AI-Native Control and Optimization of Integrated Terrestrial and Non-Terrestrial Networks

Future 5G-Advanced and 6G networks will integrate Terrestrial Networks (TNs) and Non-Terrestrial Networks (NTNs) to provide ubiquitous, reliable, and resilient connectivity. By combining terrestrial infrastructure with satellites and high-altitude platforms, integrated TN-NTN systems can extend connectivity beyond the reach of conventional terrestrial networks and support reliable communications in remote and infrastructure-sparse areas. They can also maintain service when terrestrial infrastructure is unavailable, damaged, disrupted, or under attack, which is critical for defence and security operations in contested and remote environments. Therefore, this PhD topic is strongly aligned with imec’s Aerospace and Security sector by addressing the need for secure, resilient, and autonomous connectivity across terrestrial and space-based infrastructures.

However, managing such integrated networks is challenging due to long propagation delays, satellite mobility, dynamic beam coverage, limited onboard resources, and heterogeneous network conditions.

This PhD will investigate novel AI-driven approaches for the autonomous operation and optimization of integrated TN-NTN networks. The research will explore how advanced AI paradigms, including Agentic AI and Wireless Physical-Layer Foundation Models (WPFMs), can enable scalable network awareness, reasoning, and decision-making across satellites, gateways, edge clouds, and terrestrial infrastructures [1][2]. The goal is to develop intelligent systems that continuously learn from their environment and autonomously adapt network operations to changing traffic, channel, mobility, interference, and operational conditions. Such autonomous adaptation will contribute to maintaining resilient and reliable connectivity in dynamic, disrupted, or contested environments.

A key focus of the PhD will be the design of multi-timescale AI architectures for TN-NTN systems. Different network decisions operate under very different latency constraints. Ultra-fast functions, such as radio resource scheduling, may require decisions within milliseconds, while beam management and user-beam association operate on slightly longer timescales. In contrast, network policy optimization, traffic engineering, and model lifecycle management evolve over much longer periods. The research will investigate how these functions should be distributed across satellites, gateways, edge clouds, and terrestrial infrastructure to jointly optimize performance, scalability, resilience, and operational efficiency.

The framework will address functional split architectures, intelligence placement, resource allocation, beam management, scheduling, mobility management, traffic steering, spectrum awareness, and TN-NTN dual connectivity. Particular attention will be paid to identifying which latency-critical functions should be executed onboard regenerative satellites and which functions should remain on the ground. This includes exploring how scheduling, user-beam association, power allocation, and resource allocation can benefit from onboard intelligence, while long-term policy optimization, model adaptation, and network-wide coordination are performed by edge- or ground-based controllers.

To enable practical deployment, the PhD will also investigate hardware-aware AI for satellite platforms, including model compression, pruning, quantization, knowledge distillation, and efficient adaptation of foundation models. The objective is to develop lightweight AI solutions that satisfy the strict computation, memory, energy, and latency constraints of future satellite systems while maintaining high decision quality.

The developed solutions will be evaluated through simulation and experimental validation using the IDLab 5G/6G research testbed [3], connected to a simulated or emulated NTN environment. This will enable the realistic assessment of AI-native TN-NTN architectures and provide practical guidelines for future 6G deployments, including aerospace, defence, security, emergency-response, and mission-critical communication scenarios.

Objectives:

• Design AI-native and multi-timescale architectures for autonomous TN-NTN management.
• Develop and leverage Wireless Physical-Layer Foundation Models for spectrum monitoring, localization, and inteference detection.
• Investigate intelligent placement and functional splitting of sensing, reasoning, learning, and control functions across satellites, gateways, edge clouds, and terrestrial infrastructure.
• Develop AI-driven mechanisms for beam management, radio resource allocation, scheduling, mobility management, and traffic steering.
• Investigate TN-NTN dual-connectivity mechanisms for seamless mobility, load balancing, reliability enhancement, and service continuity.
• Develop hardware-aware AI techniques, including quantization, pruning, knowledge distillation, and efficient foundation-model adaptation for satellite deployment.
• Evaluate the proposed solutions using simulation platforms and the IDLab 5G/6G testbed integrated with an NTN emulator.

Expected Results:

• A novel AI-native framework for autonomous TN-NTN operation and control.
• Multi-timescale architectures for distributing intelligence across satellites, gateways, edge clouds, and terrestrial infrastructure.
• AI-driven algorithms for beam management, resource allocation, scheduling, mobility management, and dual connectivity.
• Lightweight and deployable AI models optimized for satellite hardware constraints.
• Improved adaptability, scalability, reliability, and resource efficiency under dynamic wireless environments.
• Quantified trade-offs among network performance, control latency, computational complexity, memory footprint, energy consumption, and signaling overhead.
• Experimental validation using the IDLab 5G/6G testbed and an emulated NTN environment.
• High-impact scientific publications, reusable software components, datasets, and open research artifacts.

[1] Fontaine, J., Shahid, A., & De Poorter, E. (2024, June). Towards a wireless physical-layer foundation model: Challenges and strategies. In 2024 IEEE International Conference on Communications Workshops (ICC Workshops) (pp. 1-7). IEEE.

[2] Navidan, H., Cheraghinia, M., Fontaine, J., Seif, M., De Poorter, E., Poor, H. V., ... & Shahid, A. (2026). Toward autonomous O-RAN: A multi-scale agentic AI framework for real-time network control and management. IEEE Network.

[3] https://idlab.ugent.be/resources/cutting-edge-5g-6g-research-testbed