Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Semantic and goal-oriented edge computing for satellite Earth Observation

Beatriz Soret, Israel Leyva-Mayorga, Antonio M. Mercado-Martínez, Marco Moretti, Antonio Jurado-Navas, Marc Martinez-Gost, Celia Sánchez de Miguel, Ainoa Salas-Prendes, Petar Popovski

TL;DR

This work tackles the data deluge and resource constraints in satellite Earth Observation by integrating semantic communications with edge computing in a LEO constellation. It proposes a satellite-edge architecture where on-board semantic encoding and distributed edge processing reduce data transmission and enable real-time task execution for image reconstruction and object detection. The framework formalizes semantic and goal-oriented metrics, and demonstrates energy-time-accuracy trade-offs through simulations using real EO parameters and YOLOv8, showing when edge or hybrid edge-cloud processing is advantageous. The approach promises scalable, low-latency EO intelligence with reduced feeder-link congestion, applicable to diverse tasks such as vessel detection and tracking in dynamic environments.

Abstract

The integration of Semantic Communications (SemCom) and edge computing in space networks enables the optimal allocation of the scarce energy, computing, and communication resources for data-intensive applications. We use Earth Observation (EO) as a canonical functionality of satellites and review its main characteristics and challenges. We identify the potential of the space segment, represented by a low Earth orbit (LEO) satellite constellation, to serve as an edge layer for distributed intelligence. Based on that, propose a system architecture that supports semantic and goal-oriented applications for image reconstruction and object detection and localization. The simulation results show the intricate trade-offs among energy, time, and task-performance using a real dataset and State-of-the-Art (SoA) processing and communication parameters.

Semantic and goal-oriented edge computing for satellite Earth Observation

TL;DR

This work tackles the data deluge and resource constraints in satellite Earth Observation by integrating semantic communications with edge computing in a LEO constellation. It proposes a satellite-edge architecture where on-board semantic encoding and distributed edge processing reduce data transmission and enable real-time task execution for image reconstruction and object detection. The framework formalizes semantic and goal-oriented metrics, and demonstrates energy-time-accuracy trade-offs through simulations using real EO parameters and YOLOv8, showing when edge or hybrid edge-cloud processing is advantageous. The approach promises scalable, low-latency EO intelligence with reduced feeder-link congestion, applicable to diverse tasks such as vessel detection and tracking in dynamic environments.

Abstract

The integration of Semantic Communications (SemCom) and edge computing in space networks enables the optimal allocation of the scarce energy, computing, and communication resources for data-intensive applications. We use Earth Observation (EO) as a canonical functionality of satellites and review its main characteristics and challenges. We identify the potential of the space segment, represented by a low Earth orbit (LEO) satellite constellation, to serve as an edge layer for distributed intelligence. Based on that, propose a system architecture that supports semantic and goal-oriented applications for image reconstruction and object detection and localization. The simulation results show the intricate trade-offs among energy, time, and task-performance using a real dataset and State-of-the-Art (SoA) processing and communication parameters.
Paper Structure (12 sections, 4 figures, 1 table)

This paper contains 12 sections, 4 figures, 1 table.

Table of Contents

  1. Introduction
  2. Fundamentals of Earth Observation
  3. Data acquisition
  4. Meteorological effects
  5. Data processing
  6. LEO satellite constellations as an edge layer
  7. Semantic and goal-oriented
  8. Architecture and general procedure
  9. Metrics
  10. Energy-time-accuracy trade-offs in
  11. Performance evaluation
  12. Conclusions

Figures (4)

  • Figure 1: Sketch of vessel detection and localization.
  • Figure 2: Impact of the atmospheric turbulence in ship detection and localization using YOLOv8. The output of the algorithm are the blue bounding boxes that include coordinates and confidence about the detection and its performance is measured by the recall: the ratio of correctly detected objects to all actual objects.
  • Figure 3: Procedure of the proposed semantic and edge computing empowered .
  • Figure 4: Minimum power consumption in logarithmic scale versus number of for and computing architectures.