Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Heterogeneity: An Open Challenge for Federated On-board Machine Learning

Maria Hartmann, Grégoire Danoy, Pascal Bouvry

TL;DR

This paper addresses the open problem of cross-provider Federated Learning in Orbital Edge Computing, where heterogeneous satellites from multiple providers must collaborate on-board while facing data, device, feature, and preference heterogeneity. It conducts a systematic survey of the state-of-the-art in orbital FL, categorizing approaches by server placement and aggregation strategy, and then analyzes heterogeneity through data-, device-, feature-, and preference-based lenses, drawing on broader FL literature for potential mitigations. It highlights fairness, security (malicious participants), and standardisation as critical non-technical enablers, and offers a gap analysis and entry points for future work. The work's significance lies in guiding the design of robust, interoperable cross-provider space ML systems that can operate under ad-hoc collaborations and diverse hardware and mission constraints.

Abstract

The design of satellite missions is currently undergoing a paradigm shift from the historical approach of individualised monolithic satellites towards distributed mission configurations, consisting of multiple small satellites. With a rapidly growing number of such satellites now deployed in orbit, each collecting large amounts of data, interest in on-board orbital edge computing is rising. Federated Learning is a promising distributed computing approach in this context, allowing multiple satellites to collaborate efficiently in training on-board machine learning models. Though recent works on the use of Federated Learning in orbital edge computing have focused largely on homogeneous satellite constellations, Federated Learning could also be employed to allow heterogeneous satellites to form ad-hoc collaborations, e.g. in the case of communications satellites operated by different providers. Such an application presents additional challenges to the Federated Learning paradigm, arising largely from the heterogeneity of such a system. In this position paper, we offer a systematic review of these challenges in the context of the cross-provider use case, giving a brief overview of the state-of-the-art for each, and providing an entry point for deeper exploration of each issue.

Heterogeneity: An Open Challenge for Federated On-board Machine Learning

TL;DR

This paper addresses the open problem of cross-provider Federated Learning in Orbital Edge Computing, where heterogeneous satellites from multiple providers must collaborate on-board while facing data, device, feature, and preference heterogeneity. It conducts a systematic survey of the state-of-the-art in orbital FL, categorizing approaches by server placement and aggregation strategy, and then analyzes heterogeneity through data-, device-, feature-, and preference-based lenses, drawing on broader FL literature for potential mitigations. It highlights fairness, security (malicious participants), and standardisation as critical non-technical enablers, and offers a gap analysis and entry points for future work. The work's significance lies in guiding the design of robust, interoperable cross-provider space ML systems that can operate under ad-hoc collaborations and diverse hardware and mission constraints.

Abstract

The design of satellite missions is currently undergoing a paradigm shift from the historical approach of individualised monolithic satellites towards distributed mission configurations, consisting of multiple small satellites. With a rapidly growing number of such satellites now deployed in orbit, each collecting large amounts of data, interest in on-board orbital edge computing is rising. Federated Learning is a promising distributed computing approach in this context, allowing multiple satellites to collaborate efficiently in training on-board machine learning models. Though recent works on the use of Federated Learning in orbital edge computing have focused largely on homogeneous satellite constellations, Federated Learning could also be employed to allow heterogeneous satellites to form ad-hoc collaborations, e.g. in the case of communications satellites operated by different providers. Such an application presents additional challenges to the Federated Learning paradigm, arising largely from the heterogeneity of such a system. In this position paper, we offer a systematic review of these challenges in the context of the cross-provider use case, giving a brief overview of the state-of-the-art for each, and providing an entry point for deeper exploration of each issue.
Paper Structure (9 sections, 1 figure)

This paper contains 9 sections, 1 figure.

Table of Contents

  1. Introduction
  2. Survey results
  3. Orbital edge computing and federated learning
  4. Heterogeneity challenges of cross-provider FL
  5. Other considerations
  6. Fairness
  7. Guarding against malicious participants
  8. The role of standardisation
  9. Discussion

Figures (1)

  • Figure 1: In Federated Learning, each satellite performs on-board machine learning to train a local model (1). Only these models are transmitted via satellite link to a server (2), here based on the ground, where multiple local models are aggregated into a single global model (3). This global model is transmitted back to the satellites (4) to continue the learning process. If necessary, satellites can act as relays for one another (5).