2022 IEEE Aerospace Conference (AERO) · 2022 · 11 citations · 20 references
EngineeringComputer ArchitectureSingle ChipVirtual InstrumentationControl SystemsHardware ArchitectureHardware PlatformCubesatsSystems EngineeringInstrumentationParallel ComputingConventional InstrumentationComputer EngineeringComputer ScienceReconfigurable ArchitectureEarth Observation MissionsFpga DesignReconfigurabilityComputer VisionSystem On ChipOpen Source Components
As the number of earth observation missions in-creases, together with the number of images acquired by satel-lites, the need for optimizing onboard memory allocation and bandwidth to the ground is becoming more and more critical. Currently, large players in the space industry face this need by developing custom Instrument Control Units (ICUs), which can only be employed for the single mission they have been designed for. The unavailability of flexible, programmable, space-grade solutions ready to be integrated as building blocks keeps the costs of such units unaffordable for smaller missions and smaller industries, which do not have the resources needed to undertake custom development at a single system component level. Moreover, the development of this kind of component can increase the time to market of the system. The work presented in this paper aims at delivering an innovative fully programmable data handling and data processing System on a Chip (SoC) for modern ICUs onboard satellites, named ICU4SAT. We leveraged three highly innovative, open-source components: a RISC-V based processor as a command and control platform; the FPGA-GPU as hardware accelerator for computer vision and Artificial Intelligence algorithms; and the SpaceFibre/SpaceWire as high-speed communication modules to connect our system to the rest of the satellite. The three components are integrated into a single modular embedded architecture on a single chip, and more in particular on a space-grade FPGA. The result is a general-purpose ICU SoC, which can be potentially used to control and process data of any satellite image-based instrument. The ICU4SAT project aims to provide and fully programmable data handling and data processing modern ICU on board of satellites, also exploiting re-configurable FPGA technology. This will advance the state of the art from both the academic and industry point of view as it will: (i) Enable new mission concept by on-the-fly reconfigurability repurposing earth observation missions with no needs to launch new satellites as monitoring need change, even temporarily. (ii) Open new market opportunities for Small-Medium Enterprises since huge investments are not required to access this control unit solution.
20
TensorFlow: A system for large-scale machine learning
Martı́n Abadi, Paul Barham, Jianmin Chen et al. · arXiv (Cornell University) · 2016 · 8.8K citations · Full text
Edge Computing: Vision and Challenges
Weisong Shi, Jie Cao, Quan Zhang et al. · IEEE Internet of Things Journal · 2016 · 7.5K citations
Deep Learning for Generic Object Detection: A Survey
Li Liu, Wanli Ouyang, Xiaogang Wang et al. · International Journal of Computer Vision · 2019 · 2.7K citations · Full text
Convolutional Neural Network, Image Analysis, Machine Learning +13
John Nickolls, William J. Dally · IEEE Micro · 2010 · 945 citations
Cluster Computing, Computational Science, Gpu Architecture +14