Adaptive Cell Reselection Priority Optimization System and Method

Abstract

The invention relates to an adaptive cell reselection priority optimization system and method that dynamically determines the idle cell reselection priorities of user equipment in LTE and heterogeneous mobile networks based on real-time network conditions and user context; the system collects radio measurements with the user equipment monitoring interface (1), converts them into features with the data acquisition and preprocessing module (2), generates context, mobility and load information with the context analysis engine (3), mobility estimation module (4) and network load monitoring unit (5), learns from past behaviors via the historical data repository (6) and machine learning engine (7), generates user-specific reselection parameters with the priority optimization algorithm (8) and priority parameter generator (9), transmits the parameters to the user equipment via the RRC signaling interface (10) and provides feedback and network coordination via the performance monitoring system (11) and SON integration interface (12).

Description

1 TARIFF Adaptive Cell Reselection Priority Optimization System and Method TECHNICAL AREA 5 The invention relates to an adaptive cell reselection priority optimization system and method. The invention is particularly relevant for LTE mobile networks and for the combined use of macro, micro, pico, and femto cells. In heterogeneous network topologies, user equipment is idle in cell 10. re-election priorities depend on real-time network conditions, user context, and mobility. User-specific solutions are created taking into account cell load, signal quality, and past selection behaviors. It is related to how it is determined. PREVIOUS TECHNIQUE 15 In current LTE mobile networks, this is what user equipment does while idle. In cell re-selection processes, cell priorities are usually determined via system information blocks. It is based on the published cellReselectionPriority values. These values ​​are related to the network. It is manually configured by the operator and user 20 in the same broadcast area. This is applied jointly to their equipment. This static approach, cell reselection traffic load, radio signal conditions, and the user's status can change at the time the decision is made. taking into account mobility patterns and individual service requirements in a user-specific manner. It cannot be obtained. Therefore, even if network conditions or user behavior change, priorities remain the same. If it remains the same, the cell selection decision will become detached from the actual working conditions and the current 25 This can lead to inefficient use of radio resources. The presence of micro, pico, and femto cells alongside macrocells in heterogeneous network structures, layers of cells with different capacities and coverage characteristics together in the same area It requires management. When static cell reselection priorities are used, the result is a low 30. Small, high-capacity cells may not be preferred enough, whereas wide coverage is preferable. User density can increase in macro-level cells that provide this service. Such a load distribution can affect small... the underutilization of the available capacity of cells, physical resources in macrocells increased consumption of blocks and data transfer speed per user This can lead to a decrease. Therefore, the existing static structure is 35 depending on the changing cell loads. It is unable to optimally meet the need for real-time load balancing. 2 Where neighboring cells receive signals at or near the cell boundaries. Inappropriate fixed priority and hysteresis values ​​in the regions short-term user equipment failure. This causes repeated re-selection between cells at time intervals. This is possible. These re-election cycles, described as ping-pong, lead to unnecessary RRCs. increasing measurement and cell search operations, more network signaling resources 5 This leads to increased usage and faster battery drain of the user equipment. Frequent re-elections can negatively impact service continuity, especially for cells. This can reduce connection stability for users moving along the border. Current In systems, this behavior is considered along with past selection patterns and the current network state. an integrated mechanism that reduces the impact in a user-specific way by evaluating 10 It is not available. With a user moving at high speed in the current cell re-selection mechanisms a user moving at a constant or low speed with the same or similar priority parameters It can be evaluated. A high-speed user traveling inside the vehicle, small 15 Because it can see the cells in quick succession, it can't unnecessarily see the small cells. Prioritization leads to frequent cell turnover, an increase in the number of handovers, handover This can lead to failures and call drops. In contrast, low speed or Not directing regular users to suitable smaller cells does not make the existing capacity efficient. It can prevent its use. By estimating the user's speed and direction information, it can predict the future 20 The absence of a decision-making structure that anticipates location and mobility class, different mobility This leads to the same re-selection behavior being applied to their profiles. In addition, in traditional re-election decisions, the user's active service type is the service. 25 factors such as quality requirements, data usage patterns, battery status, and user preferences. Contextual information is not considered together. Data transfer, voice call, video streaming. and the capacity, latency, and energy requirements of different services such as IoT communication Although they differ, static priorities address these differences on a user-by-user basis. It cannot reflect this. For example, a video streaming user who requires high data transfer speeds. Managing an IoT device where energy consumption is a priority, using the same cell selection logic, 30 This can lead to suboptimal results in both use cases. A lack of contextual awareness therefore affects user experience and network resource utilization. This limits their ability to be optimized together. In terms of energy efficiency, existing static systems also eliminate unnecessary cell searches and 35 re-selection based on user equipment battery level or power saving status It does not provide a prioritization strategy that correlates cell search, measurement, and re-evaluation. 3 Unnecessary repetition of selection processes causes additional energy consumption in user equipment. This can negatively affect the operating time of devices, especially those operating in energy-saving mode. It can influence user behavior, as well as past cell selection data. The fact that these patterns are not used to predict future election needs means the system is alone. This causes the network to remain in a reactive state, only responding to the current situation. The network itself... 5 With its optimizing SON functions, it enables sufficient user-specific cell re-selection decisions. The fact that it is not carried out within a feedback loop also maintains the need for manual intervention. The aforementioned shortcomings affect RSRP, RSRQ, and RSSI obtained from user equipment. radio measurements such as; real-time load status of cells; user service, QoS, 10 considering the battery and mobility context and past cell selection behaviors together This necessitates a more integrated technical structure. Such a structure would benefit the user. predicting mobility, learning from historical data, user experience, load balancing, energy efficiency and ping-pong reduction targets under multi-objective optimization being able to consider them together, the result obtained is user-specific 3GPP re-selection 15 It can convert these into parameters and monitor the performance of the applied parameters to get feedback. It needs to be able to generate feedback. Furthermore, this feedback should be able to generate other END feedback. Coordination with functions, cell re-selection according to changing network conditions This is of technical importance in terms of continuously improving the strategy. As a result of research conducted in the literature, the code numbered US20240187982A1, “Adaptive cell Adaptive cell selection, reselection, and mobility assistance techniques A patent document titled "selection and mobility support techniques)" was found. This document outlines cell selection and reselection procedures for wireless communication systems. Preference information and mobility support information are provided by the user equipment on 25 It relates to techniques for evaluation. However, the document mentioned refers to real-time. network load, user context, past selection behavior, and machine learning outputs all come together. by evaluating 3GPP compliant cell reselection priorities for each user piece of equipment. Generating and implementing the result through multi-objective optimization and the final feedback loop. No evidence of its acquisition has been found. 30 As a result of research conducted in the literature, the "Techniques for UE" code numbered US20220030453A1 was identified. Mobility prediction based radio resource management (User equipment mobility) A patent document titled "Predictive radio source management techniques" was found. The document in question involves predicting the future mobility of user equipment and 35 It relates to the use of predicted mobility information in radio resource management operations. However, the document in question states that mobility forecasting is based on service context and cell load. 4 combined cellReselectionPriority, qRxLevMin, sIntraSearch, sNonIntraSearch and Generating threshServingLow parameters in a user-specific way and improving performance. by readjusting optimization parameters with machine learning according to the indicators No relevant evidence has been found. Ultimately, the problems mentioned above, which cannot be solved with current technology, are the subject of this technical analysis. This has made it necessary to make an innovation in the field. A BRIEF DESCRIPTION OF THE INVENTION The aim of the invention is to make cell re-selection priorities in the mobile network fixed for all users. Instead of holding onto it, it dynamically determines it according to the immediate conditions of each user's equipment. predicting future mobility and providing feedback on network performance. The goal is to provide a structure that improves in this way. The main purpose of the invention is to analyze real-time network conditions, user mobility profile, cell load, by evaluating signal quality, service requirements, and past selection behavior together The goal is to generate user-specific cell reselection priorities. Another aim of the invention is to increase the use of small cells in heterogeneous networks and to utilize macrocells. The goal is to reduce the load on the network and ensure balanced use of network resources. Another purpose of the invention is to determine mobility class, hysteresis, and time-dependent decision. The mechanisms aim to reduce ping-pong reselections and unnecessary cell searches. Another aim of the invention is to improve user experience through machine learning and multi-objective optimization. a policy of assigning balanced priorities between load balancing and energy efficiency targets to create. Another purpose of the invention is to provide feedback through KPI tracking and SON integration. This allows for automatic readjustment of model and optimization parameters. In LTE and heterogeneous mobile networks, user equipment idles during cell re-selection. on an electronic processing unit to dynamically optimize its priorities Adaptive cell reselection priority achieved via running software 35 It is an optimization system; its feature is:  RRC measurement data including RSRP, RSRQ and RSSI from user equipment mobility status, QoS requirements, and service type in real time. user equipment monitoring that collects and transfers that data for further processing. interface,  By filtering the raw data received from the user equipment monitoring interface, 5 It processes, normalizes, performs noise and missing data operations, and attributes. Data collection and preprocessing module that converts data into vectors,  Service type and data usage profile in the outputs of the data collection and preprocessing module. and by combining user status information, a user context profile is created. context analysis engine, 10  Sequential signal measurements in the outputs of the data collection and preprocessing module Evaluating according to mobility criteria and via speed and direction vectors Mobility estimation module that determines the mobility class.  Monitors PRB usage, number of active users, and data transfer speeds of cells and the network load monitoring unit, which generates cell load information using that data, 15  context profile, mobility estimation, cell load information, measurement reports and past cell re-selection decisions are recorded in a time series structure. data repository,  User selection patterns using data from the historical data repository evaluating, identifying mobility trends and prioritizing policy output 20 the machine learning engine that produces  Instantaneous network load and signal data from the output of the machine learning engine. combining user experience, load balancing, energy efficiency, and ping-pong. A priority optimization algorithm that generates priority solutions based on reduction targets.  cellReselectionPriority, 25 using the output of the priority optimization algorithm qRxLevMin, sIntraSearch, sNonIntraSearch and threshServingLow values Priority that converts to 3GPP-compliant re-election parameters including parameter generator,  User-specific reselection obtained from the priority parameter generator processing its parameters with ASN.1 encoding and sending the relevant 30 via RRC message RRC signaling interface that transmits to user equipment,  the re-election resulting from the applied re-election parameters, ping-pong, handover, data transfer speed, and connection latency It monitors indicators, compares them to target thresholds, and generates feedback. performance monitoring system, 35 6  Performance monitoring system feedback data along with other SON functions Managing network coordination and prioritizing using a machine learning engine. SON integration enables readjustment of the optimization algorithm. interface It includes. 5 The structure of the invention and the best understanding of its advantages, including additional components. This should be evaluated together with the figures explained below. BRIEF DESCRIPTION OF THE FIGURES 10 Figure 1 shows the data and control in the adaptive cell reselection priority optimization system. It is a schematic representation showing the relationship. Figure 2 shows the process flow in the adaptive cell reselection priority optimization method. This is a schematic representation showing the process. 15 The drawings do not necessarily need to be scaled and are necessary for understanding the invention. Details that are not present may have been overlooked. Furthermore, at least to a significant extent... Identical elements or elements with similar functions are numbered the same way. It is shown. 20 REFERENCE NUMBERS 1. User equipment monitoring interface 2. Data collection and preprocessing module 25 3. Context analysis engine 4. Mobility forecast module 5. Network load monitoring unit 6. Historical data repository 7. Machine learning engine 30 8. Priority optimization algorithm 9. Priority parameter generator 10. RRC signaling interface 11. Performance monitoring system 12. FINAL integration interface 35 7 1001. RSRP, RSRQ, and from user equipment by the user equipment monitoring interface. RRC measurement reports containing RSSI, mobility status, QoS requirements and service type. real-time collection of information 1002. Filtering the raw data collected by the data collection and preprocessing module. passing through, normalizing, removing noise components, outliers and missing data 5 converting them into temporal feature vectors by processing them 1003. Active service type, data usage pattern, battery by context analysis engine. dynamic user context profile by integrating status and user preferences creation 1004. Signal change in successive RRC measurements by the mobility estimation module is 10 By evaluating the user's velocity and direction vectors, their future position can be estimated. forecasting and determining mobility class 1005. The network load monitoring unit monitors the PRB usage rate of the cells, the number of active users, and Cell load information is obtained by monitoring the data transfer rates of the downlink and uplink. 15 produced 1006. Context information, mobility estimations, network load data from the historical data store, Measurement reports and past cell reselection decisions in time series data structure hiding 1007. Using data from the historical data repository by the machine learning engine Learning user cell selection patterns, predicting future activity, and 20 Development of a priority assignment policy 1008. Prioritization optimization algorithm with the outputs of the machine learning engine. Combining real-time network conditions to improve user experience, load balancing, and energy. Pareto-optimal prioritization solution under the goals of efficiency and ping-pong reduction determination 25 1009. Priority resolution determined by the priority parameter generator. cellReselectionPriority, qRxLevMin, sIntraSearch, sNonIntraSearch and threshServingLow Converting the values ​​into 3GPP compliant RRC parameters. 1010. User-specific priority parameters via the RRC signaling interface ASN.1 30 by coding and placing it in the RRC message and transmitting it to the relevant user equipment 1011. Cell reselection rate, ping-pong rate, by the performance monitoring system, handover success rate, user data transfer speed, and connection establishment latency Generating feedback based on target thresholds by monitoring indicators 1012. Feedback and network coordination information from the SON integration interface. Sharing with other END functions and machine 35 if performance targets are not met re-evaluating the parameters of the priority optimization algorithm with the learning engine adjustment 8 DETAILED DESCRIPTION OF THE INVENTION This detailed explanation describes the adaptive cell reselection priority optimization that is the subject of the invention. The system and method are solely aimed at a better understanding of the subject, with no limiting effects. 5 This is explained with examples that will not create a problem. The system described in the invention transmits RRC measurement reports and mobility data from user equipment. It collects status, QoS requirements and service type in real time according to 3GPP TS 36.331. User 10 parses compatible RRC messages and transfers the data to the next processing stage. The equipment monitoring interface (1) filters and normalizes the raw data in question, It removes noise components, identifies outliers, completes missing data, and Data collection and preprocessing module (2) that converts to feature vectors, active service type, data Dynamic user by integrating usage profile, battery status, and user preferences. The context analysis engine (3) that creates a context profile extracts the velocity and direction vector from successive measurements. mobility that predicts future location and assigns the user to a mobility class The prediction module (4) measures the PRB usage rate of the cells, the number of connected users, the data transfer rate and Network load monitoring unit (5) which generates load information by monitoring the number of RRC connections, context information, mobility estimates, network load data, measurement reports, and past re-selection data. The historical data store (6) which stores its decisions in a time series structure, 20 from the historical data Learning user choice patterns and future mobility and assigning prioritization policies. The machine learning engine (7) that developed the learning outputs and the instantaneous network status combining user experience, load balancing, energy efficiency and ping-pong reduction Priority optimization algorithm (8) which produces Pareto-optimal solutions under its goals, Priority parameter 25 converts the solution to 3GPP compliant re-selection parameters. The producer (9) places user-specific parameters in the RRC message using ASN.1 encoding. RRC signaling interface (10) that transmits to user equipment, re-selection, ping-pong, handover, by monitoring data transfer speed and connection latency indicators, returns to normal. Performance monitoring system that produces feed (11) and coordination with the said feedback The SON integration interface (12) which shares its data with other SON functions, communicates data with each other and 30 It operates within a control relationship. User equipment monitoring interface (1), RSRP, RSRQ and In addition to measurement reports containing RSSI, mobility status indicator and active carriers are also included. It can capture QCI values. Data acquisition and preprocessing module (2), Gaussian noise Temporal features reduction, Z-score normalization, and sliding window approach. It can extract and complete missing values ​​using the K-nearest neighbor algorithm. 35 It can use the context analysis engine (3), HTTP, HTTPS, RTP and SIP protocols. from the analysis, the service type, the intermittent or continuous data usage pattern, and the user. 9 The equipment can assess battery level from capability information. Mobility estimation. The module (4) can estimate the Doppler shift from the rate of change of successive RSRP values. and with an extended Kalman filter, it can determine the velocity vector; in the last ten seconds. An average speed above 30 km / h indicates high mobility, while a speed between 3-30 km / h indicates moderate mobility. Mobility levels below 3 km / h are associated with a low mobility class. Network load 5 Monitoring unit (5) monitors PRB usage rate and number of active users via S1 and X2 interfaces. and the data transfer speeds of the downlink and uplink are in one-second intervals. It can collect historical data (6), time series such as InfluxDB or TimescaleDB. It can be stored in a database structure, and each record has a timestamp, user equipment ID, The cell can be indexed by cell ID, measurement value, and context tag; the invention is a 10 The last 90 days of data are stored in its structure. The machine learning engine (7) stores the past Using the training data in the data store (6) with the Random Forest classifier with 200 trees It can learn cell selection patterns using an LSTM network with two hidden layers and 128 units. It can predict user activity in the next 60 seconds, using Q-learning-based reinforcement. The learning agent can learn the appropriate prioritization policy and 15 in anomaly detection. It can benefit from the isolation forest approach. Priority optimization algorithm (8), maximum re-selection rate, minimum signal quality, and operator policies as constraints. By using NSGA-II, we can create a Pareto-optimal solution set, improving user experience. It evaluates the score, load balancing score, and energy efficiency score together. Priority The parameter generator (9) sets the optimization score for cellReselectionPriority to 20 in the range of 0-7. It can map to a number, determine the qRxLevMin value between -70 and -22 dBm, and `sIntraSearch`, `sNonIntraSearch`, and `threshServingLow` are used to retrieve cell-specific values. It can assign. The RRC signaling interface (10) allows the user to assign the RRC Connection. Create a reconfiguration message and recreate it in the MobilityControlInfo information element. It can encode election priorities and send the message via S1-AP Downlink NAS Transport 25 It can transmit to the user equipment. Performance monitoring system (11), five minutes Improvements in re-election rate, ping-pong rate, and average data transfer speed over periods It is able to calculate that 10% of users are in the static priority control group and 90% are... It is possible to apply an A / B testing framework that manages as an adaptive priority group, and the two groups We analyzed the metrics using the Mann-Whitney U test with a significance level of p<0.05 for 30 It can make comparisons. When target KPI values ​​are not met, the performance monitoring system... Prioritization with machine learning engine (7) in line with feedback generated by (11). The hyperparameters of the optimization algorithm (8) are refined with gradient descent. It is adjustable, the learning speed can be regulated, and the model has a retraining trigger. It can be activated. SON integration interface (12), X2-AP Mobility Change Request 35 It can coordinate with neighboring eNodeBs through messages, within the scope of MLB. Load information, handover error report under MRO, and power adjustment recommendation under ICIC. They can share it. The system is part of the SON ecosystem within the modern LTE operator's network management system. It can be implemented as a mobility optimization module. Element management system 5 Measurement data and RRC via S1-AP and X2-AP with eNodeB units at the layer Configuration commands can be exchanged, MLB, MRO at the area management system layer. RACH optimization, coverage and capacity optimization, and energy saving functions. Conflict resolution can be performed between them via the SON coordination function. Enforcement of operator policies at the network management system layer, SLA monitoring, network-wide 10 Data can be provided for KPI visualization and strategic planning. Downstream interfaces S1- Control plane signaling and user context management via MME, and via X2. Exchange of load information between eNodeBs and eNodeB via TR-069 / NETCONF It can encompass the configuration; upstream interfaces provide performance via REST API. Metric presentation and configuration update, alarm and trap messages via SNMP, 15 Kafka Streams enables real-time event streaming; east-west interfaces. SON coordination protocol, policy engine interface and aggregate data export for Hadoop / Spark. It may include transfer. In one configuration of the invention, the system can be built with a cloud-native microservice architecture; data 20 Message parsing and data validation, machine learning with Kafka clients. Training service for model training and version control with distributed TensorFlow or PyTorch developers, Optimization service with parallel optimization using a high-performance computing cluster, The signaling service includes RRC message generation and ASN.1 encoding / decoding, while the monitoring service is... 25 Prometheus and Grafana-based metric collection and representation functions It can provide services such as automatic scaling and fault tolerance on Kubernetes. It can be managed with phased updates and health checks. The system has over one million users. Real-time optimization for simultaneous user equipment through horizontal scaling. It can be implemented in a way that supports; data acquisition latency is below 100 ms, machine Learning inference delay under 500 ms, optimization decision delay under 2 seconds 30 And it can be configured with end-to-end processing latency under 5 seconds. User data may be processed in compliance with GDPR, and personally identifiable information may be processed using IMSI or other methods. IMEI numbers can be anonymized through hashing, and user authentication management is integrated into the system. It can be done. Machine learning model anomaly detection against poisoning attacks, introduction 35 Verification and sanitization can be implemented, and TLS 1.3 secure communication channels are used in the interfaces. It is available to use. The system conforms to 3GPP TS 36.300, TS 36.304, TS 36.331, TS 36.413, TS 36.423 standards. 11 and can be implemented in accordance with TS 32.500 standards and Release 8 / 9 Backward reversing mechanism to static priorities in user equipment. It is able to provide compatibility. In a structured approach to the invention, real-time and user-specific prioritization results in 5 25-35% improvement in user experience, 40-50% increase in small cell usage, 20-30% reduction in macrocellular load, 60% reduction in unnecessary cell reselection. 70% reduction, 30-40% reduction in handover count at high mobility, future cell Estimation of choice requirement with 80-85% accuracy, user equipment battery consumption. 15-20 percent reduction and an 80 percent reduction in the need for manual intervention. These can be targeted. Capacity for video streaming users and energy efficiency for IoT devices. It can be prioritized and specially optimized for devices in energy-saving mode. These strategies can be implemented. Mobility status estimation 15 Mobility class re-selection number, evaluation period and speed threshold It is determined that; N_resel > N_high and Speed ​​> 30 km / h indicate high mobility, N_mid < N_resel <= N_high or 3 < Speed ​​<= 30 km / h indicates moderate mobility, N_resel <= N_mid and A speed of <= 3 km / h corresponds to a low mobility class. 20 Cell reselection sorting criterion System according to the service cell ranking criteria within the scope of 3GPP TS 36.304 section 5.2.4.3. It can dynamically optimize Q_offset values. 25 Neighboring cell sorting criterion The ranking value is calculated using quality and offset terms measured for the neighboring cell. The Q_offset term can be updated according to the adaptive optimization output. 30 Cell load index Cell load index PRB usage, normalized number of users, and normalized average data. It is obtained by weighted combination of transmission speed components. 35 12 User experience objective function The user experience score is based on predicted data transfer speed, ping-pong probability, and signal quality. It can be calculated as a weighted combination. Load balancing objective function The load balancing score will increase as the variance of cell load distribution decreases. It can be defined. Energy efficiency objective function Energy efficiency is the ratio of the expected number of cell searches to the reference value and battery level. It can be calculated using weight. qRxLevMin mapping relation Priority parameter generator (9) uses the optimization score qRxLevMin value to -70 It can map to a range of -22 dBm. Re-election rate Total number of re-selections and total user equipment within the scope of performance monitoring. It can be normalized according to the number and monitoring period. Ping-pong odds The ping-pong ratio is the total number of re-selections in the form of returning to the previous cell. It can be calculated as a ratio to the number of elections. Improvement in average data transfer speed. 13 The relative difference between the current average data transfer speed and the reference average value is a percentage. It can be monitored in this way. Load balancing index The load balancing index shows PRB usage at 0.4, and the ratio of active users to capacity is 0.3. The ratio of the average data transfer speed to the peak value can be used with a weight of 0.3. Ping-pong detection condition Cell_ID(t) is equal to Cell_ID(t-2T) and Cell_ID(t) is different from Cell_ID(tT). A ping-pong event is detected when the T value is less than the T_threshold value; The `t_threshold` setting is typically set to 10 seconds. The invention describes network slice selection between eMBB, URLLC and mMTC slices in 5G networks, according to NSSAI 15. adaptively determining information based on user context and slice load; dense Wi-Fi load balancing and band balancing between 2.4 GHz, 5 GHz and 6 GHz bands in their setups In routing and access point selection; for NB-IoT and LTE-M devices, the device... Network connection priorities are determined by PSM and based on type, power class, coverage level, and QoS profile. Optimized with eDRX parameters; LTE-NR dual connection for LTE-only, NR-20 use of only or EN-DC depending on mobility, coverage, capacity and service requirements. Factors considered in determining this include: coverage quality and roaming costs for devices with dual SIM or eSIM support. In selecting an operator based on data plan and user preference; LEO satellite and terrestrial network hybrid. In their designs, the choice between satellite beam or terrestrial cell affects signal quality, latency, and mobility. and optimized according to service continuity; V2X communication C-V2X mode 3 and mode 4 25 The choice between them depends on traffic density, vehicle speed, safety criticality, and coverage. in carrying out and facilitating the transition of corporate users between private LTE / 5G networks and the public network. in terms of location, application type, security requirements and network capacity It is applicable. The steps involved in the process are as follows:  RSRP from user equipment by user equipment monitoring interface (1), RRC measurement reports including RSRQ and RSSI, mobility status, QoS real-time collection of information including requirements and service type. (1001), 35  Filtering the raw data collected by the data collection and preprocessing module (2) passing through, normalization, removal of noise components, outlier 14 and processed through incomplete data to create temporal feature vectors. conversion (1002),  Active service type, data usage pattern, by context analysis engine (3), Dynamic user context by integrating battery status and user preferences. creation of profile (1003), 5  Signal from successive RRC measurements by the mobility estimation module (4) Estimating the user's velocity and direction vector by evaluating its change, Predicting future location and determining mobility class (1004),  Network load monitoring unit (5) monitors the PRB usage rate of the cells, active user By monitoring the number of cells and the data transfer speeds of the downlink and uplink, cell 10 Generating load information (1005),  Context information, mobility estimations, network load by historical data store (6) data, measurement reports, and time series of past cell reselection decisions. storage in data structure (1006),  The machine learning engine (7) uses data from the historical data repository (6) 15 Learning user cell selection patterns using this method will prepare future users for the future. development of mobility forecasting and prioritization policy (1007),  Priority optimization algorithm (8) of the machine learning engine (7) By combining outputs with real-time network conditions, user experience and load Pareto-20 under the goals of balancing, energy efficiency and ping-pong reduction Determining the optimal priority solution (1008),  Priority solution determined by the priority parameter generator (9) cellReselectionPriority, qRxLevMin, sIntraSearch, sNonIntraSearch and 3GPP compliant RRC parameters including threshServingLow values conversion (1009), 25  User-specific priority by the RRC signaling interface (10) The parameters are placed in the RRC message using ASN.1 encoding, and the relevant user transmission to equipment (1010),  Cell reselection rate by performance monitoring system (11), ping-pong rate, handover success rate, user data transfer speed and connection establishment 30 Monitoring lag indicators and providing feedback based on target thresholds. creation (1011),  Feedback and network coordination by the SON integration interface (12) sharing information with other END functions and performance targets Priority optimization with machine learning engine (7) when not met 35 Resetting the parameters of the algorithm (8) (1012). The system first operates using radio measurements and mobility data obtained from user equipment. and service requirements are collected via the user equipment monitoring interface (1) and data It is processed by the collection and preprocessing module (2). Context analysis engine (3), mobility The prediction module (4) and the network load monitoring unit (5) in parallel monitor the user context, predicted 5 to generate motion and cell load inputs, the data in question are in the historical data store (6) Historical election and performance data are stored along with the machine learning engine. (7) Prioritizing optimization in generating behavior and mobility predictions from past and current data. The algorithm (8) uses these outputs to find the appropriate priority solution for each user equipment. determining and priority parameter generator (9) solution 3GPP compliant re-selection 10 It converts the parameters into. The parameters are converted to the user via the RRC signaling interface (10). transmitted to the equipment, performance monitoring system (11) evaluates the resulting KPI values generating feedback and SON integration interface (12) with coordination feedback by sharing this information across the network, learning and optimization parameters can be adjusted as needed. It allows for readjustment. 15

Claims

16 REQUESTS 1. In LTE and heterogeneous mobile networks, user equipment is in an idle cell. an electronic process to dynamically optimize re-election priorities Adaptive cell regeneration is performed via software running on the unit. 5 It is a selection priority optimization system, and its feature is;  RRC measurement data including RSRP, RSRQ and RSSI from user equipment mobility status, QoS requirements, and service type in real time. the user who collects and transfers that data to subsequent processing equipment monitoring interface (1), 10  Processes the raw data received from the user equipment monitoring interface (1), normalizing, performing noise and missing data processing, and attribute processing. Data collection and preprocessing module that converts into vectors (2),  Service type, data usage in the outputs of the data collection and preprocessing module (2) By combining profile and user status information, the user context profile is 15. context analysis engine (3),  Sequential signal in the outputs of the data collection and preprocessing module (2) evaluating measurements according to mobility criteria and speed and direction vectors Mobility estimation module (4) which determines the mobility class.  The PRB usage of the cells, the number of active users, and the data transfer speeds are 20 network load that monitors and uses that data to generate cell load information. monitoring unit (5),  context profile, mobility estimation, cell load information, measurement reports and recording past cell re-selection decisions in a time series structure historical data repository (6), 25  User selection patterns using data from the historical data store (6) an assessment, mobility trend identification, and prioritization policy output. machine learning engine (7),  Instantaneous network load and signal data with the output of the machine learning engine (7) combining user experience, load balancing, energy efficiency and ping-pong 30 Priority optimization, which generates priority solutions based on reduction targets. algorithm (8),  using the output of the priority optimization algorithm (8) cellReselectionPriority, qRxLevMin, sIntraSearch, sNonIntraSearch and 3GPP compliant re-selection 35 including threshServingLow values priority parameter generator (9) which converts to its parameters, 17  User-specific reselection from priority parameter generator (9) processing its parameters using ASN.1 encoding and sending the relevant RRC message. RRC signaling interface (10) which transmits to user equipment,  the re-election resulting from the applied re-election parameters, Ping-pong, handover, data transfer speed, and connection latency are all 5. It monitors indicators, compares them to target thresholds, and generates feedback. performance monitoring system (11),  The feedback data of the performance monitoring system (11) with other SON functions using a machine learning engine to manage network coordination together. (7) enables the readjustment of the priority optimization algorithm (8) 10 SON integration interface (12) It includes.

2. The system is compliant with Request 1 and its feature is the data collection and preprocessing module (2), Gaussian noise reduction, Z-score normalization, and temporal 15 with sliding window. It implements feature extraction operations and uses K-nearest neighbors for missing data completion. It is a module that uses the algorithm.

3. The system compliant with Claim 1, and its feature is; the mobility estimation module (4), sequential RSRP Kalman 20, which estimates the Doppler shift from the rate of change of the measurements, It is a module that determines the user speed vector using a filter.

4. The system is compliant with claim 1 and its feature is that the machine learning engine (7), Random Forest classifier, LSTM neural network and Q-learning based reinforcement learning agent an engine that uses both and applies the isolation forest approach in anomaly detection 25 It is the fact that.

5. The system is compliant with claim 1 and its feature is that the priority optimization algorithm (8), maximum re-selection rate, minimum signal quality, and operator policy restrictions. An algorithm that produces a Pareto-optimal solution set with NSGA-II by evaluating it as such. 30 It is the fact that.

6. The system complies with Claim 1 and its feature is that the performance monitoring system (11) allows users one section is a static priority control group and the remaining section is an adaptive priority control group. The Mann-35 A / B test framework monitors group performance metrics as a group. It is a system that compares to the Whitney U test. 18 7. The system compliant with Request 1 is characterized by its SON integration interface (12), X2-AP Coordinating with neighboring eNodeB units via Mobility Change Request, MLB load information, MRO handover error reports, and ICIC power adjustment recommendations. It is a sharing interface.

8. In LTE and heterogeneous mobile networks, user equipment in idle state in cells an electronic process to dynamically optimize re-election priorities Adaptive cell regeneration is performed via software running on the unit. It is a selection priority optimization method, and its characteristic feature is;  User equipment monitoring interface (1) from user equipment 10 RRC measurement reports, including RSRP, RSRQ, and RSSI, indicate mobility status. Real-time information including QoS requirements and service type. collection (1001),  Raw data collected by the data collection and preprocessing module (2) filtering, normalization, removal of noise components, 15 Temporal attributes are processed by removing outliers and missing data. conversion to vectors (1002),  Active service type, data usage pattern, by context analysis engine (3), Dynamic user integration by integrating battery status and user preferences. Creation of context profile (1003), 20  Signal from successive RRC measurements by the mobility estimation module (4) by evaluating the change in the user's velocity and direction vector forecasting, predicting future location and mobility class determination (1004),  Network load monitoring unit (5) monitors the PRB usage rate of the cells, active 25 number of users and downlink and uplink data transfer speeds Generating cell load information by monitoring (1005),  Context information, mobility estimations, network by historical data store (6) load data, measurement reports, and past cell reselection decisions storage in time series data structure (1006), 30  Data from the historical data store (6) by the machine learning engine (7) Learning user cell selection patterns using this method will prepare future users for the future. development of mobility forecasting and prioritization policy (1007),  Priority optimization algorithm (8) of the machine learning engine (7) User experience by combining outputs with real-time network conditions, 35 19 under the goals of load balancing, energy efficiency and ping-pong reduction Determining the Pareto-optimal priority solution (1008),  Priority solution determined by the priority parameter generator (9) cellReselectionPriority, qRxLevMin, sIntraSearch, sNonIntraSearch and 5 3GPP compliant RRC parameters including threshServingLow values conversion (1009),  User-specific priority by the RRC signaling interface (10) The relevant parameters are embedded in the RRC message using ASN.1 encoding. transmission to user equipment (1010),  Cell reselection rate by performance monitoring system (11), ping-10 pong rate, handover success rate, user data transfer speed, and connectivity. Monitoring setup delay indicators and providing feedback based on target thresholds. creation (1011),  Feedback and network coordination by the SON integration interface (12) Sharing information with other END functions and performance targets 15 Priority optimization with machine learning engine (7) when not met Resetting the parameters of the algorithm (8) (1012) It includes the steps of the process.

9. This method complies with Claim 8 and its characteristic is that the number of re-selections for the mobility class is 20. Determined according to evaluation time and speed threshold, and the average of the last ten seconds. High mobility is indicated if the speed is above 30 km / h, and between 3-30 km / h. Moderate mobility is indicated if it is in this state, and low mobility if it is below 3 km / h. It includes the classification process step.

10. The method is compliant with Request 8, and its feature is; the last 90 days in the historical data repository (6). Time series data timestamp, user equipment ID, cell ID, measurement. This involves the process of storing the data indexed by its value and context tag.

11. The method is compliant with claim 8 and its feature is that it is performed by a machine learning engine (7) 200 30 Learning cell selection patterns with a tree-based Random Forest model, two hidden The movement over the next 60 seconds with a layer and 128-unit LSTM network. the steps involved in predicting and learning the priority policy using the Q-learning agent It includes. 35 12. The method is compliant with Claim 8 and its feature is that it is monitored by a performance monitoring system (11) five Cell reselection rate, ping-pong rate, and average data in minute periods. Calculating the transfer rate improvement and gradient when target thresholds are not met. model retraining trigger activation with descent and hyperparameter fine-tuning. It includes the steps involved in the acquisition process.

13. This method complies with Claim 8 and its characteristic is that user data is personally identifiable. 5 Anonymizing the data sections using IMSI or IMEI hashing, machine learning model Implementing anomaly detection and input verification against poisoning attacks and The steps for using TLS 1.3 secure communication channels in the interfaces It includes.

Images