a GPU and any commercially available SDR. In some examples, the system includes a single receiver and a single inferer. In some examples, the system includes multiple receivers. In some examples, the system includes multiple inferrers. In some examples, the system includes a multi-receiver configuration passing predictions into an analytics application. A multi-receiver configuration passing predictions into an analytics application may enable three topologies: monitoring of wider bandwidth, monitoring of wider geographies, and / or monitoring of the same spectrum but with different settings (e.g., antennae, gain profiles, filters, signal processing profiles). In some examples, deep learning techniques is used. Deep learning is a type of machine learning that involve training algorithms on large datasets. In some examples, the system includes a remote, scalable, interoperable sensor that forecasts whitespace in real-time, without reliance on inputs from a network.

[0089] In some examples, the system or method is a passive monitoring system or method. In some examples, the system or method outputs the path recommendation or identification to another system or method for use. In some examples, the system or method is integrated with a transmitter to apply the path recommendation or identification for transmission.

[0090] In some examples, the system or method includes cognitive software-defined radiobased spectrum monitoring, vacant channel forecasting, and / or channel assignment. In some examples, the system or method includes a radiofrequency (RF) sensing system or method, equipped with automated spectrum analytics software, which can identify activity on a wireless spectrum, then forecast with high probability spectrum availability for networks to operate on. In some examples, the system or method includes dynamic spectrum management enabling users to adapt RF communications pathways as necessary. In some examples, the system or method includes a distributed software-defined radio (SDR) spectrum sensing system integrated with a machine learning (ML) application that can identify activity on a spectrum, identify recent whitespaces (or zones of spectrum vacancy), and forecast where a high probability of whitespace is expected in the immediate future. In some examples, the system or method includes a network configuration application that is able to ensure that any operating ratios are assigned to channels that will have a high probability of availability. In some examples, the system or method includes can be scaled, with multiple receivers working together, enabling an increase in spectrum that is monitored in real-time, or an increase in geographic coverage. In some examples, the system or method includes a trained deep learning (DL) model integrated into a SDR for forecasting areas that whitespaces are expected to be present with high probability, based on an input stream of live signal data from the SDR. In some examples, the system or method includes design architecture based on commercial off-the-shelf SDR and portable computer equipment, designed for scalability with multiple receivers. In some examples, the system or method includes software components within the system for pre-processing, operating any models, logging, and sharing data and inference results. In some examples, the system or method combines emerging ML methodologies from the RF and other domains with COTS SDR hardware and open standards and uses these techniques to improve real-time wireless network configuration based on real-world measurements.

[0091] In some examples, the system or method includes a remote, scalable, interoperable sensor that forecasts whitespace in real-time, without reliance on inputs from a network. In some examples, the system or method is low cost-per channel and has low power requirements over traditional architectures which stream signal data to a large scale (e.g., hyperscalers) cloud services provider (e.g., Google Amazon Azure). In some examples, the system or method implements whitespace forecasting using generative methods for machine vision or occupancy anticipation methods for robot navigation or few-shot machine learning techniques that are able to learn accurate predictions with small data samples. In some examples, the system or method is integrated with commercially available SDR hardware and configured for interoperability with third party communication systems using open standards.In some examples, the system or method is fully integrated and includes rear-view inference, forward-looking forecasting, with the ability to be reconfigured and redeployed.

[0092] In some examples, in use the SDR is tuned to the specific spectrum, sample rate, channel, gain settings for its task. The SDR ingests data and produces a stream of IQ samples. These IQ samples can be output as (1) time series or (2) spectrograms. In both cases the outcome may be the same, and functionally they may be the same, the difference is in performance, behavior, and speed. In some examples, time series can provide a lot more detail on the time axis but have additional computational needs such as FIR filtering. In other examples, spectrograms can provide more comprehensive coverage but with much less detail and have their own additional computational needs (basis state transformation). In both cases, the data is preprocessed. For spectrogram slicing, a finite component of the stream is processed (for all stream elements), then a power spectrum distribution for each slice is obtained and passed in batches of power spectrum distributions into the ML model. Alternatively, when using time series the data goes through a number of FIR filters based on the frequency resolution that is desired, and the data is then passed through the model or series of models. The model or series of models may perform three sequential tasks: identifying vacancies in the input data for the past few “input examples” - e.g. t=-n, - n+1 ,=n+2,...-1 ,0, forecasting an estimate of where vacancies are expected for a specific number of frames in the future based on the input data (e.g. t=1 ,2,3), and recommending an optimal path through the forecasted vacancies. In some examples, the recommended path is then integrated into the desired channel assignment framework of the user.

[0093] In at least one embodiment, the spectrum assignment system 210 can operate without a Spectrum Access System 102. Without a spectrum access system, the spectrum assignment system can have a user-defined (and pre-negotiated) consumption limit. The spectrum assignment system can forecast vacancies based on the present environment, and chart possible opportunities for consumption. The spectrum assignment system 210 can then makes assignments to its own userbase staying within the consumption limit. The consumption limit may be dynamically reduced based on real-time conditions.

[0094] In the embodiment where there is no Spectrum Access System, there is only local environmental sensing and allocations are fully independent of any incumbent or spectrum assignment technology. There is no link to the Spectrum Access System, therefore no coordination is needed. Interference management can happen locally and is at the risk of the secondary user, not the SAS. Any interference that can occur from the incumbent (or other users) is not damaging to the secondary user provided it is below an acceptable threshold. Any interference from the secondary user to the incumbent are generally within the policy limits of any spectrum sharing agreement (that is, secondary users are already not transmittinga signal that can affect the incumbent). Any interference from the secondary user to other users will remain within under acceptable limits.

[0095] In at least one embodiment, the spectrum assignment system 210 can operate with a limited Spectrum Access System 102. In this embodiment, the spectrum assignment system receives real-time consumption limit updates from the SAS (which may be updated frequently or as needed), then the system forecasts vacancies based on the present environment and charts possible opportunities for consumption. The spectrum assignment system can then make assignments to its own userbase staying within the consumption limit. The consumption limit may be dynamically reduced based on real-time conditions. The spectrum assignment system reduces and / or eliminates the challenges and the risks that come from them.

[0096] In the embodiment where a Spectrum Access System is used, the SAS can provide updates on the consumption limit (which can be reduced to zero should the need arise), rather than managing assignments. A robust real-time two-way link between the users and the SAS is not needed to maintain system operation. In use case 2, secondary users function independently within the limits imposed on them.

[0097] In at least one embodiment, interference management happens with the secondary users.

[0098] In at least one embodiment, algorithmic complexity only needs to concern a secondary user and its own userbase. In at least one embodiment, the SAS does not need to rely on a vast environmental sensing system, and the environmental sensing the secondary user does is built into its system design (piggy backing off of the received uplink signals). By the same token, there is no dissemination of allocations to secondary users from the SAS. the allocation is done by the secondary user. Therefore, the spectrum assignment system 210 is more simplified, performant, and scalable, eliminating the need of an SAS to coordinate several disparate needs. In at least one embodiment, the SAS (or grouping of SAS's) manages limits, and the secondary users operate independently. In at least one embodiment, there is no need for a robust and mission critical backhaul link between secondary users and the SAS, and a limited one-way channel is sufficient.

[0099] In at least one embodiment, the spectrum assignment system solution can include multiple deep learning methodologies including state-of-the art forecasting and inference techniques from the RF domain as well as state-of-the-art forecasting techniques from other domains.

[0100] The ML can be integrated with COTS SDR hardware and open source 5G software, turning a basestation into a sensor (in addition to a communications device). This is advantageous as most integrations of Al and 5G are not within the radio or radio host butwithin a RAN Intelligent Controller. This architecture allows for very fast capture-infer-act cycles on the order of less than a millisecond, which is novel and in contrast to RIC methods which have latencies of 100ms to several seconds.

[0101] In at least one embodiment, the use of ML to perform dynamic real-time radio resource assignment (i.e., control of non-ml RAN software) is provided. In at least one embodiment, the spectrum assignment system does not rely on an SAS for high-performance dynamic spectrum sharing. The spectrum assignment system 210 is more scalable, dynamic, and performs more efficiently than traditional SAS based systems. In at least one embodiment, a remote, scalable, interoperable secondary user node (e.g. basestation or user) that inherently has a sensor that forecasts whitespace in real-time, without reliance on inputs from a SAS.

[0102] The spectrum assignment system informs wireless network configuration in real-time based on real-world measurements. The spectrum assignment system can be a low cost-per channel and have low power requirements over traditional architectures which stream signal data to hyperscalers cloud infrastructure. The spectrum assignment system can procure networks that eliminate the need for manual human intervention.

[0103] In one embodiment, the spectrum assignment system 210 forecasts whitespace using generative methods designed for machine vision, or occupancy anticipation methods developed for robot navigation, or few-shot machine learning techniques capable of making accurate predictions with limited data samples. None of these methods have been previously integrated with commercially available Software Defined Radio (SDR) hardware on an open- source 5G stack, nor have they been configured for interoperability with third-party communication systems using open standards. In one embodiment, the spectrum assignment system 210 includes rear-view inference, forward-looking forecasting, with the ability to be reconfigured and redeployed.

[0104] The spectrum assignment system is a low cost method of forecasting activity on spectrum for the purpose of channel assignment that exploits the uplink channels of a secondary user node. It enables secondary users to dynamically find the right spectrum to consume and reduce interference events. In at least one embodiment, the spectrum assignment system has the potential to maximize the usage of wireless spectrum.

[0105] In at least one embodiment, the spectrum assignment system has limited or eliminated reliance on an SAS. In at least one embodiment, the spectrum assignment system has ultra-low latency. In at least one embodiment, the spectrum assignment system has fully independent and disconnected sharing of spectrum. In at least one embodiment, the spectrum assignment system has high performance avoidance of collisions relying on forecasting ratherthan scheduling. In at least one embodiment, the spectrum assignment system has high performance efficient consumption of spectrum. In at least one embodiment, the spectrum assignment system can be deployed to multiple basestations within a network to give geographic granularity to secondary user's allocations. In at least one embodiment, the spectrum assignment system enables scalability in spectrum sharing regimes. In at least one embodiment, the spectrum assignment system can apply to any software-defined wireless technology.

[0106] A gnodeB or other basestation or radio that has been modified to pass its uplink IQ samples into one or more deep learning preprocessing engines operates with its userbase. The inference solution can operate from spectrograms or time series directly. The inference solution is designed for rapid inference, performing an inference task on the equivalent of 20- 100ms of data in under one millisecond. The inference solution is equipped with a forecasting solution on it that forecast vacancies in time and frequency on wireless spectrum a few milliseconds into the future, based on the past 20 to 100 milliseconds. The model produces a prediction in the form of a "a recommended path to consume in that forecasted duration". The path consists of time / frequency resource blocks to use in the next few milliseconds. The predictions are passed back into the gnodeB software. The gnodeB control software takes that recommended path and uses it for its own allocations when assigning resource blocks to its own users, assigning it to its own users. In essence, this gives the spectrum assignment system the ability to consume only predicted vacancies and therefore avoid collisions with other users. The system works can be used in environments that demonstrate periodicity, such as the linearly frequency modulated chirps of radar signals. Because the spectrum assignment system does not rely on assignments from an SAS, is independent, and performs at millisecond scale, it has superior performance to all other methodologies, and can provide disconnected spectrum sharing.

[0107] In at least one embodiment, a standalone link can be used. In this embodiment, the secondary user node is not a basestation or network router, but rather it can be a single standalone user that may be broadcasting or communicating with another user. The spectrum assignment system can anticipate vacancies in the same manner, then passes them along to the standalone link.

[0108] In at least one embodiment, the spectrum assignment system is configured to forecast vacancies, perform rapid inference, integrate on a gnodeB, process IQ from the gnodeB, pass predictions to the gnodeB, and make the spectrum assignments dynamically.

[0109] In at least one embodiment, a range of ML methodologies can be used such as but not limited to: time series forecasting, spectrogram, generative - infer on radio - infer on cloud- infer on gnodeB - low cost, single channel - medium cost, multi-channel. In at least one embodiment, the spectrum assignment system can be deployed to an FPGA. In at least one embodiment, the preprocessing engine 206 and forecasting model 208 can be located inside the basestation 202 and make predictions inside the gnodeB stack instead of beside it. In at least one embodiment, the preprocessing engine 206 and forecasting model 208 can be in a separate docker container on the same machine.

[0110] It will be appreciated that numerous specific details are set forth in order to provide a thorough understanding of the example embodiments described herein. However, it will be understood by those of ordinary skill in the art that the embodiments described herein may be practiced without these specific details. In other instances, well-known methods, procedures, and components have not been described in detail so as not to obscure the embodiments described herein. Furthermore, this description and the drawings are not to be considered as limiting the scope of the embodiments described herein in any way, but rather as merely describing the implementation of the various embodiments described herein.

[0111] The embodiments of the systems and methods described herein may be implemented in hardware or software, or a combination of both. These embodiments may be implemented in computer programs executing on programmable computers, each computer including at least one processor, a data storage system (including volatile memory or nonvolatile memory or other data storage elements or a combination thereof), and at least one communication interface. For example, and without limitation, the programmable computers (referred to below as computing devices) may be a server, network appliance, embedded device, computer expansion module, a personal computer, laptop, personal data assistant, cellular telephone, smart-phone device, tablet computer, a wireless device, or any other computing device capable of being configured to carry out the methods described herein.

[0112] In some embodiments, the communication interface may be a network communication interface. In embodiments in which elements are combined, the communication interface may be a software communication interface, such as those for interprocess communication (IPC). In still other embodiments, there may be a combination of communication interfaces implemented as hardware, software, and combination thereof.

[0113] Program code may be applied to input data to perform the functions described herein and to generate output information. The output information is applied to one or more output devices, in known fashion.

[0114] Each program may be implemented in a high-level procedural or object-oriented programming and / or scripting language, or both, to communicate with a computer system. However, the programs may be implemented in assembly or machine language, if desired. Inany case, the language may be a compiled or interpreted language. Each such computer program may be stored on a storage media or a device (e.g., ROM, magnetic disk, optical disc) readable by a general or special purpose programmable computer, for configuring and operating the computer when the storage media or device is read by the computer to perform the procedures described herein. Embodiments of the system may also be considered to be implemented as a non-transitory computer-readable storage medium, configured with a computer program, where the storage medium so configured causes a computer to operate in a specific and predefined manner to perform the functions described herein.

[0115] The following description is made with reference to the drawings.

[0116] Overview of the Spectrum Assignment System

[0117] The present invention provides a system and method for dynamic, real-time spectrum assignment in shared wireless environments, such as the Citizens Broadband Radio Service (CBRS) band. The system leverages machine learning-based forecasting to predict future spectrum vacancies and optimally assign time-frequency resources to user devices, thereby minimizing interference and maximizing throughput.CBRS Framework and Spectrum Access System (FIGS. 1 , 3, 13)

[0118] Referring to FIG. 1 , the CBRS system is depicted as a tiered structure managed by a Spectrum Access System (SAS) 102. The three tiers include Incumbent Access 104, Priority Access 106, and General Authorized Access 108, all coordinated within the overall CBRS framework 110. The SAS 102 dynamically allocates spectrum resources, ensuring protection for higher-tier users and efficient sharing among all participants.

[0119] FIG. 3 provides a schematic representation of the SAS 302 managing the tiered CBRS spectrum access framework 110, with base stations 324 and other network elements 306 interacting with the SAS for spectrum assignments.

[0120] FIG. 13 further illustrates spectrum sharing and coexistence within the CBRS framework, highlighting the roles of the Incumbent, Priority Access, and General Authorized Access tiers, as well as the integration of satellite and terrestrial systems.System Architecture and Components (FIGS. 2, 4, 6, 10)

[0121] FIG. 2 shows a block diagram of the spectrum assignment system 210. The system includes a base station 222 in communication with multiple user devices 224. The base station 222 is operably coupled to a spectrum forecasting system 212, which comprises a preprocessing engine 205 and a forecasting model 202. The preprocessing engine 205 receives uplink signal data from the base station 222, processes the data into frames of past-looking spectrograms, and provides these to the forecasting model202. The forecasting model 202, trained on labeled spectrogram data, infers future spectrum vacancies and generates a recommended path for resource assignment.

[0122] FIG. 4 illustrates a system diagram of a base station 406, which may be implemented using commodity hardware or an open-source stack 402. The base station 406 interfaces with user devices 404 and central network components, such as a gNB + core server, Open5GS, and srsRAN, enabling flexible deployment and integration with existing wireless infrastructure.

[0123] FIG. 6 presents a block diagram of the prediction application deployment architecture. The system includes an SDR / RF receiver, preprocess module, inference module, control module, and IO interface, all of which may be deployed on various infrastructures such as embedded devices, workstations, or cloud servers. Streamed signal data is ingested, preprocessed, and analyzed in real time to support dynamic spectrum assignment.

[0124] FIG. 10 depicts a system diagram of a gNB and a central server integrated with Open5GS, RIA docker, srsRAN docker, and connected UE devices. This configuration supports continual updates and real-time resource allocation based on the outputs of the forecasting system.Spectrum Forecasting and Resource Assignment (FIGS. 5, 7, 8, 9, 1 1 , 12)

[0125] FIG. 5 provides a graphical representation of predicted and recommended resource grids for spectrum sharing. The predicted grid shows anticipated spectrum occupancy, while the recommended path highlights the optimal sequence of resource blocks for assignment, minimizing the risk of collision and interference.

[0126] FIG. 7 illustrates a multi-threaded inference pipeline for spectrum forecasting and resource assignment. Multiple threads process incoming data, perform preprocessing, inference, and path recommendation, and communicate results to the scheduling system, enabling high-throughput, low-latency operation.

[0127] FIG. 8 presents a process diagram for spectrum analysis. The SDR is tuned to the spectrum of interest (step 1), producing a baseband IQ stream (step 2). The IQ stream is converted into a spectrogram (step 3), which is then analyzed by the trained model to forecast vacancies and recommend a transmission pathway (step 4). The model inference system may be implemented on low-cost SDR and embedded GPU hardware.

[0128] FIG. 9 shows a comparison between predicted grids and target grids, along with recommended paths for spectrum vacancy assignment. The system’s ability toclosely match the target grid and recommend efficient paths demonstrates the effectiveness of the forecasting and assignment algorithms.

[0129] FIG. 11 illustrates a network diagram depicting resource allocation between a base station and mobile devices using a resource grid. The base station assigns timefrequency resource blocks to user devices based on the recommended path, ensuring efficient and interference-aware communication.

[0130] FIG. 12 provides a schematic representation of spectrum vacancy forecasting and resource block assignment within a wireless communication system. The system dynamically identifies vacant regions in the resource grid and assigns them to user devices, adapting in real time to changing spectrum conditions.

[0131] Referring now to FIGURE 14, there is represented a methodology for collecting data for vacancy forecasting. Time series or spectrogram data, which can be synthetic, generated from one or more testbeds, or acquired by passively observing actual signals, are matched with labels that indicate vacancies (usually as a “0”) and occupancies (usually as a “1”) at a resource grid or per pixel level. The training data is produced from recordings of length L, and contain both spectrograms (represented as r x nfft) and the source time series data (represented as r x nfft), where nfft is the Fast Fourier Transform size selected for the data, and r is the number of rows within the spectrogram.

[0132] Advantageously, the system and methods described herein leverage channel state forecasting model is to anticipate the optimal MOS code for a given lookback period input. The channel state forecasting model is preferably trained to forecast the best possible modulation, gain settings, error correction code and other settings.

[0133] To achieve this, data is structured in a similarway asforforecasting, butthe means of generation and the labels differ. Generation occurs via procedural testbed generation and via procedural dataset synthesis. The data is generated or synthesized so that it can be demodulated and measured against ground truths during transmission or synthesis. Data is only labelled at the grid level, and the time and frequency bins will necessarily correspond to generated or synthesized resource grid elements.

[0134] The labels for the channel state forecasting model are the optimal MOS code configuration given the observed channel.

[0135] To achieve this, the following procedure is followed:

[0136] An OFDM signal is generated / synthesized with known settings and a known message. In agiven recording, differentknown settingsare applied overthecourse of time.

[0137] The signal is passed through a channel (synthetically or transmitted over the air).

[0138] After the recording is captured, it is received, and measured. The message is recovered and demodulated, and measurements are made.

[0139] After message recovery and measurements, bit error rate is calculated against the ground truth. For each resource element, a score is generated, either based only on the best bit error rate or a combination of bit error rate and measurements.

[0140] The MOS configuration of the transmitted cell(s) with the best scores will then be applied as labels to the label grid of that recording.

[0141] Furthermore, the system, processes and methods of the described embodiments are capable of being distributed in a computer program product comprising a computer readable medium that bears computer usable instructions for one or more processors. The medium may be provided in various forms, including one or more diskettes, compact disks, tapes, chips, wireline transmissions, satellite transmissions, internet transmission or downloading, magnetic and electronic storage media, digital and analog signals, and the like. The computer useable instructions may also be in various forms, including compiled and non-compiled code.

[0142] Various embodiments have been described herein by way of example only. Various modification and variations may be made to these example embodiments without departing from the spirit and scope of the invention, which is limited only by the appended claims.

Claims

CLAIMS1. A spectrum assignment system comprising: a base station configured to communicate with a plurality of user devices; a spectrum forecasting system, comprising: a spectrum receiverto monitor a radio spectrum and collect signal data in real time; a preprocessing engine operably coupled to the spectrum receiver and configured to process the signal data to generate a stream of frames of past-looking spectrograms; and a forecasting model, trained on spectrogram data with labeled vacancies, operably coupled to the preprocessing engine and configured to receive the stream of frames of past-looking spectrograms as input, infer future spectrum vacancies, and generate a recommended path through the inferred future vacancies for spectrum assignment; a scheduler for receiving the recommended path and assigning the path to a base station, wherein the base station is further configured to transmit uplink signal data to the preprocessing engine and the forecasting model, and to receive the recommended path for assignment to the user devices.

2. The spectrum forecasting system of claim 1, wherein the live signal data includes the last 20 milliseconds, and the path includes the next 2 milliseconds.

3. The spectrum forecasting system of claim 1, wherein the live signal data includes the last 200 milliseconds, and the path includes the next 20 milliseconds.

4. The spectrum forecasting system of claim 1, wherein the base station comprises at least one of: a router, general node base station, and a standalone access point.

5. The spectrum forecasting system of claim 1, wherein the forecasting model comprises a machine learning model.

6. The spectrum forecasting system of claim 1, wherein the machine learning model comprises at least one of: time series forecasting, spectrogram forecasting, transformers and pattern recognition models.RECTIFIED SHEET (RULE 91.1)7. The spectrum forecasting system of claim 1 , wherein the forecasting model is deployed to an edge device, FPGA or cloud server.

8. The spectrum forecasting system of claim 1, wherein the preprocessing engine and forecasting model are integrated within the base station or are deployed as a standalone inference container communicatively coupled to the base station.

9. The spectrum forecasting system of claim 1, wherein the recommended path comprises a set of time-frequency resource blocks with a probability of collision below a predetermined threshold.

10. A method of assigning spectrum vacancies, comprising: receiving uplink signal data from a base station at a preprocessing engine; processing the signal data into a stream of frames of past-looking spectrograms; inputting the stream of frames of past-looking spectrograms into a forecasting model trained on spectrum data with labeled vacancies; inferring, using the forecasting model, future spectrum vacancies; charting, using the forecasting model, a recommended path comprising a set of timefrequency resource blocks through the inferred future vacancies for assignment; communicating the recommended path to the base station; and assigning the recommended path to the user devices.

11. The spectrum forecasting method of claim 10, wherein the live signal data includes the last 20 milliseconds, and the path includes the next 2 milliseconds.

12. The spectrum forecasting method of claim 10, wherein the live signal data includes the last 200 milliseconds, and the path includes the next 20 milliseconds.

13. The spectrum forecasting method of claims 10, wherein said method further includes the step of anticipating optimal transmission parameters.

14. The spectrum forecasting method of claim 13, wherein said optimal transmission parameters include MCS.

15. A non-transitory computer-readable medium storing instructions that, when executed by one or more processors, cause the system to perform the method of any one of claims 10 to 12.RECTIFIED SHEET (RULE 91.1)16. A user device or standalone access point configured to perform the method of any one of claims 10 to 12.

17. A spectrum assignment system comprising a plurality of spectrum receivers and forecasting models distributed across multiple geographic locations, each configured to monitor a portion of the radio spectrum and provide recommended paths to a central base station.RECTIFIED SHEET (RULE 91.1)

Images