Edge-Aware RIS-Assisted Dynamic Channel Allocation With Lightweight LLM Decision Agent for Interference Mitigation in Low-Altitude Remote Sensing Networks

Low-altitude remote sensing networks are increasingly important for applications such as environmental monitoring, disaster response, infrastructure inspection, and real time sensing services. However, when many sensing nodes share limited spectrum resources, severe co-channel interference can degrade communication reliability and delay sensing-data delivery. This challenge becomes more critical in edge-enabled deployments, where control decisions must be made under strict latency, memory, and computational constraints. To address this issue, this paper proposes an LLM-enhanced Edge-aware Lightweight RIS-assisted Dynamic Channel Allocation (EL-RIS DCA) framework for interference mitigation in dense low-altitude remote sensing networks. The novelty of the proposed framework lies in a two-stage edge-control design: a lightweight large language model first generates a fast candidate decision for channel allocation and RIS phase adjustment from summarized network observations and retrieved historical patterns, and the EL-RIS-DCA module then performs feasibility verification, interference-aware refinement, and safe execution under edge constraints. Simulations are conducted using distance-based path loss with Rayleigh fading for direct and RIS-assisted cascaded links in a dense low-altitude sensing scenario with edge constrained operation. Simulation results show that the proposed method achieves higher average signal-to-interference-plus-noise ratio (SINR), lower outage probability, faster convergence, and near-optimal performance under limited computation budgets compared with the raw LLM proposal, the EL-RIS-DCA without LLM, and the considered non-RIS or centralized optimization baselines. The results also indicate that moderate RIS sizes can provide strong performance gains while keeping the computational complexity suitable for real-time edge deployment. Overall, the proposed framework offers an effective and practical solution for interference mitigation in dense and resource constrained low-altitude remote sensing networks.