AI-Based Optimization for User and Service-Level Perceived Quality

In network environments transitioning to 5G-Advanced and 6G, optimization is becoming complex due to the differing quality standards required by various services such as Extended Reality (XR), cloud gaming, and ultra-high-definition streaming.

In particular, quality degradation phenomena such as handover failures or increased latency during mobility occur repeatedly for specific users or paths, but existing cell-level optimization methods struggle to resolve these issues.

To address this, Samsung Research proposed an AI-based optimization framework. This system analyzes historical user data to learn quality degradation patterns and predicts potential issues based on these patterns during real-time network operations.

During the learning phase, multivariate time-series data such as received signal strength, quality, interference, and downlink throughput are compressed by an AI encoder and transformed into a low-dimensional embedding space.

During the operational phase, the system uses the learned patterns to detect quality degradation in advance and maintains service quality through proactive mobility control. This signifies that precise quality management considering individual user contexts is possible, rather than merely improving average cell performance.