Digital Twin-Supported Closed-Loop Autonomous RAN Optimization System and Methodology

Abstract

The invention relates to a digital twin-supported closed-loop autonomous RAN optimization system and method used in telecommunications. The main objective of the invention is to provide a digital twin-supported closed-loop autonomous RAN optimization system and method. The invention enables optimization actions to be tested first on the digital twin, and only the validated configuration to be applied in the field. In this way, the best configuration is determined without creating risk in the live network, and active traffic is protected.

Description

1 TARIFF Digital Twin-Supported Closed-Loop Autonomous RAN Optimization System and Methodology Technical Area 5 The invention describes a digital twin-supported closed-loop autonomous RAN (Real-Time Network) system used in the telecommunications field. It relates to optimization systems and methods. State of the Art 10 Current RAN optimization approaches have significant shortcomings. Field Optimization relies largely on expert knowledge and trial and error, and Since parameter changes are tested directly on the live network, failed attempts are activated. This puts subscribers and traffic at risk. Driving tests and coverage measurements are costly and time-consuming. 15 And it is not instantaneous and does not reflect the real-time status of the network. Traditional Self-optimizing network solutions are largely based on fixed rules, while dynamic ones are not. traffic distribution, complex three-dimensional deployment environment, and the multifaceted nature of Massive MIMO architecture. It is not sufficiently adaptable to the wide beam and parameter search space. The current Planning and simulation tools are mostly static and do not synchronize with live network data. 20 And the gap between simulation and real field is not systematically closed. Massive MIMO Because the number of possible combinations of slope, direction, and beam in its architecture is very large, it has to be done manually. The adjustments made do not practically achieve maximum efficiency. Problems are encountered. Creating realistic simulation environments for these regions takes considerable time. This 25 realistic 3D images are created by determining only the center coordinates of the regions. Creating digital twins of environments using open-source map data. Automating the integration of this into the platform has become a necessity in today's technology. It has arrived. In conclusion, due to the negative aspects described above and the current solutions regarding issue 30 Due to its shortcomings, it has become necessary to make improvements in the relevant technical field. Purpose of the Invention The invention was created by drawing inspiration from existing situations and addressing the aforementioned drawbacks. 35 It aims to solve the problem. 2 The main objective of the invention is a digital twin-supported closed-loop autonomous RAN optimization system and The method is to provide Another objective of the invention is to first test optimization actions on the digital twin and The goal is to ensure that only the verified configuration is applied to the field. This allows for live 5 The best configuration is determined without creating risk to the network, and active traffic is preserved. Another aim of the invention is to apply a reinforcement learning approach to the system's adherence to static rules. The goal is to enable it to continuously improve itself according to changing traffic and environmental conditions without delay; The goal is to enable the creation of a low-cost three-dimensional environment using open-source geographic data. 10 Since OpenStreetMap uses building and road data as well as open elevation data, it is subject to a commercial license. dependency decreases; field results are fed back into the model as reward and punishment signals, and The goal is to recalibrate the twin. This allows the difference between simulation and reality to be established. The accuracy decreases over time and increases; human intervention is eliminated due to the autonomous operation of the system. The goal is to minimize the time required for problem solving and shorten the problem-solving process. This situation is expert 15. It reduces dependency and operational costs. To achieve the objectives described above, the invention is a digital twin-supported closed-loop system. It is an autonomous RAN optimization system, and its features include: • Collects live KPI and measurement streams, transforms field problems into standard input, 20 Localization with input including the affected KPI set, problem type, and timestamp. Data collection and problem identification module that triggers the module, • gNB set, problem polygon and user density map affected by measurement data by removing both the area of ​​the twin environment and the neighboring cells to be included in the interference analysis determining spatial localization module, 25 • Creating a three-dimensional model of the site using open geographic and elevation data, and then applying gNB (global Numerical Network) processing its inventory and running a ray tracing-based spread engine for coverage estimation, By identifying coverage gaps and overlapping beams, the most suitable slope, direction, power, and beam are determined. digital twin engine suggesting action, • an offline coverage model trained with extensive historical data and a 30-year time window It includes a reinforcement learning agent updated with live data, with agent actions first performed digitally. experimenting on twin motors and through state space, action space and reward function A machine learning layer that validates the candidate configuration before it is applied to the field. • SON / OAM interface that applies the verified configuration onto the actual gNB, 3 • updating the machine learning layer by converting the field result into reward and penalty signals and by using the difference between the actual measurement and the digital twin engine estimate, the digital twin engine can be determined. recalibrating feedback and calibration module It includes. The invention also includes a digital twin-supported closed-loop autonomous RAN optimization method. Its characteristic is; • Data collection and problem definition module obtains live KPI stream to identify the problem. Converting to standard input and triggering the localization module, • Problem polygon and affected 10 from measurement data by the spatial localization module Defining the area that the twin environment will cover by identifying the neighboring set of cells, • Ray tracing on open geographic data and antenna inventory by digital twin engine Providing a virtual experience environment by generating a coverage estimate, • Detection of coverage gaps and overlapping beams by the machine learning layer By suggesting the most suitable slope, direction, power and beam action, 15 • The action is first verified on the digital twin engine, and only the best performance is selected. Implementing the given configuration in the field via the SON / OAM interface, • Rewards and penalties for before and after KPI comparison by the feedback module by converting it into a signal, both the agent is updated and its digital twin engine is rebuilt. calibrating 20 It includes the steps involved in the process. The structural and characteristic features and all the advantages of the invention are given in the figures below. Thanks to the detailed explanation written with references to the figures, it becomes clearer. This will be understood, and therefore the evaluation should also take these figures and detailed explanations into account. It must be done by taking 25. Figures that will help understand the invention. Figure 1 shows the general architecture of the system that is the subject of the invention. Description of Part References 30 1. Data collection and problem definition module 2. Spatial localization module 3. Digital twin engine 4. Machine learning layer 5.SON / OAM interface 35 6. Feedback and calibration module 4 Detailed Description of the Invention This detailed explanation describes the invention, a digital twin-supported closed-loop autonomous RAN. The optimization system and method preferred structures are not only better at the subject. It is explained in order to facilitate understanding. 5 Digital twin-supported closed-loop autonomous RAN optimization system, data collection and problem solving. identification module (1), spatial localization module (2), digital twin engine (3), machine learning layer (4), SON / OAM interface (5) and feedback and calibration module (6) It includes. 10 The data collection and problem identification module (1) collects live KPI and measurement streams and field It converts the problem into a standard input. This input includes the affected set of KPIs, the problem type, and Includes timestamp. Triggered by KPI threshold exceedance, periodic scan, or manual notification. It can be in this form. 15 The spatial localization module (2) renders the gNB set affected by the measurement data, the problem polygon and by creating a user density map, both the area of ​​the twin environment and the intervention analysis can be performed. It identifies the neighboring cells to be included. Digital twin engine (3), three-dimensional model of the site with open geographic data and elevation data It creates, processes the gNB inventory onto it, and performs ray tracing-based spread estimation for coverage. He starts the engine. The machine learning layer (4) is an offline coverage model trained with extensive historical data 25 and includes a reinforcement learning agent that is updated with live data from the most recent time window. The agent first tests its actions on the digital twin engine (3) and the state space, action space and verifies the candidate configuration via the reward function before it is applied to the field. The SON / OAM interface (5) applies the verified configuration onto the actual gNB. 30 The feedback and calibration module (6) converts the field result into a reward and penalty signal. Updates the machine learning layer (4) and estimates the digital twin engine (3) with actual measurement. It recalibrates the digital twin engine (3) by using the difference between them. Digital twin-supported closed-loop autonomous RAN optimization method, data collection and problem 35 identification module (1), spatial localization module (2), digital twin engine (3), machine learning layer (4), SON / OAM interface (5) and feedback and calibration module (6) It includes. Digital twin-supported closed-loop autonomous RAN optimization method; • by receiving the live KPI stream by the data collection and problem definition module (1) 5 Converting the problem into standard input and triggering the localization module (2), • Problem polygon and from the measurement data by the spatial localization module (2) By identifying the affected neighboring cell set, the area that the twin environment will cover can be defined. • Ray tracing on open geographic data and antenna inventory by digital twin engine (3) Providing a virtual experience environment by generating a coverage estimate, 10 • Coverage gaps and overlapping beams by the machine learning layer (4) By identifying and recommending the most suitable slope, direction, power and beam action, • the action must first be verified on the digital twin engine (3) and only the best performance Implementation of the given configuration to the field via the SON / OAM interface (5), • Reward and penalty comparison of KPIs before and after by feedback module (6) 15 By converting it into a signal, both the agent is updated and the digital twin engine (3) is re-engined. calibrating It includes the steps involved in the process.

Claims

6 REQUESTS 1. It is a digital twin-supported closed-loop autonomous RAN optimization system, the feature of which is; • Collects live KPI and measurement streams, transforms field problems into standard input, Localization 5 with input including affected KPI set, problem type and timestamp. Data collection and problem definition module that triggers the module (1), • gNB set, problem polygon and user density map affected by measurement data by removing both the area of ​​the twin environment and the neighboring cells to be included in the interference analysis determining spatial localization module (2), • Creates a three-dimensional model of the site using open geographic and elevation data, and then applies gNB 10 to it. processing its inventory and running a ray tracing-based spread engine for coverage estimation, By identifying coverage gaps and overlapping beams, the most suitable slope, direction, power, and beam are determined. Digital twin engine (3) that suggests action, • an offline coverage model trained with extensive historical data and in the latest time window A reinforcement learning agent that includes agent actions updated with live data, first digitally 15 experimenting on the twin motor (3) and the state space, action space and reward function machine learning that verifies candidate configuration before it is applied to the field layer (4), • SON / OAM interface (5) which applies the verified configuration onto the actual gNB, • converting the field result into reward and penalty signals, machine learning layer (4) 20 updating and using the difference between the actual measurement and the digital twin engine (3) estimate feedback and calibration module (6) that recalibrates the digital twin motor (3) It includes.

2. It is a digital twin-supported closed-loop autonomous RAN optimization method, and its characteristic is; 25 • by receiving the live KPI stream by the data collection and problem definition module (1) Converting the problem into standard input and triggering the localization module (2), • Problem polygon and from the measurement data by the spatial localization module (2) By identifying the affected neighboring cell set, the area that the twin environment will cover can be defined. • 30 ray tracing on open geographic data and antenna inventory by digital twin engine (3). Providing a virtual experience environment by generating a coverage estimate, • Coverage gaps and overlapping beams by the machine learning layer (4) By identifying and recommending the most suitable slope, direction, power and beam action, • the action must first be verified on the digital twin engine (3) and only the best performance Implementation of the given configuration to the field via the SON / OAM interface (5), 35 7 • Reward and penalty comparison of KPIs before and after by feedback module (6) By converting it into a signal, both the agent is updated and the digital twin engine (3) is re-engined. calibrating It includes the steps of the process.

Images