AI/ML for RAN optimization: anomaly detection, traffic prediction, energy saving

RAN Anomaly Detection with ML

ML-based anomaly detection in the RAN identifies degraded cells, interference events, and equipment malfunctions faster than threshold-based alarms. Unsupervised models like autoencoders and isolation forests learn normal KPI patterns (RSRP, SINR, throughput, handover success rate) per cell and flag deviations. This catches subtle degradations like antenna tilt drift or passive intermodulation that traditional alarms miss. O-RAN xApps deployed in the near-RT RIC process cell-level KPIs every 10ms-1s, enabling rapid detection. The key challenge is distinguishing true anomalies from expected…