K-AIM: KPI-Driven Adaptive Interference Management for 5G Coverage Enhancement
Abstract
Co-channel 5G new-radio deployments reuse the full band in every cell to maximize spectral efficiency, concentrating inter-cell interference at cell edges and limiting coverage. Fractional frequency reuse (FFR) and soft frequency reuse (SFR) help only when their static configuration matches the operating load, while threshold-based self-optimization leaves the state-action trade-off implicit. K-
Aim
is a KPI-driven adaptive interference management framework in which each cell infers its interference state from standard gNB counters and selects a reuse mode from an offline-calibrated state-mode cost table. A gradient-boosted detector maps a one-second KPI window to a distribution over three interference states. An outcome-aware decision rule converts this distribution into a mode action, and a hysteresis window stabilizes the loop. In system-level simulations of a 19-cell n78 network, K-
Aim
improves coverage over reuse-1 by 2.91 pp at medium load and 2.41 pp at high load. It remains within 0.3 pp of the strongest adaptive baseline and within 0.2 pp and 0.4 pp of the best static configuration, respectively. The 5%-ile SINR improves by 1.4 dB at both loads. The outcome-aware rule reduces offline coverage-loss cost by up to 39%, and a three-window hysteresis reduces switching by 64-65% relative to no hysteresis. The evaluation is limited to simulation and does not establish universal state-of-the-art or field-deployment superiority.