AI-native connectivity for cooperative autonomous systems

Main Research Question (RQ)

How can programmable intelligent networks be designed to dynamically orchestrate AI workloads and to provide seamless connectivity across heterogeneous infrastructures (terrestrial and non-terrestrial), enabling adaptive and cooperative autonomous systems?

Abstract

The rise of autonomous systems, vehicles, drones, robots, and industrial agents, demands networks that are no longer just connected resources, but programmable and intelligent infrastructures. These systems increasingly rely on seamless interaction across heterogeneous domains, from terrestrial to non-terrestrial satellite and aerial networks – the so-called 6G networks- while simultaneously executing diverse heterogenous AI workloads for perception, negotiation, and verification. Existing approaches handle network handovers and AI workload management in isolation, resulting in inefficiencies, rigidity, and limited trust. This PhD breaks new ground by fusing workload orchestration and connectivity control inside programmable networks, creating a foundation for cooperative autonomy at scale.

Application domains

The novelty of this research lies in treating AI workloads and connectivity as co-optimized, programmable resources, rather than siloed domains, unlocking efficiency, resilience, and trust for next-generation autonomy. The business potential is significant: OEMs and telecom vendors gain domain-specific solutions for trusted cooperative vehicles and robots; defence and logistics sectors access resilient multi-domain connectivity; and chip and IP vendors benefit from new programmable architectures bridging silicon, AI, and networks. By embedding intelligence and adaptability directly into the network fabric, the project positions Europe at the forefront of AI-native, programmable 6G ecosystems, with clear exploitation pathways in standardization, IP licensing, and system integration.

PhD Outlook (under consideration)

Over a 3–4 year trajectory, the research will (1) characterize AI workloads and cross-domain connectivity needs in domains such as mobility, robotics, and defence; (2) design programmable control mechanisms that jointly manage workload distribution and network adaptation; (3) prototype intelligent handover and orchestration mechanisms that exploit programmability at both network and hardware levels; and (4) validate the concepts in real-world testbeds such as IMEC’s CityLab, SmartHighway, CCAM Proving Region, etc. Key milestones include demonstrating programmable terrestrial/non-terrestrial handovers, on-demand workload orchestration linked to network conditions, and embedding interoperable verification mechanisms into programmable infrastructures.

Sub-Research Questions (SRQs):

1. Workload–network orchestration: How can programmable networks dynamically allocate AI workloads across distributed compute and communication resources to optimize latency, energy, and reliability?
2. Cross-domain connectivity: What programmable mechanisms can enable seamless and adaptive handovers between terrestrial and non-terrestrial networks in support of autonomous systems?
3. Cooperation protocols: How can programmable control planes support interaction protocols among heterogeneous autonomous agents (e.g., detection, negotiation, contracting) in real time?
4. Interoperability & verification: What methods can be embedded into programmable infrastructures to ensure the integrity of exchanged information across multiple domains?
5. Performance–adaptability trade-offs: How do programmable intelligent networks balance performance (latency, throughput, energy efficiency) with adaptability and security in safety-critical applications?