Agentic AI-Based 5G and Beyond Radio Planning Framework


Tezcan M. G., Yazar A., Karahan S. N., Osmanca M. S., Altun H. O.

IEEE Access, cilt.14, ss.72395-72413, 2026 (SCI-Expanded, Scopus)

  • Yayın Türü: Makale / Tam Makale
  • Cilt numarası: 14
  • Basım Tarihi: 2026
  • Doi Numarası: 10.1109/access.2026.3691411
  • Dergi Adı: IEEE Access
  • Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Compendex, INSPEC, Directory of Open Access Journals
  • Sayfa Sayıları: ss.72395-72413
  • Anahtar Kelimeler: Planning, Optimization, Modeling, Automation, 5G mobile communication, Cognition, Cognitive systems, Open RAN, Architecture, Computer architecture, 5G, 6G, agentic AI, large language models, radio planning, small language models
  • Boğaziçi Üniversitesi Adresli: Evet

Özet

The evolution from Fifth Generation (5G) to Sixth Generation (6G) wireless networks introduces diverse service requirements and increases radio-planning complexity. Traditional planning approaches based on static propagation models and manual optimization struggle to adapt to dense and dynamic network environments. This paper proposes an Agentic AI-based radio planning framework for 5G and beyond networks. The proposed architecture employs a multi-agent system in which specialized agents collaborate to perform intent interpretation, data aggregation, analysis, simulation, optimization, and verification tasks within a unified planning workflow. Language Models are used as reasoning engines that orchestrate domain-specific planning tools and heterogeneous data sources. The framework integrates operator intents, network measurements, geographic information, and demand-related data through interface, agent, tool, and data layers. The planning problem is formalized as a multi-objective optimization task balancing coverage, demand satisfaction, and energy efficiency. Quantitative evaluations, including preliminary simulations, control-level analysis, operator-inspired deployment scenarios, baseline comparisons, and reasoning-accuracy assessment, demonstrate the practical relevance of the proposed framework. In addition, the framework is extended to support emerging domains such as Internet of Vehicles (IoV) environments and proactive sidelink collision management in Vehicle-to-Everything (V2X) systems. The proposed approach demonstrates how agentic architectures can support scalable, explainable, and progressively autonomous radio planning for next-generation wireless networks.