From edge computing to satellite constellations, researchers gathered at ESILV to discuss how services can be deployed and optimised across increasingly distributed infrastructures. The seminar highlighted emerging research on Orbital Cloud, a field that extends cloud computing capabilities into space.
As computing infrastructures evolve beyond traditional data centres, researchers are investigating new ways to deploy services across networks spanning the cloud, the edge, and even space. These questions were at the centre of a research seminar organised by the Cloud Optimisation Working Group at ESILV, entitled Service Optimisation and Orchestration across the Computing Continuum.
The event brought together researchers from ESILV, Télécom SudParis and Université Paris Nanterre to examine current challenges in distributed computing, service placement and optimisation.
ESILV research examines computing infrastructures in orbit
The seminar opened with a presentation by Farah Aït Salaht (De Vinci Research Center / ESILV) entitled From Cloud to Constellation: Orchestrating Services at the Edge of Space.
The talk focused on the emergence of Orbital Cloud, a concept made possible by the rapid expansion of Low Earth Orbit (LEO) satellite constellations. In this model, satellites become more than communication platforms. They can host services, process data and participate directly in distributed computing systems.
Moving computing resources into orbit introduces new scientific and engineering questions. Service placement, scheduling strategies, resource allocation and autonomous decision-making must account for the continuous movement of satellites, intermittent visibility windows and constrained onboard resources.
The presentation reviewed recent advances in the field while identifying several research challenges that remain open. These preliminary investigations are expected to contribute to future research projects currently under development at ESILV and the De Vinci Research Center.
Orbital Cloud: From cloud to constellation
A research topic connected to aerospace engineering
Orbital Cloud research sits at the intersection of cloud computing, networking, optimisation and aerospace systems.
For engineering students, these developments reflect the growing convergence between digital infrastructures and space technologies. Satellite constellations are becoming platforms capable of supporting data-intensive applications, artificial intelligence workloads and distributed services.
These topics resonate with the expertise developed within ESILV’s Aeronautical & Aerospace Engineering programmes, where students study complex systems, modelling, optimisation methods and emerging technologies used across the aviation and space sectors.
Improving federated learning through network-aware deployment
The seminar also featured a presentation by Nour El-Houda Yellas (Télécom SudParis / SAMOVAR) on Function Placement for In-network Federated Learning.
Federated Learning allows multiple participants to collaboratively train machine learning models while keeping data locally stored. The approach can reduce data transfers and improve privacy, but heterogeneous computing resources and network conditions often affect performance.
The proposed In-network Federated Learning Control (IFLC) framework addresses these limitations by adapting the deployment of learning functions and hardware accelerators across distributed networks. The objective is to reduce latency variations and improve overall learning efficiency while controlling deployment costs.
Reinforcement learning for edge computing orchestration
Emmanuel Hyon (Université Paris Nanterre / LIP6) presented research on Faster Latency Constrained Service Placement in Edge Computing with Deep Reinforcement Learning.
As applications increasingly rely on edge infrastructures, determining where services should be deployed becomes a complex optimisation problem. The research combines mathematical optimisation techniques with reinforcement learning methods to identify efficient placement strategies while respecting latency constraints.
The approach contributes to the development of orchestration mechanisms that dynamically adapt to changing network conditions and resource availability.
Research at the crossroads of cloud, AI and aerospace systems
The seminar highlighted how distributed computing infrastructures are expanding across multiple environments, from terrestrial networks to edge platforms and satellite constellations.
The questions addressed by the speakers reflect a broader transformation of digital infrastructures, where optimisation, artificial intelligence and networking increasingly interact with aerospace technologies and autonomous systems.