Dynamic Interference Management for TN-NTN Coexistence in the Upper Mid-Band

The coexistence of terrestrial networks (TN) and non-terrestrial networks (NTN) in the frequency range 3 (FR3) upper mid-band presents considerable interference concerns, as dense TN deployments can severely degrade NTN downlink performance. Existing studies rely on interference-nulling beamforming, precoding, or exclusion zones that require accurate channel state information (CSI) and static coordination, making them unsuitable for dynamic NTN scenarios. To overcome these limitations, we develop an optimization framework that jointly controls TN downlink power, uplink power, and antenna downtilt to protect NTN links while preserving terrestrial performance. The resultant non-convex coupling between TN and NTN parameters is addressed by a Proximal Policy Optimization (PPO)-based reinforcement learning method that develops adaptive power and tilt control strategies. Simulation results demonstrate a reduction up to 8 dB in the median interference-to-noise ratio (INR) while maintaining over 87% TN basestation activity, outperforming conventional baseline methods and validating the feasibility of the proposed strategy for FR3 coexistence.

💡 Research Summary

**
The paper addresses the emerging challenge of spectrum coexistence in the upper mid‑band (7–24 GHz, FR3) where dense terrestrial 5G/6G networks (TN) share the same frequency resources with non‑terrestrial networks (NTN), specifically low‑Earth‑orbit (LEO) satellite systems. Existing mitigation techniques—such as interference‑nulling beamforming, advanced precoding, or static exclusion zones—depend on accurate channel state information (CSI) and assume relatively static satellite positions, making them unsuitable for the highly dynamic environment of moving satellites and mobile NTN user terminals.

To overcome these limitations, the authors formulate a joint optimization problem that simultaneously controls three key TN parameters: downlink transmit power of each gNB, uplink transmit power of user equipment, and the mechanical downtilt angle of each gNB antenna. The objective is to maximize the aggregate NTN downlink throughput while ensuring that the interference‑to‑noise ratio (INR) experienced by every NTN terminal stays below a protection threshold (typically –12 dB to –6 dB). Constraints also enforce realistic power limits, downtilt ranges, and a minimum activity level for the terrestrial network.

The system model follows 3GPP specifications. NTN path loss incorporates free‑space loss, atmospheric, ionospheric, and tropospheric losses, as well as clutter and shadow fading, based on the slant distance derived from satellite altitude and elevation angle. NTN antenna gain follows the Bessel‑function pattern defined in TR 38.811. TN propagation uses the Urban Macro (UMa) model from TR 38.901, with separate LOS/NLOS formulas and a height‑dependent breakpoint. gNB antenna gain is expressed as a function of horizontal and vertical angles and the downtilt angle, capturing how vertical beam shaping can reduce upward leakage toward satellites.

Because the resulting optimization is highly non‑convex, non‑differentiable, and couples continuous (powers, angles) and discrete (sector muting) variables, the authors resort to a reinforcement‑learning (RL) solution. They adopt Proximal Policy Optimization (PPO), a state‑of‑the‑art policy‑gradient algorithm, and implement it as a centralized agent residing in the core network.

At each decision epoch, the PPO agent observes a state vector comprising: satellite elevation angle, the fraction of NTN terminals whose INR exceeds the current threshold, the current aggregate NTN throughput, the proportion of active gNBs, and the presently used INR protection threshold. The action vector includes: (i) an updated INR threshold, (ii) a binary muting decision for each gNB sector, (iii) the downlink power for each gNB, (iv) the uplink power for each UE, and (v) the downtilt angle for each gNB. The immediate reward is a weighted sum of normalized NTN throughput, terrestrial activity level, and a penalty proportional to the fraction of interfered NTN terminals, encouraging the agent to prioritize satellite performance while preserving terrestrial service.

Simulation scenarios model a 20 × 20 km² footprint fully covered by a single LEO satellite at 600 km altitude, populated with multiple 5G gNBs and mobile NTN user terminals. The authors compare the PPO‑derived adaptive policy against three baselines: (1) static power and tilt settings, (2) dynamic exclusion‑zone muting based on satellite trajectory, and (3) interference‑nulling beamforming. Results show that the PPO approach reduces median INR by 6–8 dB relative to the baselines and maintains over 87 % of gNBs active, thereby preserving terrestrial capacity. The inclusion of sector muting as a discrete control further refines interference suppression in dense urban deployments, allowing selective silencing of the most problematic sectors without a blanket shutdown.

Key contributions of the work are:

1. A comprehensive analytical model that captures the coupled impact of TN transmit power, uplink power, and antenna downtilt on NTN interference.
2. A novel application of PPO‑based reinforcement learning to solve the resulting non‑convex joint optimization in real time.
3. Demonstrated performance gains in realistic FR3 coexistence scenarios, validating the practicality of adaptive power‑tilt‑muting control for future integrated terrestrial‑satellite networks.

The paper suggests future extensions such as multi‑agent collaborative learning for multi‑satellite, multi‑band environments, incorporation of CSI uncertainty and measurement errors in the learning loop, and field trials to verify the approach under real‑world propagation conditions. These directions aim to further enhance spectrum efficiency and service reliability as 5G/6G and NTN converge in the next generation of wireless communications.