DEVELOPMENT OF A DATABASE FOR STORAGE AND ANALYSIS OF SYNCHRONIZATION AND ERROR-CORRECTING CODING PARAMETERS IN MOBILE COMMUNICATION SYSTEMS

Abstract

This paper addresses the problem of designing a specialized database for storing, organizing, and analyzing synchronization parameters and error-correcting coding parameters in mobile communication systems. It is shown that existing data management approaches are typically focused either on physical layer signal processing, network analytics, or machine learning applications separately, and therefore do not provide an integrated representation of synchronization, channel conditions, decoding performance, and learning data within a unified framework. A multi-level database architecture is proposed, combining a relational model for structured data, object storage for large-scale signal and experiment data, and analytical processing mechanisms. The proposed approach enables simultaneous storage and analysis of synchronization parameters (time offset, frequency offset, phase error, jitter), channel parameters (SNR, Eb/No), coding parameters (code family, code rate, block length, decoding algorithm), and decoding performance metrics (BER, BLER, latency, number of iterations). The logical data model consists of the following entities: Experiment, Scenario, Run, Observation, Reinforcement Learning Transition, Artifact. The Experiment entity represents the overall context of the study. The Scenario defines a fixed configuration of channel and coding parameters. The Run corresponds to an individual execution of a scenario under specific computational conditions. The Artifact entity is used to manage metadata for large external objects such as signal recordings, datasets, and trained models. A key feature of the proposed approach is the integration of the database with reinforcement learning workflows. For this purpose, the Reinforcement Learning Transition entity is introduced, enabling representation of data as sequences of the form “state–action–reward–next state.” The system state is defined based on channel quality indicators, synchronization parameters, and decoding metrics, while actions correspond to adaptive adjustments of decoding and transmission parameters. The reward function is formulated as a trade-off between decoding accuracy and processing latency. This formalization allows direct construction of datasets for offline reinforcement learning without additional preprocessing. The practical implementation of the proposed approach relies on open and scalable technologies, including relational database systems, object storage, and machine learning frameworks. The obtained results can be used for the analysis of decoding algorithms, investigation of synchronization impairments, and development of adaptive communication systems based on artificial intelligence techniques for 5G and 6G.