Cobham Gaisler announced today that a consortium it is leading has been awarded a development contract from the European Space Agency (ESA) that will result in the design of an accelerator board suitable for Computer Vision (CV) and Artificial Intelligence (AI) applications for in-orbit space use. The consortium comprises Cobham Gaisler (Sweden), Ubotica Technologies (Ireland), QinetiQ Space nv (The Netherlands), and the National and Kapodistrian University of Athens (Greece) and they will work together on the FPGA Accelerated DSP Payload Data Processing Board activity for ESA.
The proposed high-performance compute architecture combines state-of-the-art Vision Processing Units (Myriad 2 from Intel, formerly Movidius located in Ireland), a high-capacity Field Programmable Gate Array (FPGA) and a radiation-hardened microcontroller (GR716 from Cobham Gaisler) to create a reliable system solution for space applications comparable to the performance and functionality currently only available in commercial applications.
"The focus of the ESA project is to analyze space applications with high demands on computing power, with focus on applications found within the domain of Earth Observation (EO), and to define a high-performance compute board architecture to address current and future needs of these applications", said Jan Andersson, Director of Engineering, Cobham Gaisler.
"We are very excited about entering a new domain in space computing. We have traditionally developed highly radiation-tolerant and fault-tolerant system-on-chip devices using silicon libraries specifically developed for use in harsh environments like space. We are now using our most complex and at the same time smallest device, the radiation-tolerant GR716 microcontroller to control a custom-of-the-shelf commercial grade vision processor to offer an overall reliable computing solution with a computing power that surpasses what has been used until now in the space domain. We are already looking for ways to bring this high-performance compute board to the space market”, said Sandi Habinc, General Manager, Cobham Gaisler.
“This activity is an exciting convergence of commercial edge processing compute capability and radiation tolerant management and supervisory control that facilitates the realisation of novel on-board Earth Observation applications. Myriad’s low-power software-hardware Image Signal Processing and Computer Vision pipelines enable EO data processing to be efficiently performed at streaming data rates, while adding real time inference with customisable AI models”, said Aubrey Dunne, Vice President of Engineering, Ubotica Technologies.
"New developments in the space sector have created a significant increase in the degree of accessibility, uses, and applications derived from Earth Observation missions. Commercial processing units used in ground applications continue to evolve rapidly and enable the use of complex and advanced algorithms at higher data rates. Meanwhile, capable software is playing a key role in the way we deal with the sudden increase of data volume derived from bigger constellations and higher resolution imagery. This joint development represents a step forward into the next generation of Earth Observation satellite missions and it will for sure allow new capabilities on-board future spacecraft", said Juan Pablo Ramos, Business Development Manager, QinetiQ Space nv.
"The ESA project widens our horizons regarding computing accelerators for space applications. Our experience in designing that involves high throughput accelerators and interconnections will expand to include also reliability, fault-tolerance and executing computationally complex applications in rigid conditions", said Dr Dionysis Reisis, Associate Professor, National and Kapodistrian University of Athens.
"This ESA activity aims at developing a processing board for Earth Observation applications with unprecedented capabilities for on-board image processing, through the use of a large FPGA and commercial image processors (Intel Myriad 2) – by building on synergies with other on-going ESA sponsored activities. Having access to high-performance commercial processors on-board spacecraft will enable the use of image processing and machine learning tools previously not available for space applications. Enabling this technology for Earth Observation satellites will not only enable a new level of on-board software processing performance – but may also prove to be a future enabler for new methods of data selection and reduction prior to downloading, which could in turn enable new mission concepts for small satellites", said David Steenari, On-board Payload Data Processing Engineer, European Space Agency.
Date: Oct 8, 2019