A SYSTEM AND COMPUTER-IMPLEMENTED METHOD FOR PREVENTIVE AND MOBILITY-AWARE COVERAGE AND CAPACITY OPTIMIZATION MANAGEMENT IN A NEW RADIO NETWORK.

Abstract

The invention relates to a system (100) and method for managing coverage and capacity optimization in new radio networks, integrating artificial intelligence / machine learning-based predictive processes with existing system operations, and possessing preventive and mobility awareness. The invention provides dynamic data detail leveling, spatio-temporal TRP grouping and adaptive decision-making capabilities to improve service continuity and optimize network performance in accordance with 3GPP standards.

Description

1 TARIFF NEW RADIO NETWORK WITH PREVENTIVE AND MOBILITY AWARENESS PROGRAMS A SYSTEM AND A MANAGEMENT FOR COVERAGE AND CAPACITY OPTIMIZATION. A COMPUTER-IMPLEMENTED METHOD 5 TECHNICAL FIELD The invention provides coverage and enhancement to wireless communication systems, and more specifically, new radio networks. A system for managing and predicting capacity optimization problems 10 and relates to methodology. The invention utilizes artificial intelligence and machine learning techniques, artificial intelligence / machine learning. by combining traditional, non-learning-based approaches across the radio access network preventive, context-aware and mobility-sensitive coverage and capacity It provides optimization. PREVIOUS TECHNIQUE The coverage and capacity optimization function optimizes network coverage and capacity configurations. It is used for automatic monitoring and adjustment of coverage and capacity. The optimization function monitors the network's capacity and coverage performance, 20 It automatically detects problems and takes the necessary actions, or This creates a notification for the operator. Classic coverage and capacity optimization, the network Minimizing driving tests in order to monitor and identify problems across the network (MDT), radio link error (RLF), radio condition assessment function (RCEF) reports and correlates various 5G performance metrics. Minimizing driving tests, 25 The 3GPP specifications (e.g., TS 32.421) define the network's connection from user equipment to network equipment. a mechanism that allows the collection of performance metrics during normal operation This indicates that the need for traditional driving tests is reduced. Reports minimizing test activity include signal strength, quality, handover events, and coverage. It includes metrics such as gaps. Radio link error, weak signal of user equipment 30 its quality is critical because it cannot maintain a stable radio link due to interference or mobility. This represents the situation. A radio link error sent from user equipment to the network. reports have unreliable coverage or transmission (handover) errors It is used in identifying problematic areas. Radio condition assessment function. Received Reference Signal Power (RSRP), Received Reference Signal Quality (RSRQ), and Signal-35 Radio conditions are determined based on aggregated data such as Interference and Noise Ratio (SINR). It refers to a functional entity or module responsible for evaluating and performing these analyses. 2 Coverage and Capacity Optimization (CCO), Self-Organizing Network (SON), or upper-level functions such as artificial intelligence / machine learning (AI / ML) inference engines It communicates. The management system aims to correct coverage and capacity optimization problems. For this purpose, downstream channel transmission power, digital azimuth, digital slope, and coverage pattern are adjusted. It configures its parameters. 5 The classical approach involves optimizing the scope and capacity of artificial intelligence / machine learning. Unlike other methods, it is a reactive approach rather than a preventative one. Coverage and Capacity optimization problem identification, coverage and capacity of user equipment. Optimization occurs after experiencing a problem. And resolving the problems takes 10 minutes. After the coverage and capacity optimization problem is identified, continuous monitoring and It is an iterative process that requires updating. Through the use of artificial intelligence / machine learning, earlier detection of future coverage and capacity optimization problems This will be possible and the troubleshooting process will be facilitated through communication between neighboring nodes. It will become more optimized. 15 Current methods are problematic as user equipment experiences coverage gaps or signal issues. After performance declines such as a deterioration in quality occur, these declines are identified. It is based on identifying and responding to problems. Existing solutions are based on identifying problems. For this purpose, we are correlating MDT logs, RLF reports, and performance metrics, 20 This then initiates configuration changes such as power adjustment or antenna tilt. However, these solutions lack predictive capabilities and cannot predict future network behavior or They do not take into account intercellular motility patterns. Furthermore, they are dependent on continuous monitoring. Because of this, they lead to inefficiencies in signaling and resource allocation. WO2023197245A1 uses an inference engine embedded in a network device to determine coverage and capacity. an AI / machine learning-based scope that predicts optimization problems and It explains the capacity optimization mechanism. This approach offers predictive capabilities. yes, primarily focusing on local cell optimization, artificial intelligence / machine It does not address the operational context in which learning is used. Artificial intelligence / machine learning 30 when to trust predictions and when to revert to current methods It lacks a mechanism for determining. Furthermore, the predicted problems affect neighboring cells. it does not take into account how it affects and the spatial relationship between transmission and reception points (TRPs). It does not implement a temporal coordination strategy. 35 WO2024109111A1, preventive re-enactment based on anticipated coverage changes. an AI / machine learning-powered radio access system to enable configuration. 3 This invention proposes a network (RAN) node collaboration scheme. This invention enables predictive coverage and While accepting capacity optimization, the priority is on individual cells and beams. optimizing their configurations and a broader spatio-temporal neighborhood It does not include a relationship model. Furthermore, it does not address the feedback on AI / machine learning performance. a system-level life cycle for adaptation based on feeding 5 It does not provide a management mechanism and allows for dynamic level of detail selection or hybrid operation. It does not support it. CN117837193A is a machine trained with data obtained from a specific cell and its neighbors. Channel initiative 10 by calculating initiative indicators using learning models It addresses the prediction. Although this solution uses data from surrounding nodes, the focus is on... It is limited by channel initiative and comprehensive issues such as capacity shortages or coverage gaps. It does not extend to coverage and capacity optimization metrics. This solution focuses on transmission and reception points. spatio-temporal grouping and coordinated coverage and capacity optimization decision He lacks the ability to receive. 15 WO2024028041A1, for artificial intelligence / machine learning training in the context of radio access networks. MDT relates to data collection methods. While this structure supports the training infrastructure, it also covers the scope. and online, real-time forecasting and decision-making for capacity optimization. It does not offer a solution to this issue. This invention primarily focuses on other uses of artificial intelligence. It is a provider of capabilities for various scenarios and a functional system that will perform preventive problem solving. It lacks modules. In summary, the technical problems related to the known state of the art are: artificial intelligence / machine learning. lack of operational integration between existing systems; trust level and 25 mechanisms to determine the applicability of artificial intelligence / machine learning based on context The lack of spatial coordination for managing neighboring transmission and reception points in a coordinated manner. the absence of temporal mapping; and the continuous improvement of models The absence of a lifecycle feedback architecture to provide this. These shortcomings, performance inefficiencies, high signal load, poor scalability and 30 This leads to sensitivity to inaccuracies in predictions. In conclusion, all the problems mentioned above necessitate an innovation in the relevant technical field. It has brought it to this state. 35 A BRIEF DESCRIPTION OF THE INVENTION 4 The present invention aims to eliminate the aforementioned disadvantages and introduce new technologies to the relevant technical field. to bring advantages by means of a system and a method implemented by a computer It is related. This invention predicts coverage and capacity optimization problems proactively, 5 Neighboring transmission affected by spatio-temporal analysis supported by mobility information and Defining acquisition points, dynamically selecting model complexity and input level of detail, implementing hybrid mitigation strategies and based on observed performance outputs by offering a modular structure that retrains artificial intelligence / machine learning models It addresses these problems. This integrated and adaptable approach will shape the new 10, modern and future. Smart, scalable, and standards-compliant coverage and capacity for radio applications. It provides optimization management. The invention aims to provide AI / machine learning-based predictions for existing system processes and By integrating with spatio-temporal analysis, smart, scalable and 15 in New Radio networks Preventive and mobility for coverage and capacity optimization in accordance with standards. It is about providing a system with awareness and a method implemented by a computer. All the purposes mentioned above and those that will emerge from the detailed explanation below. The present invention aims to optimize coverage and capacity in new radio networks. It relates to a system for management and a method implemented by a computer. Specifically, this invention offers an advantage over existing decision-making models based on artificial intelligence / machine learning. a management framework that integrates with logic, is preventive and has mobility awareness. The invention also provides spatio-temporal analysis, dynamic level of detail adjustment, This includes adaptive mitigation strategies and trust-based operational transition, 3GPP 25 It provides intelligent and scalable network optimization in compliance with standards. According to the invention, it is for managing coverage and capacity optimization in the new radio network. a system is provided and the system that is the subject of the invention includes the following components; at least local and neighboring transmission and reception points, user equipment and self-contained 30 Collecting network data from organizing network platforms and time-domain alignment, formatting by applying normalization, relevance filtering, and data integration to create a unified system. A data collection and preprocessing module structured to create a measurement repository; Based on the aforementioned data repository, using one or more trained models a 35 structured system designed to predict coverage and capacity optimization problems artificial intelligence / machine learning inference module; neighboring transmission and reception points that may be affected by the anticipated problems, using historical mobility routes, transfer statistics, and temporal correlation analysis a spatio-temporal analysis and TRP grouping module structured for identification; appropriate complexity depending on prediction confidence and the availability of system resources. A dynamic level-of-detail 5 system is configured to select a predictive model at the data level of detail. adjustment module; and Reconstruction commands based on prediction results and spatio-temporal analysis. or a decision and response management module structured to generate feedback responses. The system in question allows for predictable and efficient optimization of coverage and capacity in 5G networks. It enables scalable management, eliminating the need for post-event diagnosis. It eliminates the need for real-time data collection, AI / machine learning inference, and spatial analysis. By combining temporal analysis and dynamic response orchestration, for preventive mitigation. It provides a comprehensive control loop through adaptive level of detail adjustment. It achieves high prediction accuracy while conserving computational resources; this is reflected in signaling. with a significant reduction in load, increased transmission efficiency and decreased call rates 15 The solution is modular and standards-compliant, and is suitable for both existing and next-generation networks. This facilitates seamless integration into their architectures. In one application of the invention, the data collection and preprocessing module uses irrelevant elements during preprocessing. Relevance-based filtering to exclude neighboring transmission and reception points 20 It is also characterized by being structured in a way that will enable it to perform this filtering. The mechanism involves neighboring nodes that are less likely to be affected by the predicted problem. By eliminating it, it reduces the amount of feedback required from neighboring nodes; thus It reduces communication burden and transaction complexity. Scope and capacity optimization actions carried out in an application of the invention monitoring the outputs and the relationship between predicted network behavior and actual network behavior based on inconsistencies in model parameters, retraining intervals, or spatial- A lifecycle feedback structured for updating temporal weights. It is characterized by the fact that the system also includes a management module. Lifecycle back 30 The feeding management module uses continuous learning and modeling based on observed inconsistencies. By enabling adaptation, it improves forecast accuracy over time and in the long term. This reduces performance degradation. This results in lower operational slippage. This leads to a sustainable AI / machine learning integration and optimization. This increases the reliability and robustness of the process. 35 6 In one application of the invention, the lifecycle feedback management module, version including number, retraining timestamp, confidence limits, and drift statistics. It is characterized by maintaining model metadata. Metadata tracking is transparent and It supports controllable artificial intelligence / machine learning processes, explainability. increasing and complying with telecommunications industry standards or regulatory requirements 5 It also provides debugging and model rollback operations when needed. It makes things easier. In one application of the invention, the artificial intelligence / machine learning inference module will, at least periodic timing, threshold-based anomaly detection, or 10 generated by existing systems by being structured in such a way as to trigger inference based on one of the warnings It is characterized by its ability to support multiple triggering mechanisms. both its sensitivity and adaptability to gradual and sudden changes in network conditions. This increases the likelihood of timely detection and resolution of optimization problems. In one application of the invention, the scope of the artificial intelligence / machine learning inference module Predicted problem selected from gaps, cell edge interference and capacity deficiencies. types, along with the estimated timestamp and affected transmission and reception points It is characterized by providing output. The predicted problems are semantically... classification enables the system to perform targeted restructuring actions 20 This ensures that generic solutions are avoided and resource utilization is improved. It is optimized for efficient and context-aware network management. In one application of the invention, the spatio-temporal analysis and TRP grouping module can be used to group neighboring objects. By calculating the directional effect between the transmission and reception points, a spatio-temporal effect map is created. 25 It is characterized by its production. Directional impact-based impact mapping, preventive and By enabling neighborhood-sensitive mitigation, we can prevent cascading performance degradations. This reduces the need for unnecessary interventions in unaffected areas. In one application of the invention, the dynamic detail adjustment module adjusts the L1 layer beam 30 AI / machine learning models trained with metrics and L3 layer coverage summaries It is characterized by its ability to choose from among many options. It possesses an awareness of the level of detail. Model selection is an operational link between the system's prediction accuracy and computational efficiency. It ensures a balance depending on the constraints; this means high load and low resources. It enables performance improvement even under these conditions. 35 7 In one application of the invention, the dynamic level of detail adjustment module has low predictive reliability. If detected, it initiates a switch to a model with a finer level of detail. This is characterized by the shift to more detailed models in uncertain scenarios. the reliability of predictions without resorting to high-cost models by default and This ensures its accuracy; which, under normal conditions, allows for the conservation of transaction resources. 5 He knows. In one application of the invention, the decision and response management module has at least the transmission power. adjustment of digital azimuth change, digital tilt adjustment or beam shaping profile It must be configured to transmit reconfiguration commands that include modifications. 10 It is characterized by direct control over RF parameters, detected. Improving signal quality and user experience by addressing coverage and capacity deficiencies. It offers an effective method for real-time physical layer removal for this purpose. In one application of the invention, the decision and response management module affects transmission and reception. via standardized self-organizing network signal interfaces to its points It is characterized by broadcasting early warning messages. Standard early warnings, It enables seamless coordination between multiple cells and the predicted By enabling operators to respond quickly to disruption events, service continuity is ensured. The system increases transparency. 20 In one application of the invention, the data collection and preprocessing module is self-organizing. by receiving MDT logs, RLF reports, and RCEF reports from the network platform It is characterized by the use of self-organizing network platform data. Increasing the richness of input information and providing an additional 25 for more accurate and robust predictive modeling. It provides key performance indicators, thus addressing the need for comprehensive active inquiry. is being reduced to a minimum. In one application of the invention, each module is structured with data packets linked to subsequent modules. 30 by communicating using and maintaining internal logs for version control and traceability This is characterized by structured communication between modules, and the modularity of the system. This ensures traceability and sustainability; this includes version updates, bug fixes. debugging and standards-compliant integration into operator network management systems It makes things easier. 35 In one application of the invention, the management entity is compliant with 3GPP TR 38.743 and TR 38.843. by being implemented as a centralized or distributed network function that can operate in this way 8 It is characterized by flexible application options and compliance with standards. It supports scalability and includes both edge and core environments. It enables integration into various network topologies, including existing and future 5G / 6G networks. It is compatible with evolutionary pathways. According to another aspect of the invention, coverage and capacity optimization in the new radio network A computer-implemented method for its management is provided, as mentioned. The method implemented by the computer includes the following steps: at least local and neighboring transmission and reception points, user equipment and self-contained Collection and preprocessing of network data from organizing network platforms; 10 Coverage and capacity using at least one trained artificial intelligence / machine learning model Estimation of optimization problems; neighboring transmission and reception points that may be affected by the predicted problem Identification using spatio-temporal analysis; Model complexity and data detail depend on the confidence of the forecast and the available resources. adjusting the level; and Generating and transmitting reconstruction commands or feedback responses The method is a fully automated, artificial intelligence / machine learning system that enables preventive network control. It defines a supported optimization pipeline. This method improves response time, 20 minimizing delays in problem solving, both in densely populated urban areas and It provides adaptive management in busy environments such as highways. Model usage and action. By optimizing the scope, in addition to improved service quality, significant energy and bandwidth savings were achieved. Width savings are offered. In one application of the invention, the method involves the transmission and reception of unrelated neighbors during preprocessing. It also includes a relevance-based filtering step to exclude certain points. This filtering mechanism is characterized by its susceptibility to the predicted problem. By eliminating neighboring nodes with a low probability of being picked up, the required feedback from neighboring nodes is obtained. It reduces the amount of data transmitted; thus lowering the communication load and processing complexity. 30 In one application of the invention, the method relates to the predicted and observed behavior. model parameters or retraining intervals based on comparison It is characterized by also including the step of updating and monitoring real network outputs. is being done. Including real-world feedback makes predictive models more sustainable in the long term. 35 by ensuring the long-term accuracy and adaptability of the system to changing users 9 to become resilient to their behavior, network configurations, or environmental conditions It brings. In one application of the invention, model feedback tracks data drift and the model confidence limit. It is characterized by including the adjustment or retraining timing. 5 Feedback elements such as drift tracking and adaptive retraining, over time lifecycle integrity and performance of artificial intelligence / machine learning models This increases its stability, which reduces the need for manual maintenance and minimizes performance degradation. It prevents it. In one application of the invention, coverage gaps in the prediction step cause cell edge interference. It is characterized by involving the identification of problem types such as capacity deficiencies. This involves categorizing the forecast outputs, differentiating them, and identifying problem-specific solutions. This ensures that corrective actions are taken, which increases efficiency and service. It reduces interruptions and aligns better with key performance indicators. 15 In one application of the invention, the spatio-temporal analysis step involves historical transmission patterns and This includes calculating a weighted neighborhood graph based on movement traces. It is characterized by the analysis of spatio-temporal dependencies, synchronous and a context-aware 20 that enables location-aware interventions. It creates a grouping mechanism; this increases the durability of transfer (handing over) and It reduces unnecessary configuration efforts. In one application of the invention, the adjustment step is performed using L1-based and L3-based artificial intelligence / machine tools. The learning process is characterized by involving a choice between inference pipelines. 25 Model transitions based on data level detail and system state are performed, ensuring accuracy and processing power. By balancing the load, the method works in real-time under various network scenarios. This makes it efficient for applications. BRIEF DESCRIPTION OF THE FIGURE 30 Figure 1 shows a representative view of the system that is the subject of the invention. Figure 2 provides a representative view of the method implemented by the computer. 35 REFERENCE NUMBERS GIVEN IN THE FIGURE 100. System 200. Data collection and preprocessing module 300. Artificial intelligence / machine learning inference module 400. Spatial-temporal analysis and TRP grouping module 500. Dynamic detail level adjustment module 5 600. Decision and response management module 700. Lifecycle feedback management module LTRP. Local transmission and reception point. NTRP. Adjacent transmission and reception point. UE. User equipment 10 THE END. A self-organizing network platform. DETAILED DESCRIPTION OF THE INVENTION This detailed explanation of the invention does not include any information solely for the purpose of better understanding the subject. This is explained with examples that will not create a limiting effect. The invention is particularly relevant to the management of coverage and capacity optimization in new radio networks. a system (100) with mobility awareness, preventive and mobility awareness and by a computer It relates to a method applied. The system (100) is implemented within a management entity, 20 AI / machine learning-based prediction models, AI / machine learning-based It integrates with existing coverage and capacity optimization processes that are not currently available. The system (100) to provide scalable, context-aware and spatial-sensitive network optimization. temporal analysis, dynamic level of detail adjustment, decision lifecycle management, and It also includes adaptive response mechanisms. 25 The system (100) includes the following main modules: a data collection and preprocessing module (200), an artificial intelligence / machine learning inference module (300), a spatio-temporal analysis and TRP grouping module (400), a dynamic level of detail adjustment module (500), a decision and Response management module (600). The system (100) also has a lifecycle feedback management 30 It also includes module (700). These modules (200, 300, 400, 500, 600, 700) form the new radio network. within a centralized or distributed management entity that is operational The system (100) is located at local transmission and reception points (LTRP), neighboring transmission and Points of arrival (NTRP), user equipment (UE), and self-organizing network platforms It interfaces with multiple network components, including (SON). Modules (200, 35 (300, 400, 500, 600, 700), obtain network measurement data, and derive feasible predictions, Analyzing neighborhood impact, defining operational strategy, coverage and capacity. 11 To implement and evaluate the results of the optimization process, sequential and They interact in a collaborative manner. Data collection and preprocessing module (200) collects raw network data from at least four main sources. It is structured to combine: local transmission and reception points (LTRP), neighboring transmission and 5 Points of arrival (NTRP), user equipment (UE), and self-organizing network platforms (END). Local Transmission and Reception Point (LTRP) data, real-time radio source status, Coverage and capacity optimization status metrics, and optionally at the beam level. It may include performance data. Neighboring transmission and reception points (NTRP), transmission. It provides statistics, mobility event logs, and inter-cell interaction metrics. 10 User equipment (UE) based data includes Received Reference Signal Strength (RSRP), Received Reference Signal Quality (RSRQ) and Signal-to-Interference and Noise Ratio (SINR) measurements. It consists of mobility or transfer event reports. Self-organizing network. platform (SON), driving test minimization (MDT), radio link failure (RLF), radio Condition Evaluation Function (RCEF) reports and 15 in 3GPP TS 28.313 and TS 28.552. Like other defined 5G performance metrics, it contributes to previously aggregated indicators. It includes the data collection and preprocessing module (200), time domain alignment, format. normalization, relevance filtering, and a unified repository of data measurements. It is configured to perform preprocessing tasks such as integration. This repository, The downstream AI / machine learning operating mode serves as input for 20. It provides. Artificial intelligence / machine learning inference module (300), coverage and capacity optimization to predict problems before they arise at the user equipment (UE) level It is structured. Artificial intelligence / machine learning inference module (300), past and real 25 Future coverage and capacity optimization based on real-time datasets. multiple groups of varying complexity trained to infer problems It includes artificial intelligence / machine learning models. These models address coverage gaps, Problem types such as cell edge interference or capacity deficiencies, and predictions related to them. The output, along with the transmission and reception points or beams affected by the acquired timestamp, is 30. It can provide. Inference is triggered through three different mechanisms: (i) periodic artificial intelligence / machine learning timing, (ii) threshold-based anomaly detection, or (iii) existing system responding to warnings. For example, the AI / machine learning inference module (300), cell If it detects early signs of capacity reduction at the edge, this indicates high confidence. It can trigger an AI / machine learning replicate of a certain level. Artificial 35 Intelligence / Machine Learning Inference Module (300), also the confidence level of its predictions and structured to assess reliability, where confidence thresholds are not met. 12 in these situations, whether to proceed with AI / machine learning-based decisions or It dynamically determines whether or not to revert to the current decision-making logic. This The situation provides a robust recovery mechanism under uncertainty and prevents erroneous re-entry. It prevents these configurations from happening. Spatial-temporal analysis and TRP grouping module (400), probability of being affected by the same problem Artificial intelligence / machine learning to identify the secondary transmission and reception points found. It is structured to work with the prediction results. This spatio-temporal analysis and TRP grouping module (400) to determine the directional effect between transmission and reception points. It uses transfer definitions and past mobility paths. Statistically significant 10 Relevance-based filters are applied to exclude non-existent neighbors. Past coverage and capacity optimization problems over time intervals Spread trends are determined by analyzing temporal correlation. Additionally, mobility is considered. role reversals between transmission and reception points resulting from connected topology changes A multi-scenario impact analysis is also being conducted to determine this. This spatial-15 The output of the temporal analysis and TRP grouping module (400) is a spatio-temporal effect map. It is a group of high-risk transmission and reception points labeled with impact timelines. For example, if the first transmission and reception point experiences capacity shortage within 10 minutes. if estimated and the second transmission and reception point was under similar conditions to the first in the past If it is known to receive traffic overflowing from the first point, the second transmission and reception point is designated as a high-risk preventive measure. They are added to the groups. Only statistically significant adjacent transmission and reception points. Thanks to its targeted approach, the invention is an artificial intelligence / machine learning-based self-organizing system. It minimizes signaling load, a common problem for network applications. Dynamic level of detail adjustment module (500), AI / machine learning inference module 25 The level of measurement detail to be used in the inference processes carried out by (300) and It is configured to manage model complexity. Dynamic level of detail adjustment. module (500), datasets with different levels of detail, e.g., L1 layer beamforming an AI / machine learning model trained with metrics or L3 layer coverage summaries It selects from its library. When higher prediction accuracy is required or the previous 30 When the confidence level of the model run is low, artificial intelligence / machine learning inference Module (300) can restart inference with L1-based models. Dynamic detail The decisions of the level adjustment module (500) use available computing resources and delay taking into account the requirements, either increasing the model complexity accordingly or It lowers. 35 13 Decision and response management module (600), artificial intelligence / machine learning and spatio-temporal It is structured to interpret the analysis outputs and define remedial actions. When the prediction confidence is high, this decision and response management module (600) selects the transmission and They must initiate the restructuring processes at the designated timestamps at the points of purchase. It can give instructions. Reconfiguration actions; transmission power, digital azimuth, 5 This may include adjusting digital slope or beamforming profiles. Confidence If the level is low or the system load is limited, the decision and response management module (600) is available. This allows for a return to coverage and capacity optimization processes. These early warnings, To ensure compatibility with the existing network infrastructure, a standardized self-contained system is used. The organizing network is propagated using signal interfaces. 10 Lifecycle feedback management module (700), monitoring and adaptive learning It is structured to fulfill its tasks. The scope and capacity applied. recording the results of optimization actions and comparing this mitigation with the predicted effect. It evaluates whether they are consistent. If inconsistencies are observed, 15 The lifecycle feedback management module (700) updates the model parameters, It improves spatio-temporal weights or triggers the retraining process. Lifecycle feedback management module (700), data drift, retraining intervals and artificial intelligence / machine learning by enabling the tracking of operational effectiveness over time. It also supports lifecycle management. Each model; last retraining time 20 with metadata such as model version, confidence limits, and slip logs It is related. Lifecycle feedback management module (700), artificial intelligence / machine Learning models evolve in response to changing network conditions and topology differences. It ensures that he / she passes the test. The internal workflow of System (100) reflects the modular and adaptable nature of the framework. The process, data collection and preprocessing module (200) combines raw measurements from various sources an aligned, normalized, and time-synchronized data repository It is initiated by creating this repository. This repository is used for system scheduling, threshold triggers, or anomalies. 30 that selectively invoke one or more pre-trained models based on events Accessed by the AI / machine learning inference module (300). Possible coverage or inferences including estimates of capacity disruptions, mobility data, transmission assessing neighborhood-level influence using patterns and past transit logs The results of the spatio-temporal analysis and TRP grouping are transferred to the module (400). then, 35 determines the optimal resolution and model complexity for the final decision logic. It is transferred to the dynamic detail level adjustment module (500). If necessary, further Re-inferences are made using highly detailed data. Decision and response. 14 The management module (600) combines all previous outputs, evaluates the prediction confidence and ranging from complete restructuring to rollback logic or early warning deployment. It selects the elimination strategies. Lifecycle feedback management module (700), system monitoring the results across the board, keeping a record of the outcomes of all actions taken. and updates the model parameters and operational policies accordingly. Each module has 5 (200, 300, 400, 500, 600, 700), via structured data packets in the next module It communicates and maintains internal logs for version control and traceability. This is holistic. data and control flow, the system is autonomous under (100) different network conditions and topologies, It enables adaptive operation and real-time performance. This invention also contributes to the management of coverage and capacity optimization in new radio networks. It relates to a computer application method (100). The said method is as follows: It includes the following steps: Collection and preprocessing of network data; Optimizing coverage and capacity using AI / machine learning inference 15 Predicting problems, Spatial-temporal analysis to group relevant transmission and reception points, Adjusting the model level of detail depending on the context and Managing the response to outbursts. The subject of the invention is the computer-implemented method (100), elimination responses after management, the step of executing lifecycle feedback processes is also included. It also includes. The invention concerns a computer-implemented method, which initially involves data collection and preliminary processing. at least local transmission and reception points (LTRP), neighboring transmission by the processing module (200). and points of reception (NTRP), user equipment (UE), and self-organizing network These can be collected from various platforms (SON). RSRP, SINR, MDT logs, beamforming metrics and Data such as transfer statistics, filtering, normalization, and time alignment techniques It is pre-processed using and a combined input structure is created. Then, 30 one or more artificial intelligence / machine learning models, upcoming coverage and capacity AI / machine learning inference module for inferring optimization problems It is operated by (300). The artificial intelligence / machine learning inference module (300), each It determines the type, location, and timing of a potential problem. For example, a particular A coverage gap will occur at the transmission and reception point within the next 5 minutes. 35 And it can be predicted that the transmission load will increase at adjacent transmission and reception points. Then, spatio-temporal analysis and TRP grouping module (400) were performed. Spatial-temporal analysis, the effect of the predicted problem across the neighborhood. It maps the spatial-temporal patterns using movement traces and past transitions. Analysis and TRP grouping module (400), weighted relationships between transmission and reception points calculating the graph and determining which nodes should be grouped for coordinated mitigation. It defines. Then, the most suitable artificial intelligence / machine learning model and data details 5 The level can be selected by the dynamic detail level adjustment module (500). Low If trust is detected, the dynamic level of detail adjustment module (500) is L1 based. It is possible to switch to an extraction pipeline. In contrast, where resources are limited dynamic detail level adjustment module (500) in cases with L3 data and a lightweight model It can continue to process. Following this step, the decision and response management module (600) 10 It transmits intervention commands or warnings. In situations with a high level of confidence, Re-engineering to optimize coverage and capacity at affected transmission and reception points. Configuration messages are sent. In fallback scenarios, decision and response management takes place. module (600) activates the existing response mechanisms by default and handles the incident more It labels them for advanced AI / machine learning observation. Finally, the system 15 Feedback is collected by the lifecycle feedback management module (700). The applied responses are compared with actual network performance results. The models are then used accordingly. They are retrained or adjusted accordingly, and long-term operational parameters, For example, retraining frequency and model confidence intervals are recalibrated. Technical impacts and advantages of the invention include scope and capacity optimization. This includes the preventive identification and resolution of problems, and post-event recovery. It reduces dependence on existing processes. The invention predetermines the relevant transmission and reception points. By defining it, it minimizes unnecessary signaling. Through dynamic model selection ensuring efficient use of computing resources, returning to existing systems 25 It offers backward compatibility through its logic and the life of artificial intelligence / machine learning. It supports cycle adaptation. Spatial-temporal impact modeling and reliability. This invention, which incorporates model management based on intelligence, is intelligent, context-aware, and It provides scalable coverage and capacity optimization processes. Additionally, the system... (100), labor-intensive driving by utilizing user equipment mobility data 30 minimizing the need for tests and avoiding random alerts that cover the entire network. By avoiding this, it optimizes the signal loads of the self-organizing network. In this respect, Commercially viable for next-generation AI / machine learning-based network automation. It offers a solution. In general, this invention improves service continuity and reduces dropped calls. reducing transfer rates, optimizing handover processes, and implementing 6G architectures. It supports intelligent and scalable network management in line with proper evolution. 16 The scope of protection of the invention is set out in the attached claims and is not necessarily detailed in this description. It cannot be limited to what has been described for illustrative purposes. Because a person who is an expert in the field can explain the invention... Without deviating from the main theme, similar structures emerge in light of what has been described above. It is clear that he can place it.

Claims

17 REQUESTS 1. A plan to manage coverage and capacity optimization in the new radio network. A method implemented by a computer (100) and its feature is; the following steps 5 It is characterized by its inclusion of: at least local and neighboring transmission and reception points, user equipment and self-contained Collection and preprocessing of network data from organizing network platforms eclipse; coverage and 10 using at least one trained artificial intelligence / machine learning model. Predicting capacity optimization problems; neighboring transmission and reception points that may be affected by the predicted problem Determining this using spatio-temporal analysis; prediction confidence and model complexity and data detail according to available resources. adjusting the level; and 15 Generating and transmitting reconfiguration commands or feedback responses.

2. According to claim 1, a method (100) has the characteristic of; irrelevant neighbor transmission during preprocessing. and relevance-based filtering to exclude points of purchase. It is characterized by its ability to accomplish. 20 3. A method (100) according to claim 1 or 2, and its feature is; also the actual network results based on monitoring and comparing predicted behavior with observed behavior. as the step of updating model parameters or retraining intervals It is characterized by its inclusion. 25 4. A method (100) according to claim 3 has the feature of; model feedback data drift This includes monitoring, adjusting the model confidence limit, or scheduling retraining. It is characterized by...

5. A method (100) according to any of the previous requests and its feature is; estimation at least the step of coverage gaps, cell edge interference or capacity This involves identifying a type of problem related to at least one of their deficiencies. It is the characterization of the situation. 35 6. A method (100) according to any of the previous requests, and its feature is spatial- the temporal analysis step based on past transmission patterns and mobility traces It is characterized by the fact that it involves a weighted neighborhood graph calculation. 18 7. A method (100) according to any of the previous requests and its feature is; adjustment step's L1-based and L3-based AI / machine learning inference pipelines It is characterized by the fact that it involves making a choice between two options.

8. A plan to manage coverage and capacity optimization in the new radio network. The system (100) is characterized by its inclusion of the following: at least local transmission and reception points (LTRP), neighboring transmission and reception points (NTRP) is a data collection and preprocessing system configured to collect network data. module (200); user equipment (UE) and self-organizing network platforms 10 Time-based alignment, shape normalization, and relevance with (SON). Creating a unified metrics repository by applying filtering and data integration. in order to; based on the relevant measurement database using one or more trained models. 15 to predict coverage and capacity optimization problems. a structured AI / machine learning inference module (300); Adjacent transmission and reception that may be affected by the anticipated problems points, past mobility routes, transfer statistics and temporal correlation a spatio-temporal analysis structured to define using analysis and TRP grouping module (400); 20 appropriate depending on prediction confidence and the availability of system resources structured to select a predictive model at a given level of complexity and data detail. a dynamic detail level adjustment module (500); and Reconstruction based on estimation results and spatio-temporal analysis. a decision and response structured to generate commands or feedback responses 25 management module (600).

9. A system (100) according to claim 8, and its feature is; data collection and preprocessing module. (200), to exclude irrelevant neighboring transmission and reception points during preprocessing It should also be configured to perform relevance-based filtering. 30 It is characterized by...

10. A system (100) according to claim 8 or 9, and its characteristic is the scope and capacity of operation. by monitoring the outputs of optimization actions and predicting network behavior. Based on inconsistencies between the model parameters and actual network behavior, 35 for updating retraining intervals or spatio-temporal weights 19 by including a structured lifecycle feedback management module (700) It is the characterization of the situation.

11. According to claim 10, a system (100) has the feature of life cycle feedback management. module (700), version number, retraining timestamp, confidence limits and 5 It is characterized by holding model metadata, including drift statistics. It is done.

12. A system (100) according to any of claims 8 to 11, and its characteristic is; artificial intelligence / machine learning inference module (300), at least periodic scheduling, 10 threshold-based anomaly detection or one of the alerts generated by older systems It is characterized by being structured in a way that will trigger inference based on it. It is done.

13. A system (100) according to any of claims 8 to 12, and its feature is; artificial 15 intelligence / machine learning inference module (300), coverage gaps, cell Estimated problem types selected from edge interference and capacity deficiencies, along with the estimated timestamp and the affected transmission and reception points. It is characterized by producing something as output.

14. A system (100) according to any of claims 8 to 13, and its property is spatial- Temporal analysis and TRP grouping module (400), neighboring transmission and reception points by calculating the directional influence between them and generating a spatio-temporal influence map. It is the characterization of the situation.

15. A system (100) according to any of claims 8 to 14, and its feature is; dynamic detail level adjustment module (500), L1 layer beam measurements and L3 layer Selection from AI / machine learning models trained using coverage summaries. It is characterized by its ability to do things.

16. A system (100) according to any of claims 8 to 15, and its feature is; dynamic detail if low prediction confidence is detected in the level adjustment module (500) It is characterized by initiating a transition to a model with a finer level of detail.

17. A system (100) according to any of claims 8 to 16, its characteristic is; decision and response 35 management module (600), at least transmission power adjustment, digital azimuth change, This includes digital tilt adjustment or changing the beam shaping profile. It is characterized by being configured to transmit configuration commands. It is done.

18. A system (100) according to any of claims 8 to 17, its characteristic being; decision and response The management module (600) has its own standardized 5 for the affected transmission and reception points. by broadcasting early warning messages via self-organizing network signal interfaces It is the characterization of the situation.

19. A system (100) according to any of claims 8 to 18, whose characteristic is; data collection and preprocessing module (200), from the self-organizing network platform MDT 10 It is characterized by receiving logs, RLF reports, and RCEF reports.

20. A system (100) according to any of 8 to 19 of the requirements; the feature of each module It communicates with subsequent modules using structured data packets and releases It is characterized by keeping internal logs for control and traceability purposes. 15

Images