Moog GPU Products for Spacecraft
Introduction to John Schaf
I have been with Moog for 25 years. My efforts here have been focused in the design of electronics for high reliability applications. For the last 14 years, I have been focused on avionics and electronics for launch vehicles and spacecraft. My education is in electrical engineering where I have earned both Bachelor and Master degrees from the University of Buffalo. The focus of my graduate work was in high speed digital logic design for microprocessors and pipelined RISC processors.
Why did Moog develop this new GPU Processing Platform?
Moog teamed with Unibap of Sweden to develop a highly fault tolerant/ radiation tolerant graphics processing unit (GPU) processing platform for spacecraft. There is a growing need in the spacecraft industry for higher computational capability that uses less power. This technology satisfies that need.
What brought this need and what problems are we solving?
The processing of data acquired from various sensors on spacecraft, such as imagers and synthetic aperture array radar, has typically been done on the ground. Data is transmitted (downlinked) to earth stations where it is processed and analyzed. This process works when the data set is small and there is ample time to analyze the data. However, todays spacecraft sensors can generate gigabytes of data per second. It is impractical to downlink this amount of data and find interesting features in a short time frame. What is needed, as we call it in the industry, is processing at the edge. Being able to process these large data sets in near real time on the spacecraft. To do so requires either lots of traditional computing resources with its corresponding high demand for power and weight or something innovative that works efficiently and provides very high computational throughput. The graphics processing unit (GPU), the same as you might have in your Play Station IV, is capable of high throughput while requiring less power than a traditional CPU. Moog is applying this technology to spacecraft avionics to enable computing at the edge.
What are we improving from a performance aspect?
Moog is improving the data processing capability of spacecraft and at the same time lowering the power required. This trend is consistent with the needs of our customers who are always looking for increased performance while at the same time reducing the size/weight and power.
Why does this benefit our customers? What are some important facts for a customer to know?
This technology benefits both the spacecraft provider and the payload provider, who sometimes can be different organizations, such as on hosted payloads. This increase in performance allows new applications on the spacecraft such as machine learning, artificial intelligence and heterogeneous computing. These are domains that up until a few years ago could not have been imagined for deployment on spacecraft and satellites. This technology really opens up advance capability for data analysis on orbit.
For Moog, what is the future for Satellite & Spacecraft avionics?
Moog has historically been known for disrupting the satellite and spacecraft avionics market by providing higher integration, smaller/lighter systems and increased capability. By adding GPU based computing to our deep portfolio of spacecraft avionics solutions, we are looking to enable machine learning and artificial intelligence is space. True high performance computing on the edge will be realized.