System and method for slice level resource management in a network
A decentralized peer-to-peer federated learning framework among base stations addresses the challenges of dynamic UE mobility by enabling direct communication and model exchange for efficient, real-time, and secure slice-level resource management.
Patent Information
- Application Number
- PCT/IN2026/050784
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2025-05-07
- Filing Date
- 2026-05-07
- Publication Date
- 2026-11-12
AI Technical Summary
Base stations struggle to accurately forecast traffic demand and allocate slice resources due to dynamic UE mobility and frequent handovers, leading to inefficient resource management, computational intensity, and privacy concerns in centralized systems.
A decentralized peer-to-peer federated learning framework among base stations for slice-level resource management, enabling direct communication and model parameter exchange without a central aggregator, using an AI engine for real-time traffic forecasting and capacity planning.
Enables accurate, real-time, and secure slice-level resource management by distributing traffic data and computational tasks, ensuring resilience and reducing computational overhead while maintaining privacy.
Smart Images

Figure IN2026050784_12112026_PF_FP_ABST
Abstract
Description
SYSTEM AND METHOD FOR SLICE LEVEL RESOURCE MANAGEMENT IN A NETWORKFIELD OF INVENTION
[0001] The embodiments of the present disclosure generally relate to the field of peer-to-peer (P2P) learning among base stations. More particularly, the present disclosure relates to a system and a method of P2P learning among the base stations for slice level resource management in a network.BACKGROUND OF THE INVENTION
[0002] The following description of the related art is intended to provide background information pertaining to the field of the disclosure. This section may include certain aspects of the art that may be related to various features of the present disclosure. However, it should be appreciated that this section is used only to enhance the understanding of the reader with respect to the present disclosure, and not as admission of the prior art.
[0003] User equipments (UEs) move along random trajectories across a terrain, leading to frequent handovers among multiple gNBs (next generation base stations) across a network. For proper operational management of the network slice each gNB needs information regarding the UE traffic in its vicinity for calculating the slices at various future time instants. However, due to dynamic mobility of the UEs and repeated handovers, each gNB struggles to forecast traffic demand and also struggles to allocate slice resources accurately. Further, it becomes extremely difficult for each gNB in the network slice operation phase to perform resource capacity planning and modification without coordinated traffic information.
[0004] Further, existing approaches rely on centralized aggregation or heuristic -based resource planning, which are computationally intensive, time-consuming, and often inaccurate under real-time mobility conditions. Also, centralized systems introduce a single point of failure and may raise privacy concerns.
[0005] There is therefore, a need for a system and method that can mitigate at least the above mentioned issues.OBJECTS OF THE INVENTION
[0006] Some of the objects of the present disclosure, which at least one embodiment herein satisfies are listed herein below.
[0007] It is an object of the present disclosure to provide a system and a method for slice level resource management in a network by identifying one or more base station nodes and aggregating user equipment (UE) traffic measurements of each slice associated with each of the identified base station nodes for a first predetermined period, thereby enabling accurate traffic forecasting at the slice level.
[0008] It is an object of the present disclosure to provide a system that subsequently receives and exchanges information among the identified base station nodes, where the identified base station nodes are configured to communicate directly with each other in a peer-to-peer manner to generate and share model parameters without reliance on a central aggregator .
[0009] It is an object of the present disclosure to provide a system that analyzes, via an artificial intelligence engine the exchanged information and in response to a determination that the analyzed information is within a threshold condition, determines the aggregated UE traffic measurements of each slice, thereby supporting real-time slice adjustment.
[0010] It is an object of the present disclosure to provide a system that predicts, via the artificial intelligence engine, the aggregated UE traffic measurements of each slice associated with the identified base station nodes for one or more subsequent periods, based on the aggregated UE traffic measurements of each slice, thereby enabling proactive slice capacity planning.
[0011] It is an object of the present disclosure to provide a system and method that ensures resilience of the federated learning (FL) process, such that even if one base station node fails, the peer-to-peer collaboration continues without interruption.
[0012] It is an object of the present disclosure to enhance privacy by eliminating the need for centralized data collection, thereby ensuring that raw UE traffic data remains distributed across base station nodes.
[0013] It is an object of the present disclosure to reduce computational overhead by avoiding complex heuristic-based resource planning by leveraging distributed Al-driven prediction for efficient slice management.
[0014] It is an object of the present disclosure to provide a flexible participation mechanism; wherein base station nodes can dynamically join or leave the FL process based on synchronization with aggregation and exchange time windows.
[0015] It is an object of the present disclosure to provide a random gossip-based peer-to-peer parameter exchange algorithm that ensures distributed convergence of machine learning models across participating gNBs.SUMMARY
[0016] This section is provided to introduce certain objects and aspects of the present disclosure in a simplified form that are further described below in the detailed description. This summary is not intended to identify the key features or the scope of the claimed subject matter.
[0017] In an aspect, the present disclosure relates to a system for predicting user equipment (UE) traffic measurements of each slice. The system may include a processor communicatively coupled with at least one base station node, and a memory operatively coupled with the processor. The memory stores instructions which, when executed by the processor, enable the processor to coordinate federated learning operations across base station nodes. In an embodiment, the base station node may identify one or more neighbouring base station nodes capable of participating in federated learning. The base station node may aggregate UE traffic measurements of each slice associated with the base station node for a predetermined period. The base station node may transmit information associated with parameters of a model in the base station node, to the identified one or more base station nodes. Further, the base station node receives model information associated with the identified one or more base station nodes, wherein the identified one or more base station nodes are configured to communicate with each other to generate the information. The base station node further aggregates the received model information from the identified one or more base station nodes with the information of the base station node. The model information exchange is repeated until the model of the base station node converges. Finally, the base station node predicts one or more UE traffic measurements of each slice associated with the base station node for one or more subsequent periods. The model performance may be monitored by the base station to further retune the model.
[0018] In an embodiment, the system may further analyze the information using an artificial intelligence engine, and in response to a determination that the analyzed information satisfies a threshold, determines the aggregated UE traffic measurements of each slice for a second predetermined period. The system then predicts, via the artificial intelligence engine, the aggregated UE traffic measurements of each slice associated with the identified base station nodes for one or more subsequent periods based on the aggregated UE traffic measurements of each slice associated with the first and second predetermined periods.
[0019] In an embodiment, the system may be configured to transmit a request to each of the one or more base station nodes for registering with the system. Upon confirmation of the request from the base station nodes, the system registers the one or more base station nodes to enable participation in the prediction process.
[0020] In an embodiment, each of the identified one or more base station nodes may comprise a machine learning model that processes UE traffic measurements of each slice associated with the respective base station node for predicting one or more UE traffic measurements of each slice associated with the one or more base station nodes for one or more subsequent period..
[0021] In an embodiment, each of the identified base station nodes may be configured to transmit the generated model information comprising model parameters among each other, and the system may be configured to determine one or more metrics based on the aggregated parameters exchanged between the nodes.
[0022] In an embodiment, to identify the one or more base station nodes, the system may be configured to determine whether each of the base station nodes allows processing of UE traffic measurements of each slice within an aggregation time period and whether each of the base station nodes is capable of transmitting parameters to K base stations within an exchange time period.
[0023] In an embodiment, the one or more base station nodes may dynamically join or leave the federated learning process based on synchronization with the aggregation time period and the exchange time period.
[0024] In an embodiment, monitoring the model performance includes determining whether one or more performance metrics of the model are below a threshold, and in response to a positive determination, retuning to aggregate the one or more UE traffic measurements of each slice for further local data collection and retraining of the model.
[0025] In an aspect, the present disclosure relates to a method for predicting UE traffic measurements of each slice. The method may be performed by a base station node. The method may include, identifying at least one federated learning capable neighbouring base station nodes and aggregating UE traffic measurements of each slice associated with the identified base station nodes for a predetermined period. The method includes transmitting information associated with parameters of a model in the base station node, to the identified one or more base station nodes. The method further includes receiving model information associated with the identified base station nodes from the identified one or more base station nodes, and analyzing the model information, wherein the nodes are configured to communicate with eachother to generate the model information. In an embodiment, the model information is exchanged until the model of the base station node converges. The method may further includes predicting one or more UE traffic measurements of each slice associated with the base station node for one or more subsequent periods. Finally, the method includes monitoring the model performance and retuning the model information.
[0026] In an embodiment, the method may include transmitting a request to each of the one or more base station nodes for predicting UE traffic measurements of each slice, and upon confirmation of the request from the base station nodes, registering the one or more base station nodes.
[0027] In an embodiment, each of the one or more base station nodes may include a machine learning model that processes UE traffic measurements of each slice associated with the base station node for predicting one or more UE traffic measurements of each slice associated with the one or more base station nodes for one or more subsequent period.
[0028] In an embodiment, the method may include determining model information based on the aggregated parameters, wherein the model information associated with the identified one or more base station nodes are communicated among each of the identified base station nodes.
[0029] In an embodiment, the method may include determining whether each of the base station nodes allows or is capable of processing of UE traffic measurements of each slice within an aggregation time period and is capable of transmitting parameters to K base station nodes within an exchange time period, for identifying the one or more base station nodes.
[0030] In an embodiment, the method may include dynamically allowing one or more base station nodes to join or leave the federated learning process depending on synchronization with the aggregation time period and the exchange time period.BRIEF DESCRIPTION OF DRAWINGS
[0031] The accompanying drawings, which are incorporated herein, and constitute a part of this disclosure, illustrate exemplary embodiments of the disclosed methods and systems which like reference numerals refer to the same parts throughout the different drawings. Components in the drawings are not necessarily to scale; emphasis instead being placed upon clearly illustrating the principles of the present disclosure. Some drawings may indicate the components using block diagrams and may not represent the internal circuitry of each component. It will be appreciated by those skilled in the art that disclosure of such drawingsincludes the disclosure of electrical components, electronic components, or circuitry commonly used to implement such components.
[0032] FIG. 1 illustrates an example architecture (100) of a proposed system (102), in accordance with an embodiment of the present disclosure.
[0033] FIG. 2 illustrates an example block diagram (200) of the proposed system (102), in accordance with an embodiment of the present disclosure.
[0034] FIG. 3 illustrates an example flow diagram (300) representing peer to peer federated learning among base stations as enabled by the proposed system (102), in accordance with an embodiment of the present disclosure.
[0035] FIG. 4 illustrates an example schematic diagram (400) of a random peer to peer federated learning initiation implemented by the proposed system (102), in accordance with an embodiment of the present disclosure.
[0036] FIG. 5 illustrates an example schematic diagram (500) of a random peer to peer federated learning participation list implemented by the proposed system (102), in accordance with an embodiment of the present disclosure.
[0037] FIG. 6 illustrates an example schematic diagram (600) of a random peer to peer federated learning parameter exchange implemented by the proposed system (102), in accordance with an embodiment of the present disclosure.
[0038] FIG. 7 illustrates an example computer system (700) in which or with which the proposed system (102) may be implemented, in accordance with an embodiment of the present disclosure.
[0039] FIG. 8 illustrates an example high-level flow diagram of a proposed method (800) implemented by the proposed system (102), in accordance with an embodiment of the present disclosure.
[0040] The foregoing shall be more apparent from the following more detailed description of the disclosure.DETAILED DESCRIPTION OF THE INVENTION
[0041] In the following description, for explanation, various specific details are outlined in order to provide a thorough understanding of embodiments of the present disclosure. It will be apparent, however, that embodiments of the present disclosure may be practiced without these specific details. Several features described hereafter can each be used independently of one another or with any combination of other features. An individual feature may not address all of the problems discussed above or might address only some of theproblems discussed above. Some of the problems discussed above might not be fully addressed by any of the features described herein.
[0042] The ensuing description provides exemplary embodiments only and is not intended to limit the scope, applicability, or configuration of the disclosure. Rather, the ensuing description of the exemplary embodiments will provide those skilled in the art with an enabling description for implementing an exemplary embodiment. It should be understood that various changes may be made in the function and arrangement of elements without departing from the spirit and scope of the disclosure as set forth.
[0043] Specific details are given in the following description to provide a thorough understanding of the embodiments. However, it will be understood by one of ordinary skill in the art that the embodiments may be practiced without these specific details. For example, circuits, systems, networks, processes, and other components may be shown as components in block diagram form in order not to obscure the embodiments in unnecessary detail. In other instances, well-known circuits, processes, algorithms, structures, and techniques may be shown without unnecessary detail to avoid obscuring the embodiments.
[0044] Also, it is noted that individual embodiments may be described as a process that is depicted as a flowchart, a flow diagram, a data flow diagram, a structure diagram, or a block diagram. Although a flowchart may describe the operations as a sequential process, many of the operations can be performed in parallel or concurrently. In addition, the order of the operations may be re-arranged. A process is terminated when its operations are completed but could have additional steps not included in a figure. A process may correspond to a method, a function, a procedure, a subroutine, a subprogram, etc. When a process corresponds to a function, its termination can correspond to a return of the function to the calling function or the main function.
[0045] The word “exemplary” and / or “demonstrative” is used herein to mean serving as an example, instance, or illustration. For the avoidance of doubt, the subject matter disclosed herein is not limited by such examples. In addition, any aspect or design described herein as “exemplary” and / or “demonstrative” is not necessarily to be construed as preferred or advantageous over other aspects or designs, nor is it meant to preclude equivalent exemplary structures and techniques known to those of ordinary skill in the art. Furthermore, to the extent that the terms “includes,” “has,” “contains,” and other similar words are used in either the detailed description or the claims, such terms are intended to be inclusive like the term “comprising” as an open transition word without precluding any additional or other elements.
[0046] Reference throughout this specification to “one embodiment” or “an embodiment” or “an instance” or “one instance” means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment of the present disclosure. Thus, the appearances of the phrases “in one embodiment” or “in an embodiment” in various places throughout this specification are not necessarily all referring to the same embodiment. Furthermore, the particular features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.
[0047] The terminology used herein is to describe particular embodiments only and is not intended to be limiting the disclosure. As used herein, the singular forms “a”, “an”, and “the” are intended to include the plural forms as well, unless the context indicates otherwise. It will be further understood that the terms “comprises” and / or “comprising,” when used in this specification, specify the presence of stated features, integers, steps, operations, elements, and / or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof. As used herein, the term “and / or” includes any combinations of one or more of the associated listed items.
[0048] In the existing state of the art, base stations face challenges in managing slice resources due to the random trajectories of UEs and the resulting multiple handovers. Each gNB struggles to obtain accurate visibility into traffic demand beyond its immediate coverage area, making it difficult to plan slice resources for future time instants. Conventional approaches rely on centralized aggregation or heuristic-based resource planning, which are computationally intensive, time-consuming, and often inaccurate under real-time mobility conditions. Furthermore, centralized systems introduce a single point of failure and raise privacy concerns, as raw traffic data must be collected and processed at a central node.
[0049] In an aspect, the present disclosure relates to a peer-to-peer federated learning among base station nodes, such as gNBs in 5G networks and aNBs in future 6G deployments, for the purpose of slice-level resource management. In an embodiment, the present enables base stations to collaborate directly with one another, exchange model parameters, and predict user equipment (UE) traffic measurements across slices in a decentralized manner. The present disclosure is particularly relevant in scenarios where dynamic UE mobility and frequent handovers create significant uncertainty in traffic forecasting and slice capacity planning.
[0050] In an aspect, the present disclosure relates to a system and method for predicting user equipment (UE) traffic measurements of each slice. The system may include a processor communicatively coupled with at least one base station node, and a memory operativelycoupled with the processor. The memory stores instructions which, when executed by the processor, enable the processor to coordinate federated learning operations across base station nodes. In an embodiment, the base station node may identify one or more neighbouring base station nodes capable of participating in federated learning. The base station node may aggregate UE traffic measurements of each slice associated with the base station node for a predetermined period. The base station node may transmit information associated with parameters of a model in the base station node, to the identified one or more base station nodes. Further, the base station node receives model information associated with the identified one or more base station nodes, wherein the identified one or more base station nodes are configured to communicate with each other to generate the information. The base station node further aggregates the received model information from the identified one or more base station nodes with the information of the base station node. The model information exchange is repeated until the model of the base station node converges. Finally, the base station node predicts one or more UE traffic measurements of each slice associated with the base station node for one or more subsequent periods. The model performance may be monitored by the base station to further retune the model.
[0051] In an embodiment, the present disclosure mitigates the challenges of conventional approaches by introducing a decentralized, peer-to-peer federated learning framework. Base station nodes may identify peers, aggregate UE traffic measurements of each slice, and exchange model parameters directly using inter-node interfaces such as the Xn interface. An artificial intelligence engine may analyze the aggregated information, determines traffic measurements across multiple time windows, and predicts future slice-level demand. This approach eliminates reliance on a central aggregator, enhances resilience by ensuring that federated learning continues even if one node fails, and improves privacy by keeping raw traffic data distributed across nodes. The decentralized approach of the present disclosure enhances the system's resilience by distributing tasks across multiple peers, avoiding the single point of failure typical of centralized systems. Further, by exchanging the model parameters rather than raw data reduces computational overhead and mitigates privacy concerns among heterogeneous base stations from different vendors. This, the present disclosure thereby enables accurate, real-time, secure and scalable slice-level resource management in next-generation networks.
[0052] The various embodiments throughout the disclosure will be explained in more detail with reference to FIGs. 1-7.
[0053] FIG. 1 illustrates an example network architecture (100) of a proposed system (102), in accordance with an embodiment of the present disclosure.
[0054] As illustrated in FIG. 1, in an embodiment, one or more base station nodes (gNBs) (NG-RAN nodei 106, NG-RAN node2 108) may be communicatively to each other where one gNB may initiate a federated learning (FL) process by requesting a neighbor gNB if the neighbor gNB is interested in participating in the FL process for a particular task using an Xn interface. The request may also include the aggregation time period (the time window when all the gNBs update their models with the received model parameters). An exchange time period (the time window when all the gNBs share their new model parameters with each other) may also be included in the request. ‘K’ (number of gNBs to which a gNB will sends its model parameters during model Information exchange) may be decided by the initiated gNB and the list of participating gNBs may also send to all the participating gNBs so that all the gNBs may know their peers in the process.
[0055] In an embodiment, the one or more base station nodes dynamically join or leave the federated learning process based on synchronization with the aggregation time period and the exchange time period.
[0056] In an aspect, the present disclosure relates to a direct peer-to-peer communication between the one or more base station nodes for collaborative learning and slice-level resource management, without reliance on a centralized aggregator. FIG. 1 serves as a high-level schematic of the system (102) capability to support distributed intelligence across multiple nodes.
[0057] In an alternative embodiment, the network architecture (100) of the system (102) is not limited to NG-RAN nodes or gNBs. The network architecture (100) in accordance with the present disclosure, may be extended to future-generation access nodes (aNBs) in 6G or beyond, where peer-to-peer FL process may be initiated across heterogeneous base station types. Further, the communication interface may similarly evolve beyond the Xn interface to other standardized or similar inter-node interfaces, while still supporting the exchange of model parameters and aggregation windows.
[0058] In an exemplary embodiment, the network architecture (100) may involve multiple gNBs or aNBs participating simultaneously. This arrangement may also be adapted for different deployment scenarios, such as dense urban networks, rural macro cells, or private enterprise networks, where peer-to-peer collaboration enhances resilience and scalability.
[0059] Although FIG. 1 shows exemplary components of the network architecture (100), in other embodiments, the network architecture (100) may include fewer components,different components, differently arranged components, or additional functional components than depicted in FIG. 1. Additionally, or alternatively, one or more components of the network architecture (100) may perform functions described as being performed by one or more other components of the network architecture (100).
[0060] FIG. 2 illustrates an example block diagram (200) of the proposed system (102), in accordance with an embodiment of the present disclosure.
[0061] Referring to FIG. 2, the system (102) may include one or more processor(s) (202). The one or more processor(s) (202) may be implemented as one or more microprocessors, microcomputers, microcontrollers, digital signal processors, central processing units, logic circuitries, and / or any devices that process data based on operational instructions. Among other capabilities, the one or more processor(s) (202) may be configured to fetch and execute computer-readable instructions stored in a memory (204) of the system (102). The memory (204) may be configured to store one or more computer-readable instructions or routines in a non-transitory computer readable storage medium, which may be fetched and executed to create or share data packets over a network service. The memory (204) may comprise any non-transitory storage device including, for example, volatile memory such as random-access memory (RAM), or non-volatile memory such as erasable programmable read only memory (EPROM), flash memory, and the like.
[0062] In an embodiment, the system (102) may include an interface(s) (206). The interface(s) (206) may comprise a variety of interfaces, for example, interfaces for data input and output devices (I / O), storage devices, and the like. The interface(s) (206) may facilitate communication through the system (102). The interface(s) (206) may also provide a communication pathway for one or more components of the system (102). Examples of such components include, but are not limited to, processing engine(s) (208), a database (210), and a data parameter engine (212).
[0063] The processing engine(s) (208) may be implemented as a combination of hardware and programming (for example, programmable instructions) to implement one or more functionalities of the processing engine(s) (208). In examples described herein, such combinations of hardware and programming may be implemented in several different ways. For example, the programming for the processing engine(s) (208) may be processor-executable instructions stored on a non-transitory machine-readable storage medium and the hardware for the processing engine(s) (208) may comprise a processing resource (for example, one or more processors), to execute such instructions. In the present examples, the machine-readable storage medium may store instructions that, when executed by the processing resource,implement the processing engine(s) (208). In such examples, the system (102) may comprise the machine-readable storage medium storing the instructions and the processing resource to execute the instructions, or the machine-readable storage medium may be separate but accessible to the system (102) and the processing resource. In other examples, the processing engine(s) (208) may be implemented by electronic circuitry.
[0064] In an embodiment, the system ( 102) may receive a request via the data parameter engine (212). The request may be received from one or more base station nodes. The system (102) may record the request in the database (210). The system (102) may identify one or more base station nodes. The system (102) may transmit the request to each of the one or more base station nodes and upon confirmation of the request from the one or more base station nodes, register the one or more base station nodes.
[0065] In an embodiment, to identify the one or more base station nodes, the system (102) may be configured to determine if each of the one or more base station nodes allow processing of the one or more UE traffic measurements of each slice within an aggregation time period and are capable of transmitting the one or more parameters to K base stations from the one or more base station nodes within an exchange time period.
[0066] In an embodiment, each of the one or more base station nodes may include a machine learning model that processes the one or more UE traffic measurements of each slice associated with the one or more base station nodes for the second predetermined period to generate one or more parameters. Each of the identified one or more base station nodes may be configured to transmit the one or more parameters among each other and the system (102) may be configured to determine the one or more metrics based on the aggregated one or more parameters.
[0067] In an embodiment, the system (102) may aggregate one or more user equipment (UE) traffic measurements of each slice associated with each of the identified one or more base station nodes for a first predetermined period.
[0068] In an embodiment, the system (102) may subsequently receive, information associated with the identified one or more base station nodes, where the identified one or more base station nodes may be configured to communicate with each other to generate the information.
[0069] In an embodiment, the system (102) may analyze, via an artificial intelligence engine (214), the information and in response to a determination that the analyzed information is within a threshold, may determine the aggregated one or more UE traffic measurements of each slice associated for a second predetermined period.
[0070] In an alternative embodiment, the system (102) may instantaneously determine the aggregated one or more UE traffic measurements of each slice associated for a second predetermined period in response to the determination that the analyzed information is within the threshold. The term instantaneously may refer to the capability of the system (102) to perform the determination or computation without perceptible delay relative to the operational time window of the network slice. The expression does not imply absolute immediacy in a physical sense, but rather denotes that the determination is carried out within a time frame sufficiently short to enable real-time or near real-time slice-level resource management. For example, the artificial intelligence engine (24) may analyze the information and generate the aggregated traffic measurements within milliseconds or within the aggregation time period defined for the federated learning process, such that the output is available for immediate use in subsequent prediction steps. Thus, the term instantaneously is intended to encompass determinations performed with minimal latency that are operationally effective for dynamic UE traffic forecasting and slice capacity planning.
[0071] In an embodiment, the system (102) may predict, via the artificial intelligence engine (214), the aggregated one or more UE traffic measurements of each slice associated with the identified one or more base station nodes for one or more subsequent periods based on the aggregated one or more UE traffic measurements of each slice associated for the first time period and the second time period.
[0072] In an exemplary embodiment, UE traffic measurements may include various performance indicators such as Downlink (DL)ZUplink (UL) throughput, DL / UL Packet Data Convergence Protocol (PDCP) data volume, bearer level data volume, radio resource utilization metrics, packet delay and jitter statistics, handover counts, and Quality of Service (QoS) class identifiers, and likewise. The PDCP data volume refers to the total amount of user plane and control plane traffic processed at the PDCP layer, including ciphered / deciphered packets, header-compressed packets, and re-ordered service data units (SDUs). This measurement of the one or more performance indicators provides a reliable indication of actual end-user / UE traffic load at the gNB and further for slice-level resource management. By incorporating these measurements, the processor (202) can capture both throughput-oriented and protocol-layer traffic characteristics, enabling the artificial intelligence engine (214) to generate accurate predictions of UE traffic patterns across multiple time windows.
[0073] FIG. 3 illustrates an example flow diagram (300) representing peer to peer federated learning among base stations as enabled by the proposed system (102), in accordance with an embodiment of the present disclosure.
[0074] As illustrated in FIG. 3, in an embodiment, at step 306, an initiating base station node (gNB) 302 may send a solicitation request to one or more base station nodes 304 for participating in a FL process. At step 308, the all participating base station nodes may initiate local data collection based on the one or more user equipment (UE) traffic measurements of each slice associated with their own area. At step 310, all the participating base station nodes may implement a machine learning model for analyzing and processing the information associated with the one or more user equipment (UE) traffic measurements of each slice. At step 312, the participating base station nodes may select K numbers of participating gNBs. At step 314, all the participating base station nodes may send their model parameters to their selected K numbers of participating gNBs. At step 316, all the participating base station nodes may receive model parameters from the other participating base station nodes. At step 318, all the participating base station nodes may perform model aggregation based on the model parameters received from the other participating base station nodes. At step 320, all the participating base station nodes may continue with the model aggregation process until the machine learning model converges. At step 322, all the participating base station nodes whose model has converged may perform inference associated with the model aggregation process. At step 324, the gNBs whose model has converged may determine if the performance metrics is below a threshold and in response to a positive determination, may return to step 308 for further local data collection and retraining.
[0075] In an alternative embodiment, at step 306, the system (102) may solicit participation from one or more base station nodes in the FL process. The solicitation may allow for dynamic participation, wherein gNBs or future -generation aNBs can join or leave the FL process depending on their synchronization with the aggregation and exchange time windows. A base station node that successfully aligns its local training with the aggregation time period and is capable of transmitting parameters within the exchange time period may join the ongoing FL cycle, while a base station node that cannot synchronize may temporarily leave without disrupting overall convergence. This dynamic participation mechanism ensures resilience, scalability, and adaptability of the FL framework across heterogeneous deployment scenarios.
[0076] In an embodiment, at step 320, all the participating base station nodes may continue with the model aggregation process until the machine learning model converges. Once the machine learning model convergence is achieved, the aggregated model may be utilized for prediction of the UE traffic measurements of each slice for one or more subsequent periods.
[0077] In an additional embodiment, the prediction of the UE traffic measurements of each slice may be performed in either a unistep or a multistep manner.
[0078] In an embodiment, the unistep prediction refers to the process where the artificial intelligence engine (214) predicts the traffic measurements for a single subsequent time period based on the aggregated measurements of the immediately preceding time windows. This approach provides short-term forecasts that are highly responsive to dynamic changes in UE mobility and slice demand.
[0079] In another embodiment, the multistep prediction refers to the process where the artificial intelligence engine (214) predicts traffic measurements for multiple subsequent time periods in advance, based on the aggregated measurements of prior time windows. This approach enables proactive slice capacity planning over longer horizons, thereby allowing the network to anticipate sustained traffic patterns and allocate resources accordingly.
[0080] In an exemplary embodiment, the selection between the unistep and the multistep prediction may be determined dynamically by the system (102) depending on at least one of accuracy threshold, an aggregation time period, or an operational requirement of the slice. Thus, the present disclosure provides flexibility in forecasting UE traffic measurements across slices, ensuring both immediate responsiveness and long-term scalability.
[0081] In an embodiment, once the model convergence is achieved, the aggregated model may be utilized for prediction of UE traffic measurements of each slice. The prediction may be performed in either the unistep or the multistep manner as described herein. In particular, this prediction framework enables each gNB, or future-generation aNB, in the network slice operation phase to perform resource capacity planning and modification of slice resources for the next time instant (Tn+1) without reliance on complex heuristics. Thus, the present disclosure provides the prediction framework that is both efficient and accurate, thereby ensuring that slice-level resource allocation is responsive to real-time traffic demand while remaining computationally lightweight.
[0082] FIG. 4 illustrates an example schematic diagram (400) of a random peer to peer federated learning initiation implemented by the proposed system (102), in accordance with an embodiment of the present disclosure.
[0083] As illustrated in FIG. 4, in an embodiment, at step 406, a NG-RAN nodei (402) initiates the procedure by sending a peer to peer (P2P) FL PARTICIPATION REQUEST message to a NG-RAN node2 (404) to invite NG-RAN node2 (404) for participating in the FL process. Upon receipt, at step 408, the NG-RAN node2 (404) may determines whether it can synchronize with the aggregation time period and the exchange time period provided in the request. Further, the NG-RAN node2 (404) determines whether it is capable for participating for the particular Task ID mentioned in the request. If the NG-RAN node2 (404) cansynchronize with requesting gNB and is capable for the particular Task, the NG-RAN node2 (404) may participate in the peer to peer FL process.Table 1 shows the P2P FL participation request details.lE / Group Name Presence Range IE type and Semantics Criticality Assigned reference description Criticality Message Type M - 9.2.3.1 - YES reject NG-RAN node 1 M Global NG- YES reject Cell ID RAN NodeID9.2.2.3NG-RAN node2 M Global NG- YES Reject Cell ID RAN NodeID9.2.2.3Model M INTEGER TimeAggregation windowTime Window when all thegNBsupdate theirmodels withreceivedmodelparameters.Unit=secondsModel M INTEGER TimeParameters windowExchanging Time when all theWindow gNBs sharetheir modelparameterswith eachotherTask ID M An IDmapping toa particulartask (e.g.Task ID: 1map to UEtrafficforecasting)Table 1
[0084] Tables 2 shows the P2P FL participation response details.lE / Group Name Presence Range IE type and Semantics Criticality Assigned reference description Criticality Message Type M - 9.2.3.1 - YES reject NG-RAN nodei M Global NG- YES reject ID RAN NodeID9.2.2.3NG-RAN node2 M Global NG- YES reject ID RAN NodeID9.2.2.3Task ID M An IDmapping toa particulartask (e.g.Task ID:1map to UEtrafficforecasting)Response M 0..1 Indicates YES ignore whether theNode2 willparticipatein the P2PFL process(0= Will notparticipate, 1=Willparticipate)Table 2
[0085] FIG. 5 illustrates an example schematic diagram (500) of a random peer to peer federated learning participation list implemented by the proposed system (102), in accordance with an embodiment of the present disclosure.
[0086] In an embodiment, after getting response from the multiple gNBs, the nodei (502) will send the list of all participating gNBs ID to all participating gNBs, for example node2 (504). At step 506, each of the one or more base station nodes gNBs (502, 504) may know all other one or more base station nodes participating in the FL process by sending a P2P participation list.
[0087] Tables 3 shows the P2P FL participation list details.lE / Group Name Presence Range IE type and Semantics Criticality Assigned reference description Criticality Message Type M - 9.2.3.1 - YES reject NG-RAN nodel ID M - Global NG- - YES reject RAN Node ID9.2.2.3NG-RAN node2 ID M - Global NG- - YES reject RAN Node ID9.2.2.3Task ID M An ID mapping toa particular task(e.g. Task ID:1map to UE trafficforecasting)> Agreed NG- 1 - YES reject RAN Node List» Agreed NG- o.. <RAN Node Item noofAgreedNG-RANnodes >»> Agreed NG- M - Global NG- - - RAN node ID RAN Node ID9.2.2.3Number of gNBs in INTEGER number of gNBsa window to which a gNBwill sends itsmodel parametersduring one modelInformationexchange window>Initial weights - 1 - - - - List»Weights Item - 0..<noo - - - - fparameters>»>Weights - - - - - - Table 3
[0088] FIG. 6 illustrates an example schematic diagram (600) of a random peer to peer 5 federated learning parameter exchange implemented by the proposed system (102), in accordance with an embodiment of the present disclosure.
[0089] As illustrated in FIG. 6, in an embodiment, at step 602, aNG-RAN nodei (604) initiates the procedure by sending a peer to peer (P2P) FL parameter exchange to a NG-RAN node2 (604) to invite NG-RAN node2 (606) for participating in the FL process.
[0090] In an embodiment, each of the one or more base station nodes gNBs (604, 606) may prepare for the local machine learning model training by collecting local data. Further, the each of the one or more base station nodes gNBs (604, 606) may start the training of the machine learning models using the locally collected data. After local machine learning model training, each of the one or more base station nodes gNBs (604, 606) may initiate the random gossip peer to peer federated learning process based on the following algorithm. At step 602, each of the one or more base station nodes gNBs (604, 606) may exchange FL parameters to each of the P2P participating one or more base station nodes gNBs (604, 606). Each of the P2P participating gNB (for example ‘ i’th gNB) may perform the following algorithm:1. Participants= [gNBl, gNB2, gNB3, gNB4 ] (List of participants except 'i’th gNB)2. List=Participants.3. N=length_of_(List)4. Probability of choosing every gNBs in List is (1 / N)5. Choose K gNBs from the List (send model updates to these K gNBs). Updating own model by received parameters from other gNBs6. Remove the chosen gNBs from the list and update the list.7. a. If the current length_of_(List) is less than K, send next model update to the remaining gNBs in the List and load Participants into List and continue from 3.7.b. else continue from 3.8. a. Continue from 3 to 7 until model converges.8.b. If model converges in this ‘i’th gNB. This will inform other gNBs that its model is converged, so that other gNBs can remove this ‘i’th gNB from their participants list, so they can stop sending their model parameters to the ‘i’th gNB. But the ‘i’th gNB continues sending its model parameters to all other gNBs in its participants list.
[0091] In an embodiment, the foregoing algorithm illustrates a decentralized random gossip peer-to-peer federated learning framework wherein each participating gNB iteratively exchanges model parameters with a dynamically selected subset of peers until convergence is achieved. The process ensures that every gNB maintains awareness of other participants, probabilistically selects peers for parameter exchange, and updates its local model based on received information. As convergence occurs, gNBs notify peers to optimize communicationoverhead by ceasing redundant parameter transmissions, while still contributing their converged parameters to the network. This iterative and distributed approach enables resilient model training without reliance on centralized aggregation, thereby ensuring scalability, reducing computational complexity, and supporting efficient slice-level traffic prediction in real-time network operation.
[0092] Tables 4 shows the P2P FL parameter exchange details.lE / Group Name Presence Range IE type and Semantics Criticali Assigned reference description ty Criticalit y Message Type M - 9.2.3.1 - YES reject NG-RAN node 1 M Global NG- YES reject ID RAN NodeID9.2.2.3NG-RAN node2 M Global NG- YES reject ID RAN NodeID9.2.2.3Task ID M An ID mappingto a particulartask (e.g. TaskID: 1 map to UEtrafficforecasting)>Updated 1 - - - - Weights List»Weights Item 0..<noo - - - - fparameters>»>weights - - - - Table 4
[0093] FIG. 7 illustrates an example computer system (700) in which or with which the proposed system (102) may be implemented, in accordance with an embodiment of the present disclosure.
[0094] As shown in FIG. 7, the computer system (700) may include an external storage device (710), a bus (720), a main memory (730), a read-only memory (740), a mass storage device (750), a communication port(s) (760), and a processor (770). A person skilled in the art will appreciate that the computer system (700) may include more than one processor and communication ports. The processor (770) may include various modules associated with embodiments ofthe present disclosure. The communication port(s) (760) may be any of an RS- 232 port for use with a modem-based dialup connection, a 10 / 100 Ethernet port, a Gigabit or10 Gigabit port using copper or fiber, a serial port, a parallel port, or other existing or future ports. The communication ports(s) (760) may be chosen depending on a network, such as a Local Area Network (LAN), Wide Area Network (WAN), or any network to which the computer system (700) connects.
[0095] In an embodiment, the main memory (730) may be Random Access Memory (RAM), or any other dynamic storage device commonly known in the art. The read-only memory (740) may be any static storage device(s) e.g., but not limited to, a Programmable Read Only Memory (PROM) chip for storing static information e.g., start-up or basic input / output system (BIOS) instructions for the processor (770). The mass storage device (750) may be any current or future mass storage solution, which can be used to store information and / or instructions. Exemplary mass storage solutions include, but are not limited to, Parallel Advanced Technology Attachment (PATA) or Serial Advanced Technology Attachment (SATA) hard disk drives or solid-state drives (internal or external, e.g., having Universal Serial Bus (USB) and / or Firewire interfaces).
[0096] In an embodiment, the bus (720) may communicatively couple the processor(s) (770) with the other memory, storage, and communication blocks. The bus (720) may be, e.g. a Peripheral Component Interconnect PCI) / PCI Extended (PCI-X) bus, Small Computer System Interface (SCSI), USB, or the like, for connecting expansion cards, drives, and other subsystems as well as other buses, such a front side bus (FSB), which connects the processor (770) to the computer system (700).
[0097] In another embodiment, operator and administrative interfaces, e.g., a display, keyboard, and cursor control device may also be coupled to the bus (720) to support direct operator interaction with the computer system (700). Other operator and administrative interfaces can be provided through network connections connected through the communication port(s) (760). Components described above are meant only to exemplify various possibilities. In no way should the aforementioned exemplary computer system (700) limit the scope of the present disclosure.
[0098] FIG. 8 illustrates an example high-level flow diagram of a proposed method (800) implemented by the proposed system (102), in accordance with an embodiment of the present disclosure.
[0099] In an embodiment, the method (800) for predicting user equipment (UE) traffic measurements of each slice may be performed by one or more base station nodes such as gNB (302, 402, 502, 604) and its peers (304, 404, 504, 606). The method (800) begins with identifying (802), by a base station node, at least one federated learning capable neighbouringbase station node. The identification may be performed by the system (102) through solicitation and registration messages as described in FIG. 3 at step 306 and FIG. 4 at step 406.
[0100] In an embodiment, once the participating base station nodes are identified, aggregating (804) one or more UE traffic measurements of each slice associated with each of the identified base station nodes for a predetermined period. Each participating node may aggregate UE traffic measurements of each slice associated with its own coverage area. This aggregation corresponds to the local data collection step 308 of FIG. 3 and forms the basis for subsequent analysis and prediction.
[0101] In an embodiment, the method (800) includes transmitting (806), by the base station node, information associated with parameters of a local machine learning model to the identified neighbouring base station nodes. In parallel, receiving (808), by the base station node, model information associated with the identified neighbouring base station nodes. This exchange of parameters is performed via peer-to-peer communication and corresponds to steps 314 and 316 of FIG. 3.
[0102] In an embodiment, the method (800) includes aggregating (810), by the base station node, the received model information with the information of the base station node. The method further includes repeating (812) the model information exchange until the model of the base station node converges. This iterative aggregation corresponds to step 320 of FIG. 3 and ensures distributed convergence across participating nodes.
[0103] In an embodiment, once convergence is achieved, predicting (814), by the base station node, one or more UE traffic measurements of each slice associated with the base station node for one or more subsequent periods. The prediction may be performed in either a unistep or a multistep manner, thereby enabling both short-term responsiveness and long-term slice capacity planning.
[0104] In an embodiment, the method (800) includes monitoring (816), by the base station node, the performance of the model. In response to a determination that one or more performance metrics are below a threshold, the base station node may retune the model parameters by returning to step 804 for further local data collection and retraining. This ensures that the prediction framework remains accurate under dynamic UE mobility conditions.
[0105] In an embodiment, the method (800) may include transmitting a request to each of the one or more base station nodes for predicting UE traffic measurements of each slice. Upon confirmation of the request, the system (102) registers the base station nodes for participation in the federated learning cycle.
[0106] In an embodiment, each of the one or more base station nodes may comprise a machine learning model that processes the UE traffic measurements of each slice for the second predetermined period to generate one or more parameters. These parameters may include throughput, bearer-level KPIs, or PDCP data volume.
[0107] In an embodiment, the method (800) may include determining information based on the aggregated parameters, wherein the parameters are communicated among each of the identified base station nodes. This ensures that the federated learning process is distributed and resilient.
[0108] In an embodiment, the method (800) may include determining whether each of the base station nodes allows processing of UE traffic measurements within an aggregation time period and is capable of transmitting parameters to K base stations within an exchange time period. This determination ensures that only synchronized nodes participate in the federated learning cycle, thereby enabling dynamic participation.
[0109] In an embodiment, the method (800) may include dynamically allowing one or more base station nodes to join or leave the federated learning process depending on synchronization with the aggregation time period and the exchange time period.
[0110] While considerable emphasis has been placed herein on the preferred embodiments, it will be appreciated that many embodiments can be made and that many changes can be made in the preferred embodiments without departing from the principles of the disclosure. These and other changes in the preferred embodiments of the disclosure will be apparent to those skilled in the art from the disclosure herein, whereby it is to be distinctly understood that the foregoing descriptive matter is to be implemented merely as illustrative of the disclosure and not as a limitation.ADVANTAGES OF THE INVENTION[oni] The present disclosure allows each base station node in the network slice operation phase to perform resource capacity planning and modification in the network slices for Tn+1 without using any complex heuristics which are time consuming and computationally inaccurate.
[0112] The present disclosure ensures that federated learning is not affected even if one node failure occurs and enhances privacy because of the absence of a central aggregator. This resilience is achieved through the decentralized peer-to-peer architecture, which eliminates dependence on a central aggregator.
[0113] The present disclosure ensures that privacy is maintained while federated learning. This is achieved as in the present disclosure, raw traffic data remains distributed across base station nodes rather than being collected at a central point, and only model parameters are exchanged among distributed base station nodes. The absence of a central aggregator reduces the risk of data leakage, improves scalability, and ensures that the federated learning process remains robust across heterogeneous deployment scenarios.
[0114] The present disclosure provides flexibility by supporting both unistep and multistep prediction modes. The dynamic selection between these modes allows the system to adapt to varying operational requirements, traffic fluctuations, accuracy thresholds, and aggregation time periods, thereby strengthening the overall efficiency of slice resource management.
Claims
We Claim:
1. A system (102) for predicting user equipment (UE) traffic measurements of each slice, comprising:a processor (202) communicatively coupled with at least one base station node; a memory (204) operatively coupled with the processor (202), wherein said memory (204) stores instructions which, when executed by the processor (202), cause the processor (202) to:identify, by the at least one base station node, at least one federated learning capable neighbouring one or more base station nodes; aggregate, by the base station node, one or more UE traffic measurements of each slice associated with the base station node for a predetermined period;transmit, by the at least one base station node, information associated with parameters of a model in the base station node, to the identified one or more base station nodes;receive, by the at least one base station node, model information associated with the identified one or more base station nodes from the identified one or more base station nodes;aggregate, by the at least one base station node, the received model information from the identified one or more base station nodes with the information of the base station node;repeat, by the at least one base station node, model information exchange until the model of the base station node converges;predict, by the at least one base station node, one or more UE traffic measurements of each slice associated with the base station node for one or more subsequent periods; andmonitor, by the at least one base station node, the model performance and retuning the model.
2. The system (102) as claimed in claim 1, wherein the system (102) is configured to transmit a request to each of the one or more base station nodes for registering with the system (102) and upon confirmation of the request from the one or more base station nodes, register the one or more base station nodes.
3. The system ( 102) as claimed in claim 2, wherein each of the identified one or more base station nodes comprise a machine learning model that processes the one or more UE traffic measurements of each slice associated with the one or more base station nodes for predicting one or more UE traffic measurements of each slice associated with the one or more base station nodes for one or more subsequent period.
4. The system (102) as claimed in claim 3, wherein each of the identified one or more base station nodes are configured to transmit the model information among each other and the system (102) is configured to determine the one or more metrics based on the aggregated one or more parameters.
5. The system ( 102) as claimed in claim 3 , wherein to identify the one or more base station nodes, the system (102) is configured to determine if each of the one or more base station nodes allow processing of the one or more UE traffic measurements of each slice within an aggregation time period and are capable of transmitting the one or more parameters to K base stations from the one or more base station nodes within an exchange time period.
6. The system (102) as claimed in claim 1, wherein the one or more base station nodes can dynamically join or leave the federated learning process based on synchronization with an aggregation time period and an exchange time period.
7. The system (102) as claimed in claim 1, wherein monitoring the model performance comprises determining whether one or more performance metrics of the model are below a threshold, and in response to a positive determination, returning to aggregate the one or more UE traffic measurements of each slice for further local data collection and retraining of the model.
8. A method (800) for predicting user equipment (UE) traffic measurements of each slice, the method comprising:identifying (802), by a base station node, at least one federated learning capable neighbouring base station nodes;aggregating (804), by the base station node, one or more UE traffic measurements of each slice associated with the base station nodes for a predetermined period;transmitting (806), by the base station node, information associated with parameters of a model in the base station node, to the identified one or more base station nodes;receiving (808), by the base station node, model information associated with the identified one or more base station nodes from the identified one or more base station nodes;aggregating (810), by the base station node, the received model information from the identified one or more base station nodes with the information of the base station node;repeating (812), by the base station node, model information exchange until the model of the base station node converges;predicting (814), by the base station node, one or more UE traffic measurements of each slice associated with the base station node for one or more subsequent periods; andmonitoring, by the base station node, the model performance and retuning the model.
9. The method (800) as claimed in claim 8, comprising transmitting, a request to each of the one or more base station nodes for predicting the UE traffic measurements of each slice and upon confirmation of the request from the one or more base station nodes, registering the one or more base station nodes.
10. The method (800) as claimed in claim 9, wherein each of the one or more base station nodes comprise a machine learning model that processes the one or more UE traffic measurements of each slice associated with the one or more base station nodes for predicting one or more UE traffic measurements of each slice associated with the one or more base station nodes for one or more subsequent period.
11. The method (800) as claimed in claim 8, wherein the model information associated with the identified one or more base station nodes are communicated among each of the identified one or more base station nodes.
12. The method (800) as claimed in claim 8, comprising determining whether each of the one or more base station nodes is capable of processing of the one or more UE traffic measurements of each slice within an aggregation time period and of transmitting the one or more parameters to K base station nodes within an exchange time period, for participation in the federated learning process.
13. The method (800) as claimed in claim 8, comprising dynamically allowing one or more base station nodes to join or leave the federated learning process depending on synchronization with an aggregation time period and an exchange time period.
