Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

FPGA-Based Neural Network Accelerators for Space Applications: A Survey

Pedro Antunes, Artur Podobas

TL;DR

This survey analyzes the state-of-the-art in FPGA-based neural network accelerators for onboard space applications, highlighting the dual architectural paradigms of dataflow and time-multiplexed designs and the dominance of CNNs for Earth-observation tasks. It consolidates 47 studies, detailing hardware platforms, quantization strategies, datasets, and performance/power metrics, and identifies gaps such as limited online unsupervised learning and sparse use of SNNs or 3D convolutions. The work emphasizes radiation tolerance, energy efficiency, and fault tolerance as critical constraints, and suggests directions like broader NN topologies, radiation-hardened designs, and standardized benchmarks to advance practical deployment. Overall, the paper provides a comprehensive, technically focused reference to guide researchers and practitioners in designing robust, efficient onboard AI accelerators for space missions.

Abstract

Space missions are becoming increasingly ambitious, necessitating high-performance onboard spacecraft computing systems. In response, field-programmable gate arrays (FPGAs) have garnered significant interest due to their flexibility, cost-effectiveness, and radiation tolerance potential. Concurrently, neural networks (NNs) are being recognized for their capability to execute space mission tasks such as autonomous operations, sensor data analysis, and data compression. This survey serves as a valuable resource for researchers aiming to implement FPGA-based NN accelerators in space applications. By analyzing existing literature, identifying trends and gaps, and proposing future research directions, this work highlights the potential of these accelerators to enhance onboard computing systems.

FPGA-Based Neural Network Accelerators for Space Applications: A Survey

TL;DR

This survey analyzes the state-of-the-art in FPGA-based neural network accelerators for onboard space applications, highlighting the dual architectural paradigms of dataflow and time-multiplexed designs and the dominance of CNNs for Earth-observation tasks. It consolidates 47 studies, detailing hardware platforms, quantization strategies, datasets, and performance/power metrics, and identifies gaps such as limited online unsupervised learning and sparse use of SNNs or 3D convolutions. The work emphasizes radiation tolerance, energy efficiency, and fault tolerance as critical constraints, and suggests directions like broader NN topologies, radiation-hardened designs, and standardized benchmarks to advance practical deployment. Overall, the paper provides a comprehensive, technically focused reference to guide researchers and practitioners in designing robust, efficient onboard AI accelerators for space missions.

Abstract

Space missions are becoming increasingly ambitious, necessitating high-performance onboard spacecraft computing systems. In response, field-programmable gate arrays (FPGAs) have garnered significant interest due to their flexibility, cost-effectiveness, and radiation tolerance potential. Concurrently, neural networks (NNs) are being recognized for their capability to execute space mission tasks such as autonomous operations, sensor data analysis, and data compression. This survey serves as a valuable resource for researchers aiming to implement FPGA-based NN accelerators in space applications. By analyzing existing literature, identifying trends and gaps, and proposing future research directions, this work highlights the potential of these accelerators to enhance onboard computing systems.

Paper Structure

This paper contains 62 sections, 2 equations, 11 figures.

Table of Contents

  1. Introduction
  2. Background
  3. Hardware Systems for Space Missions
  4. Reconfigurable Hardware Systems
  5. Neural Networks
  6. FPGA-based Neural Network Accelerators
  7. Literature Review
  8. Real-time runway detection in satellite images using multi-channel PCNN
  9. FPGA Architecture for Deep Learning and its application to Planetary Robotics
  10. FPGA Based Implementation of Convolutional Neural Network for Hyperspectral Classification
  11. An Exploration of FPGA based Multilayer Perceptron using Residue Number System for Space Applications
  12. Ship Classification from SAR Images Based on Deep Learning
  13. Efficient Convolutional Neural Network Weight Compression for Space Data Classification on Multi-fpga Platforms
  14. Artificial Intelligence Implementation on Voice Command and Sensor Anomaly Detection for Enhancing Human Habitation in Space Mission
  15. A Lightweight Hyperspectral Image Anomaly Detector for Real-Time Mission
  16. ...and 47 more sections

Figures (11)

  • Figure 1: Abstract comparison of CPU, GPU, FPGA, and NPU.
  • Figure 2: Tree representation of the different fault tolerance design techniques.
  • Figure 3: Abstract representation of the structure of an FPGA, showing logic blocks, programmable interconnect, and IO pad.
  • Figure 4: Dataflow architecture of an FPGA-based NN accelerator.
  • Figure 5: Time-multiplexed architecture of an FPGA-based NN accelerator.
  • ...and 6 more figures