Samsung and NTT Docomo Validate User-Level AI-RAN

The companies used aggregated data from mobile devices, described in the release as MDT, or minimization of drive tests, to analyze user context and movement trends. NTT Docomo said the system can detect early signs of throughput degradation and alter settings when predicted speeds could fall below a threshold needed by an application.

The stated example is video streaming: if the system predicts lower transmission speed that could produce buffering or reduced video quality, it can apply a different network configuration. Samsung and NTT Docomo also reported joint work spanning AI-RAN development, commercial-network drive-test data collection in Japan, and data-measurement process improvements.

UPI and BusinessKorea reported that the pair also developed a more selective data-processing approach for AI-RAN operations. Rather than processing all wireless-environment information together, the approach processes information needed for a specific user problem. That design consideration is material because inference and data pipelines in radio networks operate under latency and capacity constraints that differ sharply from offline network planning workflows.

The reported 5.9-percentage-point reduction is based on simulations, not a reported commercial deployment or a population-wide service-quality measurement. The sources do not provide the number of users, traffic mix, model architecture, prediction horizon, throughput threshold, or confidence intervals behind the result. Those details would be necessary for practitioners to evaluate generalization across cell density, mobility patterns, radio bands, and application classes.

BusinessKorea reported that Samsung and NTT Docomo submitted their data-processing procedure to 3GPP in February. Separately, NTT Docomo's release characterizes the validation as a step toward predictive network operations as AI-driven optimization becomes more prevalent in 6G discussions.

For network engineers, the work illustrates a shift from cell-wide radio policy toward per-user optimization driven by telemetry and prediction. Comparable systems commonly face tradeoffs among prediction accuracy, data minimization, control-loop latency, fairness across users, and the risk that an optimization for one device changes congestion conditions for others. The available reporting does not establish how the validated system addresses those tradeoffs.

The announcement concerns a validation using Japanese network and test-field data. Samsung and NTT Docomo have not reported a commercial rollout timeline or customer-facing availability in the cited materials.