6G-SANDBOX

6GMeasurOps conducted >3,000 experiment runs at the 6G-SANDBOX Athens/OTE Academy site, covering diverse packet sizes, intervals, durations, and load conditions in three different layers. In the physical layer, RSRP, RSRQ and SINR remained stable with minimal fluctuations, showing reliable mid-band 5G coverage. Measurements confirmed strong radio stability even under stress traffic. In the transport layer, TCP uplink throughput was consistently ~80 Mbps, while TCP downlink throughput was ~15–18% lower than uplink. UDP flows showed larger variance, confirming sensitivity to congestion/load and aggregated RTTs were in line with commercial thresholds; retransmission rates were measurable but within acceptable ranges. Regarding application layer results, MQTT outperformed video streaming in latency and jitter — ideal for IoT, URLLC, and low-power services.

while video streaming suffered more from jitter and bandwidth fluctuations, highlighting its sensitivity to load. Application-level RTT evolutions over hours confirmed stable MQTT but variable video flows behavior.

Finally, regarding AI prediction, the LSTM predictor achieved high accuracy (SMAPE>90%) for uplink throughput and RTT.

Overall, the dual-agent architecture proved robust and reproducible. By combining PHY, transport, and application measurements with AI forecasting, the platform went beyond benchmarking — it delivered predictive QoS intelligence.

6GMeasurOps can have significant technical and scientific impact, since: a) it demonstrated a portable, extensible measurement and prediction platform that works in real 5G NSA and SA networks; b) it Introduced a benchmarking + predictive QoS framework, bridging reactive and proactive network management. c) it created open-source software (QoSCOPE) for reproducible experiments; d) it provides validated datasets for future 6G research and contributions to 3GPP/ETSI standardisation.