About


What is the CELESTE project about?

The vision of CELESTE is to enable intelligent, efficient, and resilient resource orchestration in future Three-Dimensional (3D) networks that seamlessly integrate terrestrial, airborne, and spaceborne infrastructures. To deliver realistic and practical solutions, CELESTE embraces the inherent uncertainty of these environments, which arises from mobility, stochastic and time correlated dynamics, intermittent resource availability, and fluctuating service demands. To address these challenges, the project moves beyond static and centralized control paradigms and envisions a holistic framework based on distributed optimization algorithms with limited information exchange and near real time operation. Through this approach, CELESTE aims to identify operational regimes that balance efficiency and resilience, ultimately supporting reliable communication and computing services in highly dynamic 6G network scenarios.

Objectives

To achieve its goal, CELESTE targets three dimensions:

Conference Topics

CELESTE will thoroughly analyze and accurately model the inherent system uncertainties with particular emphasis on the (a) stochastic and time-varying wireless channel modeling due to ground device or airborne node mobility, (b) stochastic computing service request arrival rate, and (c) long-term constraint of battery energy backup of the devices on the ground or the airborne nodes. These uncertainty factors will be tackled both separately and jointly, serving as a basis for efficient resource orchestration.

CELESTE will introduce a learning approach, according to which the individual decision making entities will learn their most beneficial strategies in terms of the resource orchestration and computation offloading by relying solely on their past actions and the respective rewards of their satisfactory strategies.

CELESTE will scrutinize efficient and resilient operational points, easily derived from the designed distributed resource orchestration algorithms, such as the generalized NE points expressed as probability distributions on action profiles. This concludes a wider space of possible solution outcomes, offering the required flexibility given the dynamicity of the underlying 3D network.

Project Originality

CELESTE advances beyond existing approaches by developing distributed optimization frameworks that are inherently resilient to the incomplete and imperfect information that characterizes real 3D network environments. Unlike conventional solutions that rely on centralized control or full system knowledge, CELESTE explicitly tackles uncertainty, mobility, and large-scale operation, addressing challenges that are often overlooked in current research. A key innovation of the project is the shift from strict Nash Equilibrium (NE) solutions to more flexible equilibrium concepts, such as Correlated Equilibria (CE), which offer a richer and more flexible set of operating points suited to the dynamic nature of 3D networks. To support practical implementation, CELESTE explores regret-based Reinforcement Learning (RL) methods that allow network entities to adapt rapidly to changing conditions by minimizing instantaneous regret rather than optimizing long-term objectives. These methods will be applied and validated in emerging 3D network resource orchestration scenarios that jointly consider radio and computing resource allocation, while also laying the foundations for broader distributed optimization and control problems under uncertainty.


Expected Results & Research Project Impact

CELESTE is expected to deliver research outcomes that support the practical realization of seamlessly integrated 3D networks. Owing to the adaptability, flexibility and autonomous capabilities of the CELESTE framework, the vision of 3D and self-managed 6G networks will be grounded in a solid theoretical framework. CELESTE will not only etsablish this framework but also has the potential to influence industrial approaches to next-generation network technologies. Its outcomes can guide the evolution of network architectures and protocols toward more efficient, resilient, and energy-aware operation, ultimately benefiting network providers through reduced operational complexity and end-users through improved service reliability and quality.