Towards the Use of Artificial Intelligence on the Edge in Space Systems: Challenges and Opportunities
The DOI matches its Crossref record; 43 of the 48 references it cites were found in a registry or an index.
Abstract
The market for remote sensing space-based applications is fundamentally limited by up- and downlink bandwidth and onboard compute capability for space data handling systems. This article details how the compute capability on these platforms can be vastly increased by leveraging emerging commercial off-the-shelf (COTS) system-on-chip (SoC) technologies. The orders of magnitude increase in processing power can then be applied to consuming data at source rather than on the ground allowing the deployment of value-added applications in space, which consume a tiny fraction of the downlink bandwidth that would be otherwise required. The proposed solution has the potential to revolutionize Earth observation (EO) and other remote sensing applications, reducing the time and cost to deploy new added value services to space by a great extent compared with the state of the art. This article also reports the first results in radiation tolerance and power/performance of these COTS SoCs for space-based applications and maps the trajectory toward low Earth orbit trials and the complete life-cycle for space-based artificial intelligence classifiers on orbital platforms and spacecraft.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
In the citing paper
“Secondly, these models (or certain DL layers) may not be supported [125] by the AI accelerators used in current space systems like FPGA-based [126,127] accelerators.”
From A survey on deep learning-based monocular spacecraft pose estimation: Current state, limitations and prospects (2023) · cites this paper as [126]
In the citing paper
“However, most existing algorithms are tested on workstations and large server clusters and very limited evaluations have been conducted on edge systems with FPGA [126,127] or GPU [174][175][176]-based AI accelerators for space applications.”
From A survey on deep learning-based monocular spacecraft pose estimation: Current state, limitations and prospects (2023) · cites this paper as [126]
Citations
Cites
Edge computing: Vision and challenges
Shi, W., Cao, J., Zhang, Q., Li, Y., & Xu, L. (2016) · IEEE Internet of Things Journal · doi:10.1109/jiot.2016.2579198 · Cited by 8,733 · cited 1×
“…neural networks (DNNs) on board spacecraft, analyzing the possible benefits in terms of bandwidth downlink due to the on-the-edge processing of data [5]: a novel AI-based architecture for EO satellites is proposed, which embeds AI algorithms for consuming data at source rather than on the ground,…”
BackgroundSentinel-2: ESA's Optical High-Resolution Mission for GMES Operational Services
Drusch, M., Del Bello, U., Carlier, S., Colin, O., Fernandez, V., Gascon, F., Hoersch, B., Isola, C., Laberinti, P., Martimort, P., et al. (2012) · Remote Sensing of Environment · doi:10.1016/j.rse.2011.11.026 · Cited by 4,815 · cited 2×
“[44] M.”
MethodIn-Datacenter Performance Analysis of a Tensor Processing Unit
Jouppi, N. P., Young, C., Patil, N., Patterson, D., Agrawal, G., Bajwa, R., Bates, S., Bhatia, S., Boden, N., Borchers, A., et al. (2017) · Proceedings of the 44th Annual International Symposium on Computer Architecture · doi:10.1145/3079856.3080246 · Cited by 4,537
Deep Learning in Remote Sensing: A Comprehensive Review and List of Resources
Zhu, X. X., Tuia, D., Mou, L., Xia, G.-S., Zhang, L., Xu, F., & Fraundorfer, F. (2017) · IEEE Geoscience and Remote Sensing Magazine · doi:10.1109/mgrs.2017.2762307 · Cited by 3,400
Convergence of Edge Computing and Deep Learning: A Comprehensive Survey
Wang, X., Han, Y., Leung, V. C. M., Niyato, D., Yan, X., & Chen, X. (2020) · IEEE Communications Surveys & Tutorials · doi:10.1109/comst.2020.2970550 · Cited by 1,498 · cited 1×
“The COTS hardware solutions for DNN inference on the edge can be classified based on their core computing unit [26]: 1) ASIC-based hardware that integrates custom AI chips; 2) GPU-based hardware that uses the massive Towards the Use of Artificial Intelligence on the Edge in Space Systems: Challenges and Opportunities”
MethodFINN- R: An End-to-End Deep-Learning Framework for Fast Exploration of Quantized Neural Networks
Blott, M., Preußer, T. B., Fraser, N. J., Gambardella, G., O’brien, K., Umuroglu, Y., Leeser, M., & Vissers, K. (2018) · ACM Transactions on Reconfigurable Technology and Systems · doi:10.1145/3242897 · Cited by 446 · cited 5×
“Furthermore, quantization and pruning techniques can be efficiently applied to compress the model [35], even improving its accuracy compared to the original one [32].”
BackgroundMethodInfluentialA survey and assessment of the capabilities of Cubesats for Earth observation
Selva, D., & Krejci, D. (2012) · Acta Astronautica · doi:10.1016/j.actaastro.2011.12.014 · Cited by 430 · cited 3×
“In this scenario, one of the most significant limiting factors for the remote sensing market, especially for miniaturized spacecraft, is represented by the up-and downlink bandwidth and onboard compute capability on satellites combined with ground station availability [3], [4].”
BackgroundA survey of FPGA-based accelerators for convolutional neural networks
Mittal, S. (2020) · Neural Computing and Applications · doi:10.1007/s00521-018-3761-1 · Cited by 429 · cited 2×
“For more complex DNNs, solutions relying on external memories to store coefficients and intermediate results are generally exploited [33].”
BackgroundModern Small Satellites-Changing the Economics of Space
Sweeting, M. N. (2018) · Proceedings of the IEEE · doi:10.1109/jproc.2018.2806218 · Cited by 382 · cited 3×
“The improved computing capabilities have increased the scope of possible applications Earth observation (EO), while the miniaturization of processing platforms has permitted to shrink the size and mass of spacecraft, enabling the production of small but capable satellites, in mass classes well below the 10 Kg (SmallSa…”
BackgroundSpacecraft Systems Engineering
(2011) · Wiley eBooks · doi:10.1002/9781119971009 · Cited by 342
The impact of new technology on soft error rates
Dixit, A., & Wood, A. (2011) · 2011 International Reliability Physics Symposium · doi:10.1109/irps.2011.5784522 · Cited by 299 · cited 2×
“On the other hand, flop error rates decrease as technology nodes decrease down to at least 28 nm [48].”
BackgroundOrbital Edge Computing: Nanosatellite Constellations as a New Class of Computer System
Denby, B., & Lucia, B. (2020) · Proceedings of the Twenty-Fifth International Conference on Architectural Support for Programming Languages and Operating Systems · doi:10.1145/3373376.3378473 · Cited by 273 · cited 7×
“Such an approach is called “bent-pipe” communication paradigm, which involves sending the data to the ground following a command transmitted by the specific ground station [39].”
BackgroundMethodInfluentialDeep Neural Network Approximation for Custom Hardware: Where We've Been, Where We're Going
Wang, E., Davis, J. J., Zhao, R., Ng, H.-C., Niu, X., Luk, W., Cheung, P. Y. K., & Constantinides, G. A. (2020) · ACM Computing Surveys · doi:10.1145/3309551 · Cited by 166
A digital CMOS design technique for SEU hardening
Baze, M. P., Buchner, S. P., & McMorrow, D. (2000) · IEEE Transactions on Nuclear Science · doi:10.1109/23.903815 · Cited by 122 · cited 2×
“Micro-semi FPGAs exploit a flash-based design to increase TID tolerance and a specific CMOS design to guarantee high SEEs resistance [11].”
MethodHigh-Performance Embedded Computing in Space: Evaluation of Platforms for Vision-Based Navigation
Lentaris, G., Maragos, K., Stratakos, I., Papadopoulos, L., Papanikolaou, O., Soudris, D., Lourakis, M., Zabulis, X., Gonzalez-Arjona, D., & Furano, G. (2018) · Journal of Aerospace Information Systems · doi:10.2514/1.i010555 · Cited by 120 · cited 3×
“Space-qualified devices shall be designed to be tolerant to deep vacuum conditions, strong vibrations, and higher temperature ranges, to mitigate electrostatic discharges risks, and to guarantee sufficient heat dissipation exclusively through thermal conduction [6], [7].”
BackgroundSelf-citationAlways-on Vision Processing Unit for Mobile Applications
Barry, B., Brick, C., Connor, F., Donohoe, D., Moloney, D., Richmond, R., O'Riordan, M., & Toma, V. (2015) · IEEE Micro · doi:10.1109/mm.2015.10 · Cited by 106 · cited 3×
“Myriad 2 is a system-in-package (SiP) device that contains a 28-nm SoC die fabricated in a standard LPCMOS TSMC manufacturing process along with a stacked 512 MB Low Power Double Data Rate (LPDDR)3 dynamic random access memory (DRAM) die [31].”
MethodInfluentialSelf-citationThe Final Frontier: Deep Learning in Space
Kothari, V., Liberis, E., & Lane, N. D. (2020) · Proceedings of the 21st International Workshop on Mobile Computing Systems and Applications · doi:10.1145/3376897.3377864 · Cited by 86 · cited 6×
“[24] V.”
BackgroundMethodInfluentialA Survey of AI Accelerators for Edge Environment
Li, W., & Liewig, M. (2020) · Advances in intelligent systems and computing · doi:10.1007/978-3-030-45691-7_4 · Cited by 70 · cited 3×
“These accelerators target mobile devices (e.g., Qualcomm’s Snapdragon 855, HiSilicon’s Kirin 970, and MediaTek’s Helio P90), IoT edge applications (e.g., AI AM5729 of BeagleBone) or the general use case (e.g., Intel’s Myriad 2 and Google’s Coral) [28].”
BackgroundEarly Identification of Oil Spills in Satellite Images Using Deep CNNs
Krestenitis, M., Orfanidis, G., Ioannidis, K., Avgerinakis, K., Vrochidis, S., & Kompatsiaris, I. Y. (2018) · Lecture notes in computer science · doi:10.1007/978-3-030-05710-7_35 · Cited by 63 · cited 2×
“[47] M.”
BackgroundDinelli, G., Meoni, G., Rapuano, E., Benelli, G., & Fanucci, L. (2019) · International Journal of Reconfigurable Computing · doi:10.1155/2019/7218758 · Cited by 60 · cited 7×
“Furthermore, quantization and pruning techniques can be efficiently applied to compress the model [35], even improving its accuracy compared to the original one [32].”
BackgroundMethodInfluentialSelf-citationResource Characterisation of Personal-Scale Sensing Models on Edge Accelerators
Antonini, M., Vu, T. H., Min, C., Montanari, A., Mathur, A., & Kawsar, F. (2019) · Proceedings of the First International Workshop on Challenges in Artificial Intelligence and Machine Learning for Internet of Things · doi:10.1145/3363347.3363363 · Cited by 55 · cited 3×
“A benchmark between the different COTS for DL is shown in[28], [29], and [30].”
BackgroundHally, B., Wallace, L. O., Reinke, K. J., Jones, S., & Skidmore, A. K. (2018) · International Journal of Digital Earth · doi:10.1080/17538947.2018.1497099 · Cited by 54 · cited 2×
“[46] B.”
BackgroundHeavy Ions Induced Single Event Upsets Testing of the 28 nm Xilinx Zynq-7000 All Programmable SoC
Tambara, L. A., Kastensmidt, F. L., Medina, N. H., Added, N., Aguiar, V. A. P., Aguirre, F., Macchione, E. L. A., & Guazzelli, M. A. (2015) · IEEE Radiation Effects Data Workshop (REDW) · doi:10.1109/redw.2015.7336716 · Cited by 52 · cited 1×
“At 28 nm, several radiation campaigns have been conducted to assess the performance of the Xilinx Zynq 7000 COTS SoC wherein testing was conducted with heavy ions [50],[51].”
MethodEsposito, M., Conticello, S. S., Pastena, M., & Carnicero Dominguez, B. (2019) · CubeSats and SmallSats for Remote Sensing III · doi:10.1117/12.2532262 · Cited by 42 · cited 3×
“[43] M.”
MethodRoadmap for On-Board Processing and Data Handling Systems in Space
Furano, G., & Menicucci, A. (2017) · Springer eBooks · doi:10.1007/978-3-319-54422-9_10 · Cited by 41 · cited 7×
“[10] G.”
BackgroundMethodInfluentialSelf-citationAdvantages and Limitations of Fully on-Chip CNN FPGA-Based Hardware Accelerator
Dinelli, G., Meoni, G., Rapuano, E., & Fanucci, L. (2020) · IEEE International Symposium on Circuits and Systems proceedings · doi:10.1109/iscas45731.2020.9180867 · Cited by 38 · cited 3×
“Generally, the advantages of FPGAs compared to other COTS strictly depend on many factors, such as the architecture, model size, and the number of bits used for representation [30], [32], [34], [35].”
BackgroundMethodSelf-citationNAND flash storage technology for mission-critical space applications
Fabiano, M., & Furano, G. (2013) · IEEE Aerospace and Electronic Systems Magazine · doi:10.1109/maes.2013.6617096 · Cited by 37 · cited 2×
“[41] M.”
BackgroundSelf-citationA Neural Network Based Fault Management Scheme for Reliable Image Processing
Biasielli, M., Bolchini, C., Cassano, L., Koyuncu, E., & Miele, A. (2020) · IEEE Transactions on Computers · doi:10.1109/tc.2020.2965518 · Cited by 28 · cited 2×
“On the contrary, for those applications requiring reprocessing the image, the detection of “slightly corrupted” images allows saving time due to image processing pipeline [45].”
BackgroundMethodSpace Processor Radiation Mitigation and Validation Techniques for an 1800 MIPS Processor Board
Hillman, R. S., Swift, G. M., Layton, P., Conrad, M., Thibodeau, C., & Irom, F. (2003) · ESASP · Cited by 24
Tambara, L. A., Akhmetov, A. O., Bobrovsky, D. V., & Kastensmidt, F. L. (2015) · European Conference on Radiation and Its Effects on Components and Systems · doi:10.1109/radecs.2015.7365643 · Cited by 22 · cited 2×
“At 28 nm, several radiation campaigns have been conducted to assess the performance of the Xilinx Zynq 7000 COTS SoC wherein testing was conducted with heavy ions [50],[51].”
MethodA low power keyword spotting algorithm for memory constrained embedded systems
Benelli, G., Meoni, G., & Fanucci, L. (2018) · IFIP/IEEE International Conference on Very Large Scale Integration · doi:10.1109/vlsi-soc.2018.8644728 · Cited by 21 · cited 2×
“In addition, the inference of many AI models is computationally intensive [36], since they require a high number of operations per second to respect the latency requirements typical of many onboard applications.”
BackgroundSelf-citationSurvey of Processors for Space
Ginosar, R. (2012) · Cited by 21
Rapid Prototyping of Deep Learning Models on Radiation Hardened CPUs
Blacker, P., Bridges, C., & Hadfield, S. (2019) · NASA/ESA Conference on Adaptive Hardware and Systems (AHS) · doi:10.1109/ahs.2019.000-4 · Cited by 17 · cited 1×
“For instance, Blacker et al. [25] developed a tool to facilitate the porting of DNNs onboard the LEON 3 processor. computing parallelism of the GPUs to improve performance at the cost of higher energy consumption; and 3) FPGA-based hardware that significantly increases the performance per watt but…”
MethodDing, L., Gerardin, S., Bagatin, M., Bisello, D., Mattiazzo, S., & Paccagnella, A. (2015) · Nuclear Instruments and Methods in Physics Research Section A Accelerators Spectrometers Detectors and Associated Equipment · doi:10.1016/j.nima.2015.03.021 · Cited by 17 · cited 1×
“…challenges due to ionizing radiation: space-borne devices shall take into account both single event effects (SEEs) caused by the charge deposited by ionizing particles [8], and effects due to long time exposure to radiation, such as total ionizing dose (TID) and total-not ionizing dose [9], [10].”
BackgroundEvaluating research for disruptive innovation in the space sector
Summerer, L. (2012) · Acta Astronautica · doi:10.1016/j.actaastro.2012.08.009 · Cited by 14 · cited 2×
“Costs ($) reduction has great importance for the advent of the New Space era [1], [13].”
BackgroundRCU2 — The ALICE TPC readout electronics consolidation for Run2
Alme, J., Alt, T., Bratrud, L., Christiansen, P., Costa, F., David, E., Gunji, T., Kiss, T., Langoy, R., Lien, J., et al. (2013) · Journal of Instrumentation · doi:10.1088/1748-0221/8/12/c12032 · Cited by 14 · cited 1×
“…work on radiation effects on FPGA-based systems has been conducted under the auspices of the ALICE project within the umbrella of CERN research [49], where parts of the collider sensing equipment are managed by Xilinx Virtex-II COTS FPGA-controlled subsystems that are exposed to high energy…”
MethodA selective mapper for the mitigation of SETs on rad-hard RTG4 flash-based FPGAs
Sterpone, L., Azimi, S., & Du, B. (2016) · European Conference on Radiation and Its Effects on Components and Systems · doi:10.1109/radecs.2016.8093152 · Cited by 13 · cited 2×
“[14] L.”
BackgroundWertz, P., Hespeler, B., Kiessling, M., & Hagmanns, F. J. (2016) · Cited by 12
Heavy Ion Latch-up Test on dsPIC Microcontroller to be used in ExoMars 2020 Mission
Manzano, P. D. C., Álvarez, M., Sampedro, S., Rivas, M. J., Traseira, I., Manzano, J., Mingo, J. R. D., Martín-Ortega, A., & Santiuste, N. A. (2017) · IEEE Radiation Effects Data Workshop (REDW) · doi:10.1109/nsrec.2017.8115481 · Cited by 11 · cited 3×
“Furthermore, modern silicon in space faces specific challenges due to ionizing radiation: space-borne devices shall take into account both single event effects (SEEs) caused by the charge deposited by ionizing particles [8], and effects due to long time exposure to radiation, such as total ionizing dose (TID) and tota…”
BackgroundTotal-ionizing-dose radiation response of partially-depleted SOI devices
Rezzak, N., Zhang, E. X., Alles, M. L., Schrimpf, R. D., & Hughes, H. L. (2010) · IEEE International SOI Conference (SOI) · doi:10.1109/soi.2010.5641057 · Cited by 10 · cited 1×
“Space-qualified devices should then apply Radiation-Hardened By Design (RHBD) [11] or Radiation-Hardened By Process (RHBP) techniques, for example, by exploiting flash-based and silicon on insulator [12] processes.”
MethodSimplified Procedures for COTS TID Testing: A Comparison Between 90Sr and 60Co
Menicucci, A., Malatesta, F., Capua, F. D., Campajola, L., Casolaro, P., Furano, G., Mascio, S. D., & Ottavi, M. L. (2018) · IEEE Nuclear and Space Radiation Effects Conference (NSREC) · doi:10.1109/nsrec.2018.8584279 · Cited by 8 · cited 3×
“SoCs in general present specific challenges for radiation characterization campaigns compared to other silicon devices such as memories (see [6], [10], and [20] for background).”
BackgroundSelf-citationEstimating the Downlink Data-Rate of a CCSDS File Delivery Protocol IP Core
Meoni, G., Valverde, A., Magistrati, G., & Fanucci, L. (2020) · Lecture notes in electrical engineering · doi:10.1007/978-3-030-37277-4_60 · Cited by 6 · cited 3×
“[19] G.”
BackgroundMethodSelf-citationGuideline for Single-Event Effect (SEE) Testing of System on a Chip (SOC) Devices
Guertin, S. M. (2018) · NASA Technical Reports Server (NASA) · Cited by 3
Research works printed but not linked: 2
- chen, 0, Learning efficient object detection models with knowledge distillation, Proc Adv Neural Inf Process Syst, 742
- esposito, 2019, Hyperscout-2: Highly integration of hyperspectral and thermal sensing for breakthrough in-space applications, ESA Earth Observation ?-Week
Cited by
Mahboob, S., & Liu, L. (2024) · IEEE Communications Surveys & Tutorials · doi:10.1109/comst.2023.3347145 · Cited by 255
“In [315], minimum flow maximum residual path-based network flow algorithm is used to find out the optimal network routing path for satellites.”
BackgroundMethodGiuffrida, G., Fanucci, L., Meoni, G., Batic, M., Buckley, L., Dunne, A., van Dijk, C., Esposito, M., Hefele, J., Vercruyssen, N., et al. (2022) · IEEE Transactions on Geoscience and Remote Sensing · doi:10.1109/tgrs.2021.3125567 · Cited by 225
“radiation cross sections across a range of energies [17].”
BackgroundMethodSelf-citationEmotion recognition in EEG signals using deep learning methods: A review
Jafari, M., Shoeibi, A., Khodatars, M., Bagherzadeh, S., Shalbaf, A., García, D. L., Gorriz, J. M., & Acharya, U. R. (2023) · Computers in Biology and Medicine · doi:10.1016/j.compbiomed.2023.107450 · Cited by 240
Cloudscout: a deep neural network for on-board cloud detection on hyperspectral images
Giuffrida, G., Diana, L., de Gioia, F., Benelli, G., Meoni, G., Donati, M., & Fanucci, L. (2020) · Remote Sensing · doi:10.3390/rs12142205 · Cited by 176
“• On-board filtering of unuseful data, relaxing the strict bandwidth requirements typical of modern/future Earth Observation applications [6–8]; • Preliminary decision taken directly on board, without the need for a human operator; • Mission reconfigurability, changing only the weights of the network[6]; • Continuous…”
BackgroundMethodInfluentialSelf-citationPauly, L., Rharbaoui, W., Shneider, C., Rathinam, A., Gaudillière, V., & Aouada, D. (2023) · Acta Astronautica · doi:10.1016/j.actaastro.2023.08.001 · Cited by 121
“However, most existing algorithms are tested on workstations and large server clusters and very limited evaluations have been conducted on edge systems with FPGA [36, 77] or GPU [72, 122, 15]-based AI accelerators for space applications.”
BackgroundMethod
Metadata
- Authors
- Furano, G., Meoni, G., Dunne, A., Moloney, D., Ferlet-Cavrois, V., Tavoularis, A., Byrne, J., Buckley, L., Psarakis, M., Voss, K.-O., & Fanucci, L.
- Affiliations
- European Space Agency (FR)
- University of Pisa (IT)
- Intel (Ireland) (IE)
- University of Piraeus (GR)
- GSI Helmholtz Centre for Heavy Ion Research (DE)
- ORCID
- Venue
- IEEE Aerospace and Electronic Systems Magazine
- Published
- 2020-12-01
- Volume
- 35
- Issue
- 12
- Pages
- 44-56
- ISSN
- 0885-8985 (ISSN-L)
- 1557-959X
- OpenAlex
- W3113229535
- Cited by
- 259
- References
- 48
- Citations by year
- 2026: 26 · 2025: 58 · 2024: 66 · 2023: 57 · 2022: 32 · 2021: 18 · 2020: 2
- FWCI
- 9.36
- Type
- journal article
- Language
- en
- Open access
- closed
- Licence
- Locations
- IEEE Aerospace and Electronic Systems Magazine (publishedVersion)
- CINECA IRIS Institutial research information system (University of Pisa) (submittedVersion)
- Topics
- Physical Sciences › Engineering › Electrical and Electronic Engineering › Space Technology and Applications
- Social Sciences › Political Science and International Relations › Space exploration and regulation
- Physical Sciences › Engineering › Aerospace Engineering › Spacecraft Design and Technology
- Keywords
- system-on-chip, commercial off-the-shelf hardware, on-board processing, Earth observation, remote sensing, radioresistance
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.