Deployable Hierarchical ML Traffic Steering for O-RAN RICS

Current mobile networks often fail to achieve their full performance potential, as user devices commonly experience data rates well below their possible capabilities. This limitation arises from conventional cell selection mechanisms, which often fail to identify and utilize higher-performing candidate cells, resulting in suboptimal use of network resources. In this paper, we introduce a hierarchical machine learning-based traffic steering solution for O-RAN-enabled 5G and beyond networks. The proposed solution supports network-assisted cell selection, guiding user devices toward more optimal cells. The proposed solution is developed in accordance with O-RAN Alliance design principles and operates without any modifications to existing 3GPP signaling or to mobile devices. Experimental evaluations on a small-scale testbed demonstrate that the proposed approach can improve the median throughput by up to 75% in certain traffic scenarios while also improving overall network fairness.