System and method of managing and reducing inter-cell interference

Abstract

A method of managing and reducing Inter-Cell Interference (ICI), between plurality of Base Stations (BSs) in data communications network, the method, being carried out at Central Unit (CU) which coordinates the BS transmissions, comprising steps of receiving Channel State Information (CSI) from each BS, where each BS receives the CSI from each of plurality of User Equipments (UEs), where CSI is in compressed, Orthogonality Capturing, Representation (OCR) and inputting the received OCR information into plurality of OCR translation models, where one OCR translation model is assigned to each BS, where each OCR translation model is Artificial Intelligence (AI) based model which, for each respective BS, maps respective OCRs for each UE to global OCR space, using output of each OCR translation model to coordinate scheduling of transmission from each BS and sending resulting scheduling instructions from coordination to each BS to thereby manage and reduce ICI.

Description

[0001] SYSTEM AND METHOD OF MANAGING AND REDUCING INTER-CELL INTERFERENCE

[0002] TECHNICAL FIELD

[0003] The present disclosure relates generally to the field of wireless communication systems; and more specifically, to a system and method of managing and reducing inter-cell interference.

[0004] BACKGROUND

[0005] In modem wireless communication systems, the increasing density of Base Stations (BSs) and User Equipments (UEs) has led to significant challenges in managing Inter-Cell Interference (ICI). The ICI occurs when multiple BSs transmit to UEs in overlapping coverage areas using the same resource blocks, resulting in degraded Quality of Service (QoS), particularly for UEs located at the edges of BSs (or cells). The interference problem is further exacerbated in dense urban deployments and large-scale networks, where the proximity of BSs and the number of active UEs are substantially higher. A widely adopted approach to reduce the ICI is Scheduling Coordination (SC). In SC, the BSs share Channel State Information (CSI) with a Central Unit (CU) or neighbouring BSs to coordinate their transmission schedules. This ensures that the UEs in the shared coverage areas are assigned distinct resource blocks, minimizing interference. The existing SC solutions often involve acquiring CSI at each BS, followed by exchanging the information with the CU to enable centralized scheduling decisions. This process introduces considerable overhead due to the high dimensionality of CSI, particularly in dense deployments with multiple UEs and BSs. The centralized SC communication systems rely heavily on detailed CSI feedback from the UEs, which includes quantized reports on the wireless channel conditions. Such data is significant for making accurate scheduling decisions but results in considerable communication overhead between BSs and the CU. Furthermore, the UEs located at the cell edges often require CSI acquisition from multiple BSs, increasing the complexity and scalability issues of current SC approaches. As a result, reducing the CSI overhead while ensuring effective ICI management remains a significant challenge for next-generation wireless networks. The traditional SC methods, such as those based on Coordinated Multi-Point (CoMP) transmission and reception, have been proposed to address these challenges by optimizing resource allocation at the CU. However, the traditional methods are often limited by the sheer volume of CSI data that must be collected, shared, and processed. This necessitates the development of approaches that can maintain the benefits of centralized coordination while minimizing the associated data transfer and computational requirements. Consequently, there exists a technical problem of excessive CSI feedback overhead, inefficient management of ICI, and limited scalability in existing scheduling coordination mechanisms for dense wireless communication systems.

[0006] Therefore, in light of the foregoing discussion, there exists a need for an improved scheduling coordination framework that can efficiently manage ICI by reducing CSI feedback overhead through compressed representations, while maintaining the accuracy and scalability required for dense and next-generation wireless networks.

[0007] SUMMARY

[0008] The present disclosure provides a system and method of managing and reducing inter-cell interference. The present disclosure provides a solution to the existing problem of excessive CSI feedback overhead, inefficient management of inter-cell interference, and limited scalability in existing scheduling coordination mechanisms for dense wireless communication systems. An aim of the present disclosure is to provide a solution that overcomes at least partially the problems encountered in the prior art and provides an improved scheduling coordination framework that can efficiently manage inter-cell interference by reducing CSI feedback overhead through compressed representations while maintaining the accuracy and scalability required for dense and next-generation wireless networks.The object of the present disclosure is achieved by the solutions provided in the enclosed independent claims. Advantageous implementations of the present disclosure are further defined in the dependent claims.

[0009] In one aspect, the present disclosure provides a method of managing and reducing Inter-cell Interference (ICI), between a plurality of Base Stations (BSs) in a data communications network. The method, is carried out at a Central Unit (CU) which coordinates the BS transmissions, and comprising steps of receiving Channel State Information (CSI) from each BS, where each BS receives the CSI from each of a plurality of User Equipments (UEs), where the CSI is in a compressed, Orthogonality Capturing Representation (OCR). The method further involves inputting the received OCR information into a plurality of OCR translation models, where one OCR translation model is assigned to each BS, where each OCR translation model is an Artificial Intelligence (Al) based model which, for each respective BS, maps respective OCRs for each UE to a global OCR space. The method further involves using an output of each OCR translation model to coordinate the scheduling of transmission from each BS and sending resulting scheduling instructions from the coordination to each BS to thereby manage and reduce ICI.

[0010] The disclosed method for managing and reducing ICI in a data communications network introduces an advanced framework designed to address the limitations of conventional scheduling coordination techniques. The method, executed at the CU, leverages an approach by utilizing CSI in a compressed format called the OCR, which each BS receives from the associated plurality of UEs. The received OCR data is processed through dedicated OCR translation models at the CU, where each translation model is an Al-driven model assigned to a specific BS. The Al models map the respective OCRs for each UE into a unified global OCR space, preserving the essential characteristics of the CSI while significantly reducing overhead. Using the output from the translation models, the CU coordinates the scheduling of transmissions across all BSs, ensuring enhanced resource allocation and minimal ICI. The resulting scheduling instructions are then transmitted back to each BS, enabling seamless and efficient management of interference in dense wireless environments. The method combines innovative Al-based compression with centralized coordination to deliver a scalable, effective solution that enhances network performance and supports next-generation communication systems.

[0011] In an implementation form, in addition to receiving the OCR information from each BS, the CU also receives labels for training the Al based models at the CU.

[0012] The CU receives labels to ensure that the Al models are trained effectively, allowing for improved accuracy in predicting user behavior and maximizing resource allocation. This approach leads to enhanced performance in managing interference effects and improves the scheduling decisions made by the CU, resulting from the effective training of Al models using the received labels.

[0013] In a further implementation form, the labels including global data.

[0014] The utilization of global data enhances the accuracy of the Al-based translation models, resulting in improved performance in converting the CSI representations to a unified domain.

[0015] In a further implementation form, the global data includes global position coordinates.

[0016] The utilization of global position coordinates with the OCRs enables precise localization and improved data alignment, resulting in enhanced performance of the translation models.

[0017] In a further implementation form, the global data includes time references.

[0018] The inclusion of time references is required for correlating user data across the plurality of base stations, enabling accurate tracking of user behavior and resource allocation. This approach enhances the precision of scheduling decisions by ensuring that the data reflects real-time user interactions, improving resource management and communication efficiency.In a further implementation form, the time references are timestamps.

[0019] The use of timestamps enhances the precision of data synchronization, leading to improved scheduling decisions and more efficient communication between the CU and associated BSs.

[0020] In a further implementation form, the global data includes user identifiers (IDs).

[0021] This approach is employed to ensure accurate user representation and facilitate the integration of data from multiple sources, which is required for maximizing network performance. The method enhances the efficiency of data processing by enabling the conversion of diverse user data into a common format, thereby improving the accuracy of subsequent analyses and predictions.

[0022] In a further implementation form, the OCR is integrated as part of a structured representation of the Wireless Environment State Information (WESI)-AI based air interface.

[0023] The OCR data is seamlessly utilized within the WESI-AI architecture, improving scalability, reducing CSI overhead, and enabling more accurate and efficient scheduling coordination in dense network environments. The result of the integration of OCR is improved responsiveness and adaptability of the air interface, leading to enhanced transmission efficiency and reliability in varying wireless conditions.

[0024] In a further implementation form, each BS uses representation learning techniques to extract the compressed OCR from each CSI.

[0025] This allows the OCR to effectively mimic the behavior of CSI while significantly reducing data overhead. As a result, the method provides an improved scalability, minimized network congestion, and enabled efficient coordination of transmissions at the CU, ultimately enhancing network performance and reducing ICI.

[0026] In a further implementation form, the compressed OCR has lower dimensionality as compared to the CSI prior to extraction and compression.

[0027] By reducing the dimensionality, the method significantly decreases the amount of data transmitted from the plurality of BSs to the CU, improving scalability and efficiency in dense network deployments. This approach allows for more effective scheduling and interference management while maintaining high network performance.

[0028] In a further implementation form, each BS employs representation learning techniques to extract the compressed OCR from each CSI, using a loss function that minimizes the difference between the Euclidean distance of a pair of OCRs and the orthogonality coefficient computed from the corresponding pair of CSI measurements.

[0029] By minimizing the loss function, the representation learning techniques ensure that the compressed OCRs retain key orthogonality characteristics of the original CSI. This enables the CU to make accurate scheduling decisions without requiring full CSI transmission, thus reducing CSI overhead, improving scalability, and enhancing network performance.

[0030] In a further implementation form, the CU uses dedicated signalling to the plurality of BSs to trigger OCR collection at the CU. By using dedicated signalling, the CU ensures that OCRs from all BSs are collected in a timely and synchronized manner, allowing the Al-based translation models at the CU to map these representations into a unified space for effective scheduling coordination.In another aspect, there is provided a system comprising means adapted for carrying out all the steps described above with respect to the method.

[0031] The comprehensive implementation of the system enables the scalable and adaptive deployment of Al-driven wireless communication solutions, particularly in dense network environments with significant interference management.

[0032] In yet another aspect, there is provided a computer program comprising instructions for carrying out all the steps of the method described above, when the computer program is executed on a computer system.

[0033] The implementation of the computer program allows for seamless automation, real-time adaptability, and scalability in dense network environments, ensuring effective resource management and improved wireless communication performance.

[0034] It is to be appreciated that all the aforementioned implementation forms can be combined.

[0035] It has to be noted that all devices, elements, circuitry, units and means described in the present application could be implemented in the software or hardware elements or any kind of combination thereof. All steps which are performed by the various entities described in the present application as well as the functionalities described to be performed by the various entities are intended to mean that the respective entity is adapted to or configured to perform the respective steps and functionalities. Even if, in the following description of specific embodiments, a specific functionality or step to be performed by external entities is not reflected in the description of a specific detailed element of that entity which performs that specific step or functionality, it should be clear for a skilled person that these methods and functionalities can be implemented in respective software or hardware elements, or any kind of combination thereof. It will be appreciated that features of the present disclosure are susceptible to being combined in various combinations without departing from the scope of the present disclosure as defined by the appended claims.

[0036] Additional aspects, advantages, features and objects of the present disclosure would be made apparent from the drawings and the detailed description of the illustrative implementations construed in conjunction with the appended claims that follow. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The summary above, as well as the following detailed description of illustrative embodiments, is better understood when read in conjunction with the appended drawings. For the purpose of illustrating the present disclosure, exemplary constructions of the disclosure are shown in the drawings. However, the present disclosure is not limited to specific methods and instrumentalities disclosed herein. Moreover, those skilled in the art will understand that the drawings are not to scale. Wherever possible, like elements have been indicated by identical numbers.

[0038] Embodiments of the present disclosure will now be described, by way of example only, with reference to the following diagrams wherein:

[0039] FIG. 1A is a network environment diagram of a system for managing and reducing Inter-Cell Interference (ICI), in accordance with an embodiment of the present disclosure;

[0040] FIG. IB is a block diagram that illustrates various exemplary components of a Central Unit (CU), in accordance with an embodiment of the present disclosure;

[0041] FIG. 2 is a flowchart of a method of managing and reducing ICI between a plurality of Base Stations (BSs) in a data communications network, in accordance with an embodiment of the present disclosure;

[0042] FIG. 3 illustrates an implementation scenario of orthogonality capturing representation (OCR) in a multi-user Multiple Input Multiple Output (MIMO) system for managing the ICI, in accordance with an embodiment of the present disclosure;FIG. 4 is a diagram that represents the OCR-aided scheduling approach using translation models at the CU in a scenario involving the plurality of BSs, in accordance with an embodiment of the present disclosure;

[0043] FIG. 5 is a flowchart that depicts a series of operations performed in OCR-based centralized scheduling coordination, in accordance with an embodiment of the present disclosure;

[0044] FIG. 6 is a flowchart that depicts a series of operations performed in training phase of the OCR-aided scheduling coordination, in accordance with an embodiment of the present disclosure; and

[0045] FIG. 7 is a flowchart that depicts a series of operations performed for the integration of OCR into Wireless Environment State Information (WESI)-AI based air interface architecture, in accordance with an embodiment of the present disclosure. In the accompanying drawings, an underlined number is employed to represent an item over which the underlined number is positioned or an item to which the underlined number is adjacent. A non-underlined number relates to an item identified by a line linking the non-underlined number to the item. When a number is non-underlined and accompanied by an associated arrow, the non-underlined number is used to identify a general item at which the arrow is pointing.

[0046] DETAILED DESCRIPTION OF EMBODIMENTS

[0047] The following detailed description illustrates embodiments of the present disclosure and ways in which they can be implemented. Although some modes of carrying out the present disclosure have been disclosed, those skilled in the art would recognize that other embodiments for carrying out or practicing the present disclosure are also possible.

[0048] FIG. 1A is a network environment diagram of a system 100 for managing and reducing Inter-Cell Interference (ICI), in accordance with an embodiment of the present disclosure. The system 100 includes key components, such as a Central Unit (CU) 102, a plurality of Base Stations (BSs) 104, and a plurality of User Equipments (UEs) 106. The plurality of BSs 104 communicates with the plurality of UEs 106 through a communication network 108. Moreover, there is shown a plurality of Orthogonality Capturing Representation (OCR) translation models 110 inside the CU 102.

[0049] The system 100 for managing and reducing the ICI is designed to coordinate transmissions across the plurality of BSs 104 within a wireless communication system (e.g., the system 100). The system 100 operates through the CU 102 that receives Channel State Information (CSI) from each BS of the plurality of BSs 104, where the CSI is represented in a compressed form known as OCR. By using advanced artificial intelligence (Al)-based translation models (i.e., the plurality of OCR translation models 110), where each translation model is assigned to a specific BS, the system 100 maps the BS-specific OCRs into a unified global OCR space. This mapping enables the CU 102 to accurately analyze the spatial and orthogonality characteristics of the channels. Based on this analysis, the CU 102 coordinates the scheduling of transmissions from each BS of the plurality of BSs 104, generating and sending optimized scheduling instructions to manage and reduce the ICI effectively.

[0050] The CU 102 may be referred to as a pivotal component in the system 100 that coordinates the management of ICI across the plurality of BSs 104. The CU 102 functions as a coordinator that receives compressed CSI in the form of OCR from the plurality of BSs 104. The OCRs are processed through the plurality of OCR translation models 110 to map the data into a unified global OCR space, enabling accurate analysis and coordination. Additionally, the CU 102 handles the training data, such as global position coordinates, timestamps, and user IDs, while employing dedicated signalling to trigger OCR collection from the plurality of BSs 104. Ultimately the CU 102 serves as the intelligence hub for optimizing network performance and minimizing interference between cells.

[0051] The plurality of BSs 104 may be referred to as multiple wireless communication nodes within a communication system (i.e., the system 100) that facilitate data transmission and reception between the plurality of UEs 106 through the communication network 108. Each base station of the plurality of BSs 104 is equipped with suitable logic, circuitry, antennas, and / or interfaces that is configured to support wireless communication and process the CSI received from each of the plurality of UEs 106. Eachof the plurality of BSs 104 is responsible for collecting and compressing the CSI into the OCR, which is then sent to the CU 102 for scheduling coordination. Examples of each of the plurality of BSs 104 may include, but are not limited to, a macro base station, a micro base station, a small cell, a femtocell, or any other communication infrastructure supporting wireless connectivity in a cellular or non-cellular network environment.

[0052] The plurality of UEs 106 refers to multiple devices that can access the communication network 108, each functioning as an endpoint for data transmission and reception. The plurality of UEs 106 may include suitable logic, circuitry, interfaces and / or code that is configured to transmit the CSI to the plurality of BSs 104. Examples of the plurality of UEs 106 may include, but are not limited to, a smart phone, a laptop, a tablet, an Intemet-of-Things (loT) device, a Machine Type Communication (MIC) device, a computing device, a server, a drone, a customized hardware for wireless telecommunication, a transceiver, or any other portable or non-portable electronic device.

[0053] The communication network 108 refers to an interconnected infrastructure that enables wireless communication between the plurality of BSs 104 and the plurality of UEs 106. The communication network 108 provides the framework for transmitting and receiving data, managing resources, and coordinating network operations. The communication network 108 includes various components, such as the core network, backhaul connections, and air interfaces that link the plurality of BSs 104 and the plurality of UEs 106. The communication network 108 facilitates the exchange of CSI, which is required for efficient scheduling and interference management. Examples of the communication network 108 may include, but are not limited to, a cellular network (e.g., a 5G, or 5G NR network, such as sub 6 GHz, cmWave, or mmWave communication network), a cloud network, a Local Area Network (LAN), a vehicle-to-network (V2N) network, a Metropolitan Area Network (MAN), and / or the Internet.

[0054] The plurality of OCR translation models 110 refers to a set of Artificial Intelligence (Al)-based models implemented at the CU 102 to facilitate efficient coordination of downlink transmissions across the plurality of BSs 104. The plurality of OCR translation models 110 are designed to map the OCRs, which are compressed forms of CSI generated at the plurality of BSs 104, into a unified global OCR space. The plurality of OCR translation models 110 ensures efficient data transmission, resource management, and interference mitigation across the network. Examples of applications for the plurality of OCR translation models 110 include but are not limited to, multi-user Multiple Input Multiple Output (MIMO) systems, centralized scheduling coordination, and advanced Al-based air interface architectures.

[0055] The system 100 is designed to implement OCR-based scheduling coordination, where each BS extracts and transmits compressed CSI representations to the CU 102 using Al-based techniques. The CU 102, in turn, employs dedicated signalling to trigger OCR collection, processes the received data through translation models, and optimizes scheduling decisions to minimize ICE The system 100 ensures efficient data transmission, reduced CSI feedback overhead, and enhanced network performance.

[0056] FIG. IB is a block diagram that illustrates various exemplary components of a central unit, in accordance with an embodiment of the present disclosure. FIG. IB is described in conjunction with elements from FIG. 1A. With reference to FIG. IB, there is shown a block diagram that illustrates various exemplary components of the CU 102. The CU 102 includes a memory 112, a network interface 114, and a processor 116 in addition to the plurality of OCR translation models 110. In an implementation scenario, the plurality of OCR translation models 110 may be stored inside the memory 112.

[0057] The memory 112 refers to a storage component integrated within the CU 102 that is responsible for storing data, algorithms, and models required for network operations. The memory 112 may store the training data, labels, and other global information, such as user identifiers, timestamps, and position coordinates, required for training and updating the Al-based translation models (such as the plurality of OCR translation models 110). The memory 112 ensures efficient data access and retrieval,enabling the CU 102 to coordinate scheduling, manage resources, and mitigate interference across the network. The memory 112 may include suitable logic, circuitry, and / or interfaces that is configured to store machine code and / or instructions executable by the processor 116. Examples of implementation of the memory 112 may include, but are not limited to, an Electrically Erasable Programmable Read-Only Memory (EEPROM), Random Access Memory (RAM), Read Only Memory (ROM), Hard Disk Drive (HDD), Flash memory, a Secure Digital (SD) card, Solid-State Drive (SSD), a computer readable storage medium, and / or CPU cache memory. A computer readable storage medium for providing a non-transient memory may include, but is not limited to, an electronic storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing.

[0058] The network interface 114 refers to a communication component within the CU 102 that facilitates data exchange between the CU 102 and the plurality of BSs 104 in the network. The network interface 114 is responsible for transmitting and receiving compressed OCRs from the plurality of BSs 104, as well as sending scheduling instructions and other signalling messages back to the plurality of BSs 104. The network interface 114 ensures seamless connectivity and efficient data transfer, enabling the CU 102 to collect the required information for managing the ICI and coordinating transmissions. The network interface 114 supports various communication protocols and interfaces, such as Ethernet, optical links, or wireless backhaul, to ensure robust and high-speed connectivity across the network. Examples of the network interface 114 may include but are not limited to, a wireless transceiver, an antenna, a radio transceiver, and the like.

[0059] The processor 116 may include suitable logic, circuitry, and / or interfaces that is configured to execute instructions stored in the memory 112. Examples of the processor 116 may include, but are not limited to an integrated circuit, a co-processor, a microprocessor, a microcontroller, a complex instruction set computing (CISC) processor, an application-specific integrated circuit (ASIC) processor, a reduced instruction set (RISC) processor, a very long instruction word (VLIW) processor, a central processing unit (CPU), a state machine, a data processing unit, and other processors or circuits. Moreover, the processor 116 may refer to one or more individual processors, processing devices, a processing unit that is part of a machine.

[0060] The system 100 achieves efficient data processing and transmission coordination while addressing the key challenges of scalability, data overhead, and complexity. By compressing the high-dimensional CSI into OCRs, the system 100 reduces the communication overhead between the plurality of BSs 104 and the CU 102. Furthermore, the use of Al-based translation models (i.e., the plurality of OCR translation models 110) ensures accurate mapping with minimal latency, even in dense network deployments with the plurality of BSs 104 and the plurality of UEs 106. The system 100 is equipped to handle the dynamic demands of modem wireless communication systems, ensuring effective ICI management while maximizing resource allocation and network performance. These advancements enable the system 100 to support high-performance, scalable, and energy-efficient operations in diverse wireless communication scenarios.

[0061] In operation, there is provided the system 100 for managing and reducing the ICI, between a plurality of BSs 104 in a data communications network. The system 100 comprises the CU 102 configured to coordinate data transmissions with the plurality of BSs 104. The CU 102 is configured to receive CSI from each BS, where each BS receives the CSI from each of the plurality of UEs 106, where the CSI is a compressed and OCR. The channel state information refers to the data that characterizes the current conditions of a communication channel, including parameters, such as signal strength and interference levels. The OCR (or compressed CSI) retains the required characteristics of the original CSI, specifically capturing the channel's orthogonality properties, which are significant for interference management. By receiving the CSI from the plurality of BSs 104 in the compressed form (i.e., the OCR) at the CU 102, the system 100 leads to minimization of data transfer requirements while preserving the information required for effective scheduling and resource coordination.

[0062] The CU 102 is further configured to input the received OCR information into the plurality of OCR translation models 110, where one OCR translation model is assigned to each BS, where each OCR translation model is an artificial intelligence (Al)based model which, for each respective BS, maps respective OCRs for each UE to a global OCR space. The global OCR space may be referred to as a unified representation domain where OCRs from different BSs are mapped into a common feature space, enabling seamless integration and comparability across the network. The plurality of OCR translation models 110, plays a significant role in mapping the OCRs of the plurality of UEs 106 to a unified global OCR space. This mapping is required because the OCRs received from the plurality of BSs 104 may not be directly comparable due to their local encoding. For example, in a multi-BS scenario, the OCRs obtained at each BS cannot be directly exploited by the CU 102 to coordinate the downlink scheduling and avoid ICI, as the Euclidean distance between two OCRs corresponding to two different BSs is not guaranteed to be representative for the orthogonality between the corresponding channels. Therefore, one OCR translation model assigned to each BS is deployed at the CU 102. The OCR translation model maps the OCRs of the plurality of UEs 106 to the global OCR space. By transforming them into the global OCR space, the transformation enables the CU 102 to analyze the relationships and orthogonality characteristics between the plurality of UEs 106 across the plurality of BSs 104. The mapping of the respective OCRs of each UE into the global OCR space is shown and described in detail, for example, in FIG.

[0063] 4.

[0064] The CU 102 is further configured to use an output of each OCR translation model to coordinate the scheduling of transmission from each BS. After receiving the OCRs from each BS, each OCR is mapped to the global OCR space using the respective translation model in the plurality of the translation models 110, the CU 102 is configured to use this unified representation to analyze the spatial and orthogonality relationships among the plurality of UEs 106 connected to the plurality of BSs 104. The CU 102 is further configured to coordinate the scheduling for downlink transmissions by computing distances in the global OCR space. By leveraging the outputs of the plurality of OCR translation models 110, the system 100 ensures that the scheduling decisions are informed, accurate, and scalable, even in complex network environments with overlapping BS coverage. Further, the CU 102 is configured to send resulting scheduling instructions from the coordination to each BS to thereby manage and reduce the ICI. The ICI arises when the plurality of BSs 104 transmits data over the same resource blocks, leading to signal overlap and degradation in network performance, particularly for UEs located at cell edges. By generating scheduling instructions based on the processed OCRs and their mappings in the global OCR space, the CU 102 ensures that each BS transmits data in a manner that minimizes interference with neighbouring BSs. These instructions dictate when and how each BS should utilize its transmission resources, such as time slots, frequency bands, or spatial beams, to avoid overlapping with other BSs in shared coverage areas.

[0065] Consider a downlink multi-user MIMO system where each BS is equipped with NBantennas and each of the plurality of UEs 106 are equipped with Nvantennas. Typically, users are scheduled in the same transmission block based on the channel orthogonality coefficient, which, for the pair of UEs (u, it') is defined as given in Equation (1)

[0066]

[0067] where Hn uand Hn uiare complex CSI matrices of dimension Nvx NBacquired at the BS n corresponding to users u and u' , respectively. The lower yu u< indicates better orthogonality, thus reduced interference when such u and u' are scheduled on the same transmission block.

[0068] Specifically, for a multi-carrier communication system (e.g., Orthogonal Frequency Division Multiplexing) where transmissions occur over Nscsubcarriers, the channel model and user pairing need to consider both the spatial and frequency characteristics. The orthogonality coefficient can be identified as the normalized cross-correlation between two users' channel vectors averaged over all subcarriers, as expressed in Equation (2)

[0069]

[0070] or the minimum cross correlation among the subcarriers, as shown in Equation (3)

[0071]

[0072] where j is the subcarrier index and f indicates the Hermitian operator.

[0073] In accordance with an embodiment, in addition to receiving the OCR information from each BS, the CU 102 also receives labels for training the plurality of OCR translation models 110 at the CU 102. The labels, such as user Identifiers (IDs), timestamps, and global positioning data, are required for the training of the plurality of OCR translation models 110 as the labels provide context and enable the CU 102 to accurately align and pair user representations across different BSs. The training process involves minimizing the Euclidean distance between OCR representations in a unified domain using a loss function, ensuring that the CU 102 can effectively coordinate scheduling decisions to mitigate ICI. By leveraging this label-enriched training approach, the CU 102 ensures that the plurality of OCR translation models 110 (or the Al models) are robust and capable of handling real-world scenarios, including dynamic channel conditions and dense network environments.

[0074] In accordance with an embodiment, the labels include global data. The global data refers to information that is accessible and applicable across multiple nodes or components within a communication network, enabling interoperability and comprehensive analysis of network performance and behavior. The labels received by the CU 102 include global data, such as global positioning coordinates, time references, and user IDs. Specifically, the global data is used to match user representations from different BSs and establish paired representations for users at cell edges. By incorporating the global data in training of the plurality of OCR translation models 110, the CU 102 ensures that the OCR-based scheduling models are accurate and capable of mitigating ICI by coordinating scheduling decisions based on the spatial and temporal dynamics of users across the network. In accordance with an embodiment, the global data includes global position coordinates. The global position coordinates may be referred to as a set of numerical values that represent a specific location on the earth's surface, typically expressed in latitude and longitude, which can be utilized for navigation, mapping, and location-based services. The global data received by the CU 102 includes global position coordinates, which are significant for training the plurality of OCR translation models 110 (or the Al-based models) and enabling effective scheduling coordination. The global position coordinates provide spatial context for the plurality of UEs 106 across different BSs, allowing the CU 102 to align and match OCRs for users located at cell edges. For example, if Ztand Zj represent OCRs of two UEs received from the plurality of BSs 104, the global position coordinates of the two UEs are used to identify whether the users are in overlapping coverage areas or near cell boundaries. By incorporating the global position coordinates as part of the training labels, the CU 102 can accurately model the spatial relationships between the plurality of UEs 106 and optimize scheduling decisions to mitigate the ICI.

[0075] In accordance with an embodiment, the global data includes time references. The time references refer to specific indicators or markers that denote particular points or intervals in time within a communication network, facilitating synchronization and coordination of data transmission and processing activities. The global data received by the CU 102 includes time references, which play a pivotal role in training the plurality of OCR translation models 110 (or the Al-based models) for scheduling coordination. The time references allow the CU 102 to synchronize and associate OCRs received from different BSs with the dynamic behavior of the plurality of UEs 106 over time. For example, the OCRs Z;and Zj, corresponding to users i and j from different BSs, are matched based on their time references to ensure temporal alignment when training the plurality of OCR translation models 110. The time references ensure that the paired representations correspond to the same temporal instances, allowing the plurality of OCR translation models 110 (or the Al models) to capture real-time channel dynamics and user mobility patterns. This temporal granularity enhances the CU's ability to coordinate scheduling decisions effectively, mitigating the ICI and optimizing resource allocation in dynamic and dense network environments.In accordance with an embodiment, the time references are timestamps. The timestamps refer to digital markers that indicate the specific date and time at which an event occurs within a communication network, facilitating the accurate sequencing and retrieval of data packets or messages. The timestamps are utilized to represent time references, aligning with a specified duration defined in the 3GPP standard. Each timestamp corresponds to a specific time and frequency granularity, allowing for precise characterization of physical quantities. The timestamps are required for associating the OCRs received from different BSs with specific UEs at precise time intervals. By including timestamps as part of the metadata accompanying the OCRs, the CU 102 can effectively align and synchronize data collected from multiple BSs. This alignment is required for training the AI-based OCR translation models (i.e., the plurality of OCR translation models 110), particularly when matching user representations from different BSs to map the collected OCRs into the global OCR space. The utilization of timestamps minimizes temporal discrepancies, ensuring that the orthogonality characteristics captured at different BSs correspond to the same time frame, thereby improving the accuracy of the Al-based OCR translation models (i.e., the plurality of OCR translation models 110).

[0076] In accordance with an embodiment, the global data includes user identifiers (IDs). The user IDs (or the IDs of UEs associated to a specific BS) refer to unique alphanumeric strings or codes assigned to individual users within a communication network, facilitating the identification and authentication of users (i.e., the plurality of UEs 106) for access to network resources and services. The user IDs are vital for uniquely identifying UEs associated with the plurality of BSs 104 and aligning their OCRs during data processing at the CU 102. By including the user IDs (or the IDs of the plurality of UEs 106) in the global data for training of the plurality of OCR translation models 110, the system 100 ensures that the Al-based OCR translation models can reliably match OCRs corresponding to the same UE, even when collected from different BSs. This matching process is significant for mapping the BS-specific OCRs into the global OCR space, allowing the CU 102 to analyze and coordinate scheduling decisions more accurately.

[0077] In accordance with an embodiment, the OCR is integrated as part of a structured representation of the Wireless Environment State Information (WESI)-AI based air interface. The WESI refers to data that characterizes a wireless communication environment's operational conditions and characteristics through various parameters, such as signal strength, interference levels, and user density. The WESI framework is designed to represent the state of the wireless environment in a structured and multi-dimensional format, where each element corresponds to a specific physical property, such as time, frequency, or spatial dimensions. The WESI-AI based air interface utilizes structured representations of environmental state information to enable Al-driven decision-making in network operations. By incorporating the OCR into the WESI-AI framework, the system 100 effectively captures and processes channel orthogonality characteristics within a compressed and structured format. The integration is vital because it allows the Al models (i.e., the plurality of OCR translation models 110) at the CU 102 to predict future OCRs, thereby enabling proactive scheduling and interference management. The integration is achieved by specifying the OCRs as a WESI property with defined dimensionality, size, and space / time / frequency span, along with a translation flag indicating BS-specific data that requires transformation.

[0078] In accordance with an embodiment, each BS uses representation learning techniques to extract the compressed OCR from each CSI. The representation learning technique refers to a set of techniques in machine learning that enable the automatic discovery of representations or features from raw data, facilitating improved performance on various tasks, such as classification and clustering. The extraction of the compressed OCR from each CSI is significant because the raw CSI data is high-dimensional and requires considerable bandwidth for transmission to the CU 102. By leveraging representation learning techniques, the BSs can transform the CSI into a lower-dimensional OCR while preserving the required orthogonality characteristics for interference management. The extraction is achieved through an Al-based compression function, trained using CSI data pairs, and further, optimizing a loss function that minimizes the difference between the orthogonality coefficient of the original CSI and the Euclidean distance in the compressed OCR space. This allows the OCR to effectively mimic the behavior of CSI while significantly reducing data overhead. As a result, the system 100 manifests scalability, minimized network congestion, andefficient coordination of scheduling transmissions at the CU 102, which ultimately leads to enhanced network performance and reduced ICI.

[0079] In accordance with an embodiment, the compressed OCR has lower dimensionality as compared to the CSI prior to extraction and compression. The lower dimensionality is vital to minimize data transmission overhead between the plurality of BSs 104 and the CU 102 while preserving the pivotal orthogonality characteristics required for scheduling and interference management. The OCR for each CSI is specifically designed to reduce dimensionality while effectively mitigating interference effects. By transmitting only the OCRs from the plurality of BSs 104 to the CU 102, the system 100 minimizes the volume of data transmitted, as the original CSI typically has higher dimensionality. The plurality of OCR translation models 110 at the CU 102 further processes the compressed OCR, mapping them to a unified domain for efficient scheduling decisions. This approach is utilized to enhance data transfer efficiency, particularly in MIMO-OFDM systems where the CSI can be cumbersome due to its high dimensionality.

[0080] The OCR aims at replicating the orthogonality coefficient (represented in Equation (1), Equation (2), and Equation (3)) in the space of compressed, BS-specific CSI representations.

[0081] Let V.nbe the set of UEs associated with a BS n. The CSI matrix Hn u, u G V.nis acquired at the BS n and compressed into a low-dimensional representation Zn uwith a compression function ( / (■) such that g(Hn u) = Zn u.

[0082] The compression function is an Al model trained by giving as input a set of couples of CSI matrices {(Hn u,

[0083]

[0084] ant^ utilizing a loss function, expressed in Equation (4)

[0085]

[0086] The loss function minimizes the distance between the orthogonality coefficient of a pair of users (u, u') and the Euclidean distance between the corresponding representations Zn uand Z,l veffectively mimicking orthogonality behavior in a Euclidean space.

[0087] In accordance with an embodiment, each BS employs representation learning techniques to extract the compressed OCR from each CSI, using a loss function that minimizes the difference between the Euclidean distance of a pair of OCRs and the orthogonality coefficient computed from the corresponding pair of CSI measurements. The loss function refers to a mathematical function that quantifies the difference between the predicted values and the actual values in a model, guiding the optimization process. The Euclidean distance refers to a straight-line distance between two points in a multi-dimensional space, calculated as the square root of the sum of the squared differences of their coordinates. The orthogonality coefficient refers to a measure that quantifies the degree of orthogonality between two vectors, indicating how much they are independent of each other in a vector space. Instead of transmitting the CSI matrices in their original form having high dimensionality, each BS compresses the data into a more compact form that maintains the key relationships between different users' channel conditions. The compression process is guided by a loss function (expressed in Equation (4)) that ensures the similarity between OCRs reflecting the actual level of interference between users, as originally determined from the CSI measurements. By training the Al models (i.e., the plurality of OCR translation models 110) to minimize discrepancies between the compressed OCR distances and the interference patterns in the original CSI, the system 100 ensures that the CU 102 can accurately coordinate scheduling decisions without requiring high-volume CSI transmissions.

[0088] In accordance with an embodiment, the CU 102 uses dedicated signalling to the plurality of BSs 104 to trigger OCR collection at the CU 102. The signalling mechanism is vital for coordinating the retrieval of compressed CSI representations from the plurality of BSs, ensuring that the CU 102 has the required data to make informed scheduling decisions. Without centralized triggering, OCR transmissions from different BSs might be unsynchronized, leading to incomplete or outdated information,which can impact interference management and network efficiency. By using dedicated signalling, the CU 102 ensures that OCRs from the plurality of BSs 104 are collected in a timely and synchronized manner, allowing the Al-based translation models (i.e., the plurality of OCR translation models 110) at the CU 102 to map these representations into a unified space for effective scheduling coordination.

[0089] In accordance with an embodiment, a computer program comprising instructions for carrying out all the steps of the method described above, when the computer program is executed on a computer system. The computer program enables the automated operation of the OCR-based scheduling coordination process within a wireless communication network. Specifically, the program is designed to control the plurality of BSs 104 and the CU 102, facilitating the extraction of compressed CSI representations at the BSs, coordinating their transmission to the CU 102 using dedicated signalling, and processing the received OCRs through Al-based translation models. By executing these instructions, the computer program ensures that the CU 102 can analyze and coordinate scheduling decisions, while minimizing ICI and enhancing overall network efficiency. Thus, the system 100 employs OCR-aided scheduling coordination in a scenario with the plurality of BSs 104 offers significant advantages in managing the ICI while maximizing network performance. By leveraging the OCR, each BS can locally process and compress CSI into low-dimensional, interference-aware representations, reducing the signalling overhead associated with traditional CSI-based coordination. These compressed OCRs are transmitted to the CU 102, where Al-driven translation models (i.e., the plurality of OCR translation models 110) standardize the OCRs into the global OCR space, ensuring comparability across multiple BSs. This approach enables the CU 102 to make precise scheduling decisions based on the orthogonality between UE channels, minimizing interference and improving overall throughput. Moreover, the integration of predictive capabilities through WESI-AI architecture allows the system 100 to anticipate future channel conditions and make proactive scheduling decisions, further enhancing resource allocation and network scalability. The combined effect of distributed OCR generation at the plurality BSs 104 and centralized coordination at the CU 102 results in a highly efficient, scalable, and interference-resilient network architecture, ensuring reliable connectivity even in dense deployment scenarios.

[0090] FIG. 2 is a flowchart of a method of managing and reducing ICI between a plurality of BSs in a data communications network, in accordance with an embodiment of the present disclosure. FIG. 2 is described in conjunction with elements of the FIGs. 1 A, and IB. The method 200 includes steps 202 to 208 and is being carried out at the CU 102.

[0091] The method 200 is provided for enabling the CU 102 to manage and reduce ICI between the plurality of BSs 104 in the wireless communication network by utilizing Al-based processing of CSI. The method 200 comprises receiving CSI from each BS, where each BS acquires the CSI from the plurality of UEs 106 and compresses the received CSI into an OCR using representation learning techniques. The OCR captures the interference characteristics of the channel while reducing the data overhead. Additionally, the method 200 involves inputting the received OCR information into the plurality of OCR translation models 110, where each OCR translation model is assigned to a respective BS and maps the OCRs of its associated UEs to a global OCR space. Each OCR translation model is an Al-based model that enables the CU 102 to transform the localized OCRs into a unified representation suitable for inter-BS coordination. Moreover, the method 200 supports using the output of each OCR translation model to coordinate the scheduling of transmissions across BSs, ensuring that UEs experience minimal interference. Furthermore, the method 200 comprises sending scheduling instructions derived from the coordinated scheduling process to each BS, thereby optimizing transmission efficiency and reducing ICI. The method 200 also supports receiving additional training labels, such as global position coordinates, time references, and user identifiers (IDs), which enhance the accuracy of the Al-based translation models. By leveraging Al-driven OCR processing and translation, the method 200 enables the CU 102 to efficiently manage scheduling decisions while significantly reducing CSI overhead and improving network performance.

[0092] At step 202, the method 200 comprises receiving CSI from each BS, where each BS receives the CSI from each of the plurality of UEs 106, where the CSI is in a compressed, OCR. The step 202 involves the reception of the CSI at the CU 102 from theplurality of BSs 104, where the CSI is in the compressed OCR form. The reception of the CSI in the compressed OCR form is executed using a predefined communication protocol that facilitates efficient data exchange between the plurality of BSs 104 and the CU 102. EachBS processes the raw CSI received from the associated UEs, which is typically represented as a complexvalued matrix, and applies an Al-based representation learning model to extract the OCR.

[0093] At step 204, the method 200 comprises inputting the received OCR information into the plurality of OCR translation models 110, where one OCR translation model is assigned to each BS, where each OCR translation model is an artificial intelligence (Al) based model which, for each respective BS, maps respective OCRs for each UE to the global OCR space. The step 204 involves processing the received OCRs at the CU 102 using Al-based OCR translation models (i.e., the plurality of OCR translation models 110), which map the BS-specific OCRs for each UE to the global OCR space. The OCR translation is executed using a trained Al model that ensures consistency across the plurality of BSs 104 by aligning OCR representations in a common space. Each OCR translation model processes the received OCRs using a transformation function that maps the received OCRs from the local BS-specific space to the global OCR space so that the plurality of UEs 106 across different BSs can be accurately compared for interference-aware scheduling. The transformation allows the CU 102 to coordinate transmission scheduling efficiently and mitigate ICI by leveraging the standardized global OCR representation.

[0094] At step 206, the method 200 further comprises using an output of each OCR translation model to coordinate the scheduling of transmission from each BS. The CU 102 utilizes the transformed OCR information to analyze the spatial and orthogonality relationships among the plurality of UEs 106 served by different BSs. Since the Euclidean distances between OCRs in the global space reflect the orthogonality of the corresponding CSI matrices, the CU 102 can make informed scheduling decisions to reduce ICI.

[0095] At step 208, the method 200 further comprises sending resulting scheduling instructions from the coordination to each BS to thereby manage and reduce ICI. The CU 102 determines the efficient transmission scheduling for each BS by analyzing the orthogonality relationships between the plurality of UEs 106 and provided that UEs with high cross-correlation are not scheduled in the same transmission block. Once the scheduling decisions are finalized, the CU 102 generates scheduling instructions and transmits them to the corresponding BSs via a dedicated signalling protocol. Each BS then follows these instructions to adjust the downlink transmissions, accordingly.

[0096] In accordance with an embodiment, in addition to receiving the OCR information from each BS, the CU 102 also receives labels for training the Al based models. The Al models, particularly the plurality of OCR translation models 110, require periodic updates to improve their accuracy in mapping BS-specific OCRs to a global OCR space. The received labels serve as ground truth data, allowing the CU 102 to refine the translation process and enhance the performance of interference-aware scheduling.

[0097] In accordance with an embodiment, the labels include global data. The labels play a significant role in the training phase of AI-based models (i.e., the plurality of OCR translation models 110) at the CU 102. Specifically, the global data is used to match user representations from the plurality of BSs 104 and establish paired representations for users at cell edges.

[0098] In accordance with an embodiment, the global data includes global position coordinates. The global position coordinates provide spatial context for UEs across the plurality of BSs 104, allowing the CU 102 to align and match OCRs for UEs located at cell edges.

[0099] In accordance with an embodiment, the global data includes time references. The time references allow the CU 102 to synchronize and associate OCRs from the plurality of BSs 104 with the dynamic behavior of the associated UEs over time. In accordance with an embodiment, the time references are timestamps. The timestamps are significant for associating OCRs transmitted from the plurality of BSs 104 to the CU 102. By using timestamps, the CU 102 ensures temporal alignment of OCRs for UEs located at cell edges, allowing effective training and inference in Al-based models.In accordance with an embodiment, the global data includes user identifiers (IDs). The user IDs serve as a key reference for tracking and distinguishing UEs when processing OCRs at the CU 102. By including user IDs as part of the global data, the CU 102 can correctly associate the received OCRs with specific UEs across different BSs, ensuring accurate mapping in the global OCR space.

[0100] In accordance with an embodiment, the OCR is integrated as part of a structured representation of the Wireless Environment State Information (WESI)-AI based air interface. The WESI-AI air interface utilizes structured representations of environmental state information to enable Al-driven decision-making in network operations. By incorporating OCR into this framework, the method 200 effectively captures and processes channel orthogonality characteristics within a compressed and structured format.

[0101] In accordance with an embodiment, each BS uses representation learning techniques to extract the compressed OCR from each CSI. This allows the OCR to effectively mimic the behavior of CSI while significantly reducing data overhead. As a result, the method 200 improves scalability, minimizes network congestion, and enables the CU 102 for efficient scheduled coordination of transmissions.

[0102] In accordance with an embodiment, the compressed OCR has lower dimensionality as compared to the CSI prior to extraction and compression. The reduction in dimensionality is vital for minimizing communication overhead and computational complexity in wireless networks while preserving the characteristics of the CSI required for ICI management.

[0103] In accordance with an embodiment, each BS employs representation learning techniques to extract the compressed OCR from each CSI, using a loss function that minimizes the difference between the Euclidean distance of a pair of OCRs and the orthogonality coefficient computed from the corresponding pair of CSI measurements. Instead of transmitting full CSI matrices, each BS compresses the received CSI from the associated UEs into a more compact form that maintains the key relationships between different users' channel conditions. The compression process is guided by the loss function that ensures the similarity between OCRs reflects the actual level of interference between users, as originally determined from CSI measurements. In accordance with an embodiment, the CU 102 uses dedicated signalling to the plurality of BSs to trigger OCR collection at the CU 102, The signalling mechanism is vital for coordinating the retrieval of compressed CSI representations from BSs, ensuring that the CU 102 has the required data to make informed scheduling decisions.

[0104] The steps 202 to 208 are only illustrative, and other alternatives can also be provided where one or more steps are added, or one or more steps are provided in a different sequence without departing from the scope of the claims herein.

[0105] There is provided a computer program product comprising instructions for carrying out all the steps of the method 200 when said computer program is executed on a computer system. The computer program is implemented as an algorithm, embedded in a software stored in the non-transitory computer-readable storage medium having program instructions stored thereon, the program instructions being executable by the one or more processors in the computer system to execute the method 200. The non-transitory computer-readable storage means may include, but are not limited to, an electronic storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. Examples of implementation of computer-readable storage medium, but are not limited to, an Electrically Erasable Programmable Read-Only Memory (EEPROM), a Random Access Memory (RAM), a Read Only Memory (ROM), a Hard Disk Drive (HDD), a Flash memory, a Secure Digital (SD) card, a Solid-State Drive (SSD), a computer-readable storage medium, and / or a CPU cache memory.

[0106] The present disclosure introduces an Al-driven and scalable approach to ICI management and scheduling coordination in multiuser MIMO systems. Unlike the conventional methods that rely on direct CSI feedback, which incurs significant overhead and scalability challenges, the present disclosure leverages OCRs to extract compressed yet interference-relevant CSI features ateach BS. These OCRs are then transmitted to the CU 102 using dedicated signalling, where Al-based translation models (i.e., the plurality of OCR translation models 110) align them into a unified space for efficient scheduling decisions. By replacing high-dimensional CSI feedback with structured OCR representations, the method 200 significantly reduce signalling overhead while preserving critical interference characteristics. This results in enhanced spectral efficiency, improved scheduling accuracy, and reduced computational complexity compared to conventional CSI-based coordination techniques. Additionally, the disclosed Al-driven translation mechanism enables adaptive scheduling that dynamically responds to network conditions, ensuring real-time interference mitigation and optimized resource allocation capabilities that were previously limited in traditional approaches.

[0107] FIG. 3 illustrates an implementation scenario of orthogonality capturing representation (OCR) in a multi-user MEMO system for managing ICI, in accordance with an embodiment of the present disclosure. FIG. 3 is described in conjunction with elements from FIGs. 1 A, IB and 2. With reference to FIG. 3, there is shown an implementation scenario 300 of the OCR in a multi-user MEMO system for managing the ICI. The implementation scenario 300 includes two distinct stages of the OCR-aided scheduling coordination, a first stage is: distributed OCR processing at the BS level (left side) and a second stage is: centralized scheduling coordination at the CU 102 (right side).

[0108] The distributed OCR processing at the BS level depicts that each of the plurality of BSs 104 independently process the CSI received from their connected UEs. Each BS extracts a compressed representation called the OCR from the raw CSI, using AI-driven compression techniques. This process ensures that the interference-related information is retained while significantly reducing data dimensionality. FIG. 3 visually represents the individual BSs computing and maintaining their respective OCRs before transmitting them to the CU 102. The CU 102 is responsible for coordinating scheduling decisions across the plurality of BSs 104. The hexagonal cells in the FIG. 3 symbolize different coverage areas of the BSs, while the arrows indicate the data flow from the plurality of BSs 104 to the CU 102. The visualization highlights the OCR-aided approach distributes the CSI processing at the BS level while enabling efficient, interference-aware scheduling through compressed yet information-rich representations.

[0109] The centralized scheduling coordination at the CU 102 illustrates how the CU 102 utilizes the received OCRs for centralized scheduling decisions. The CU 102 applies Al-based translation models (i.e., the plurality of OCR translation models 110) to map the BS-specific OCRs into the global OCR space, ensuring the compressed representations from different BSs are comparable and interference-aware scheduling decisions can be made. In the global OCR space, the CU 102 evaluates the Euclidean distances between the OCRs to assess user orthogonality and interference potential across the network. The CU 102 then determines the optimal scheduling strategy and sends the corresponding instructions back to the plurality of BSs 104. The centralized scheduling approach leverages the benefits of OCR compression at the BS level while ensuring accurate coordination at the CU 102. Compared to traditional full-dimensional CSI-based methods, the OCR-aided scheduling coordination offers reduced signalling overhead, improved scalability for dense deployments and enhanced scheduling accuracy through orthogonality-based representations. Thus, the OCR-aided scheduling coordination combines distributed CSI compression at the BS level with centralized scheduling at the CU 102. This dual-level approach improves efficiency, reduces data transfer requirements, and optimizes interference management across the network.

[0110] FIG. 4 is a diagram that represents the OCR-aided scheduling approach using translation models at the CU in a scenario involving a plurality of BSs, in accordance with an embodiment of the present disclosure. FIG. 4 is described in conjunction with elements of the FIGs. 1A, IB, 2, and 3. With reference to FIG. 4, there is shown a diagram 400 that represents the OCR-aided scheduling approach using the plurality of OCR translation models 110 at the CU 102 in a scenario involving the plurality of BSs 104. The diagram 400 includes three blocks, such as a first block 402 that represents a global OCR space (Q ). a second block 404 that represents the plurality of OCR translation models 110, processing flow, and scheduling coordination, and a third block 406 that represents communication between the plurality of BSs 104 and the CU 102.Referring to FIG. 4, the first block 402 demonstrates the transformation of local OCRs from multiple base stations (i.e., from a first BS to a / Vth BS) into a unified, standardized global OCR space (ff). Since each BS has a unique channel environment and measurement process, their respective OCRs are not inherently comparable. The global OCR space

[0111]

[0112] serves as the standardized domain where all OCRs from different BSs are mapped. This transformation ensures that OCRs from different BSs are aligned, making them comparable and suitable for centralized scheduling.

[0113] Furthermore, the second block 404 represents the plurality of OCR translation models 110. Each BS has a dedicated Al-based translation model, for example, a translation model represented as fn^g is associated with the nth BS, which is responsible for converting the nth BS OCR representations into a globally recognized format. Each BS has its own OCR translation model fn^g, trained to map its local OCRs into the global OCR space Q. The translation model ensures that differences in local channel measurements, noise conditions, and interference patterns do not affect the comparability of OCRs across BSs. These Al-based models (i.e., the plurality of OCR translation models 110) learn the relationship between local OCR representations and their corresponding placement in the global OCR space. Further, the OCR data from each BS is processed through its respective translation model before being used for scheduling coordination. For the processing, each BS collects OCR data for its connected users {Zn u). The OCR data is sent to the CU 102, which processes it through the corresponding OCR translation model fn^g. The translated OCRs are then mapped into a common format within the global OCR space (ff). The CU 102 performs further computations based on these transformed representations. The final translated OCRs are passed to the scheduling coordination (SC) algorithm for decision-making. The SC algorithm evaluates the orthogonality between users' channels in the global OCR space. It determines which users should be scheduled together to minimize ICI and maximize network efficiency. The CU 102 makes scheduling decisions based on computed distances in the global space, ensuring that users who are orthogonal (non-interfering) are scheduled together.

[0114] The third block 406 demonstrates the communication between the plurality of BSs 104 and the CU 102. Each BS independently collects OCR data and sends it to the CU 102. The CU 102 processes and translates these OCRs before making scheduling decisions. The CU 102 then informs each BS of its scheduling assignments, ensuring that all BSs coordinate their transmissions effectively. The real-time exchange allows for efficient interference management and optimized resource allocation across the network.

[0115] FIG. 5 is a flowchart that depicts a series of operations 502 to 512 performed in OCR-based centralized scheduling coordination, in accordance with an embodiment of the present disclosure. FIG. 5 is described in conjunction with elements of the FIGs. 1 A, IB, 2, 3, and 4.

[0116] At operation 502, the centralized scheduling coordination at the CU 102 begins by receiving OCR data from the plurality of BSs (i.e., from the first BS to the N th BS) and processing the received OCR data to make efficient scheduling decisions. The data includes the OCRs of multiple UEs in set 'U1from the first BS to the OCRs of the UEs in set 11Nfrom the N th BS. The CU 102 acts as the central processing unit, which uses the plurality of OCR translation models 110 (fn-^g) and Scheduling Coordination (SC) algorithm to process the received OCRs for efficient scheduling decisions.

[0117] At operation 504, the CU 102 is configured to transmit a signalling request to each Base Station (i.e., from BS 1 to BS IV). The signaling request specifies the set of UEs (Tl^, ... , 'UN) for which OCRs are required. This signaling request triggers the plurality of BSs 104 to prepare and send the requested OCR data back to the CU 102 for further processing.

[0118] At operation 506, each BS (i.e., from BS 1 to BS N) is configured to retrieve the required OCRs (Zn u) for the UEs within its coverage area. The retrieved OCRs are inherently BS-specific and cannot be directly compared across different BSs. Once OCRs are retrieved, each BS transmits the OCR data ( Zn u]ueu) to the CU 102 for translation and further processing.At operation 508, the CU 102 is configured to process the received OCRs through the respective OCR translation models (f1^g,f2^g> — Each model maps the BS-specific OCRs into a unified global OCR space (Q). ensuring that the OCRs from different BSs become comparable. The translated OCRs ({^gjU}ueK1, ■ ■ ■ , {^g,u}ueztN)arcthen used as standardized inputs for the scheduling coordination algorithm.

[0119] At operation 510, the CU 102 is configured to execute the scheduling coordination (SC) algorithm using the translated OCR data from the global OCR space. The SC algorithm processes the global OCR space (Zgu) to determine optimal scheduling decisions, ensuring efficient interference management and improved network coordination.

[0120] At operation 512, finally, the CU 102 is configured to transmit scheduling instructions back to each BS, guiding each BS on resource allocation and scheduling strategies for their UEs, ensuring effective interference management and optimized network performance. This real-time feedback loop leads to maintain overall system efficiency and scalability.

[0121] FIG. 6 is a flowchart that depicts a series 600 of operations 602 to 612 performed in a training phase of the OCR-aided scheduling coordination, in accordance with an embodiment of the present disclosure. FIG. 6 is described in conjunction with elements of the FIGs. 1A, IB, 2, 3, 4, and 5. The series of operations 602 to 612 is performed to train the plurality of OCR translation models 110 at the CU 102.

[0122] At operation 602, the CU 102 begins by preparing inputs for the SC algorithm and the plurality of OCR translation models 110. The input to SC algorithm includes the OCRs of the UEs in the set 'U1from the first BS to the OCRs of the UEs in set

[0123]

[0124] from the Nth BS.

[0125] At operation 604, the CU 102 is configured to transmit a signalling request to each Base Station (i.e., from BS 1 to BS IV). The signaling request specifies the set of UEs

[0126]

[0127] TlN) for which OCRs and associated labels are required. This signalling request triggers the plurality of BSs 104 to prepare and send the requested OCR data and the associated labels back to the CU 102 for further processing.

[0128] At operation 606, each BS (e.g., from BS 1 and BS N ) is configured to retrieve the OCRs (Zn u) and the corresponding labels (c / ln u) from the associated UEs, such as the global data (e.g., user IDs, timestamps, and global position coordinates). The retrieved OCRs and labels are then transmitted from each BS to the CU 102 for processing.

[0129] At operation 608, the CU 102 is configured to perform label representation matching using the received OCRs and labels. The CU 102 utilizes user IDs and timestamps to match user representations across the plurality of BSs 104, ensuring the formation of paired OCR representations {Zn u,Zm u} for UEs located at the cell edge. These matched representations are required for advanced training of the plurality of OCR translation models 110 to standardize the OCRs into the global OCR space (ff). At operation 610, the CU 102 is configured to train the plurality of OCR translation models 110

[0130]

[0131] using the matched OCRs and labels. The training process optimizes a loss function that minimizes the Euclidean distance between the OCR representations of the same user from different BSs when projected into the global OCR space (ff). For example, the loss function may be expressed in Equation (5)

[0132]

[0133] At operation 612, the CU 102 completes the training of the plurality of OCR translation models 110, enabling them to map BS-specific OCRs into the unified global OCR space (ff) accurately. Once trained, these models facilitate efficient scheduling and interference management in real-time deployment by ensuring comparability of OCR data across different BSs.FIG. 7 is a flowchart 700 that depicts a series of operations performed for the integration of OCR into Wireless Environment State Information (WESI)-AI based air interface architecture, to enable predictive and proactive scheduling, in accordance with an embodiment of the present disclosure. FIG. 7 is described in conjunction with elements of the FIGs. 1A, IB, 2, 3, 4, 5, and 6.

[0134] At operation 702, the WESI property list is configured to include multiple properties, each corresponding to a distinct physical quantity that characterizes the air interface. Among these properties, the OCR is specifically listed as WESI Property k. The OCR property is defined by key parameters, including:

[0135] Representation dimensionality (d^) Defines the number of dimensions used to represent the OCR feature.

[0136] Representation size (in bits): Specifies the bit-length of the OCR representation.

[0137] Space / Time / Frequency span: Determines the range over which the OCR property is defined.

[0138] Translation flag (TFk): Indicates that the OCR property is base station (BS)-specific and requires translation before use in a common representation space.

[0139] At operation 704, the predicted WESI token sequence yL\-f' + P] is generated. The token sequence yL\-f' + P] serves as an input for structured WESI representation learning, enabling the prediction of future properties at a given time step [ + P], This predictive capability allows the CU 102 to anticipate future OCR values and make scheduling decisions in advance. At operation 706, the WESI structured representation is configured to aggregate individual representations of the D properties defined in the WESI property list. Each individual representation captures a specific physical quantity at a given time and frequency granularity.

[0140] At operation 708, the WESI feature 1 representation learning modules are configured to process and generate structured representations for each property.

[0141] At operation 710, the OCR feature, in particular, undergoes a dedicated OCR representation learning step, where the raw OCR data is processed and mapped into a structured form. This structured representation, denoted as Z!k> [T + P] , provides a refined characterization of the OCR property at the future time instance [T + P] .

[0142] At operation 712, the final structured WESI representation Z\ f + P] is generated, consisting of the individual representations [T + P] : Representation of WESI Feature 1.

[0143] Z(kl[T + P] : Representation of the OCR feature.

[0144] Z(l> >f f' + P] : Representation of WESI Feature D.

[0145] These final structured representations serve as inputs for predictive scheduling at the CU 102, enabling proactive resource allocation based on forecasted air interface conditions.

[0146] Modifications to embodiments of the present disclosure described in the foregoing are possible without departing from the scope of the present disclosure as defined by the accompanying claims. Expressions such as "including", "comprising", "incorporating", "have", "is" used to describe and claim the present disclosure are intended to be construed in a non-exclusive manner, namely allowing for items, components or elements not explicitly described also to be present. Reference to thesingular is also to be construed to relate to the plural. The word "exemplary" is used herein to mean "serving as an example, instance or illustration". Any embodiment described as “exemplary” is not necessarily to be construed as preferred or advantageous over other embodiments and / or to exclude the incorporation of features from other embodiments. The word "optionally" is used herein to mean "is provided in some embodiments and not provided in other embodiments". It is appreciated that certain features of the present disclosure, which are, for clarity, described in the context of separate embodiments, may also be provided in combination in a single embodiment. Conversely, various features of the present disclosure, which are, for brevity, described in the context of a single embodiment, may also be provided separately or in any suitable combination or as suitable in any other described embodiment of the disclosure.

Claims

CLAIMS1. A method (200) of managing and reducing Inter-Cell Interference, ICI, between a plurality of Base Stations, BSs (104) in a data communications network, the method (200), being carried out at a Central Unit, CU (102) which coordinates the BS transmissions, comprising steps of:receiving Channel State Information, CSI from each BS, where each BS receives the CSI from each of a plurality of User Equipments, UEs (106), where the CSI is in a compressed, Orthogonality Capturing, Representation, OCR;inputting the received OCR information into a plurality of OCR translation models (110), wherein one OCR translation model is assigned to each BS, wherein each OCR translation model is an Artificial Intelligence, Al based model which, for each respective BS, maps respective OCRs for each UE to a global OCR space; using an output of each OCR translation model to coordinate the scheduling of transmission from each BS; and sending resulting scheduling instructions from the coordination to each BS to thereby manage and reduce ICI.

2. The method (200) of claim 1, wherein, in addition to receiving the OCR information from each BS, the CU (102) also receives labels for training the Al based models at the CU (102).

3. The method (200) of claim 2, wherein the labels include global data.

4. The method (200) of claim 3, wherein the global data includes global position coordinates.

5. The method (200) of claim 3, wherein the global data includes time references.

6. The method (200) of claim 5, wherein the time references are timestamps.

7. The method (200) of claim 3, wherein the global data includes user Identifiers (IDs).

8. The method (200) of claim 1, wherein the OCR is integrated as part of a structured representation of the Wireless Environment State Information, WESI-AI based air interface.

9. The method (200) of claim 1, wherein each BS uses representation learning techniques to extract the compressed OCR from each CSI.

10. The method (200) of claim 9, wherein the compressed OCR has lower dimensionality as compared to the CSI prior to extraction and compression.

11. The method (200) of claim 9, wherein each BS employs representation learning techniques to extract the compressed OCR from each CSI, using a loss function that minimizes the difference between the Euclidean distance of a pair of OCRs and the orthogonality coefficient computed from the corresponding pair of CSI measurements.

12. The method (200) of claim 1, wherein the CU (102) uses dedicated signalling to the plurality of BSs (104) to trigger OCR collection at the CU (102).

13. A system (100) comprising means adapted for carrying out all the steps of the method (200) according to any preceding method claim.

14. A computer program comprising instructions for carrying out all the steps of the method (200) according to any preceding method claim, when said computer program is executed on a computer system.

Images