The Future of Edge Computing in Space

By Marcel Lariviere - August 11, 2021

Spacecraft in Orbit

The satellites we’re sending to orbit today have a major setback. They do not have the resources for on board data processing and are forced to send raw data to terrestrial data centers to be processed. 


Why is this an issue? Well, the rising number of satellites being launched into space results in the downlink pipeline to the ground becoming increasingly constricted. Additionally, space missions that operate farther from Earth, such as the moon or Mars, have to consider the delay in the time it takes data to be transmitted. The average round-trip communication time is 2.6 seconds between the moon and Earth, and 12.5 minutes between Mars and Earth, both of which are unacceptable time-delays for systems that are supposed to be operating in real-time. If there is an unforeseen problem with a spacecraft heading to Mars, operators on Earth won’t know for 12.5 minutes, and then even if they react straight away it’ll be another 12.5 minutes before any instructions get back to the spacecraft – a lot that can go wrong during that time. 


To address these setbacks, the processing location of data generated from satellites is starting to shift from traditional terrestrial data centers to edge computing, or localized computing (i.e. data processing that takes place on the satellite itself). Edge computing allows for data to be automatically and instantly analysed so that only relevant information is sent to the ground. Processing satellite data on-orbit through edge computing not only decreases network loads on the ground, but also allows for low-latency actionable insights for time-sensitive missions. 


For example, the possibilities and applications of processing photos in space can have immensely positive impacts for life on Earth. Some use cases include wildfire detection, illegal fishing vessel monitoring, humanitarian aid, disaster relief, agriculture monitoring, and ocean trash monitoring to name a few. Current state-of-the-art satellites take 5+ hours from the time that a picture is captured to when it is accessible to the end-user. That 5 hour time delay is what the industry refers to as data latency. Data latency can mean the difference between a catastrophic fire and one that is quickly contained or knowing the location of fishing vessels that could be operating illegally. Processing photos directly on the satellite could decrease the data latency to only 30 minutes. 


Looking forward, edge computing in space will become essential for space operations including predictive maintenance, autonomous satellite docking, on-orbit assembly, orbital refueling, swarm satellite constellation management, and deep space missions.


Predictive Maintenance:
A relatively new concept in manufacturing where machine learning tools use historical and real-time data from various parts of a machine to anticipate problems before they occur. This concept could be applied to a satellite that uses edge computing to continuously monitor the performance and condition of the satellite, while offloading complex data analytics and empowering operators to be proactive in replacing satellites. Using edge computing, operators would be notified about potential part failures before they occur, thereby minimizing downtime and improving the efficiency of the satellite. 


On-Orbit Operations:

For other space operations, such as satellite docking, on-orbit assembly, orbital refueling, and swarm satellites, real-time data analysis through edge computing is essential. These complex operations involve a large number of sensors, cameras, and electrical and mechanical hardware that produce and communicate data between the various components. One small error in the data processing or transmittals could lead to catastrophic failures of the entire mission. Sending all data to the ground for processing is tedious, inefficient, and increases the risk of failure, making edge computing an ideal solution to ensure the success of all operations. 

Deep Space Missions:

Deep space missions outside of Earth’s orbit, such as NASA’s Artemis program that is preparing to return humans to the moon by 2024, have specifically expressed interest in developing AI at the extreme edge [NASA, SBIR]. SpaceX has also announced plans  to make regular visits to Mars starting as soon as 2024 [SpaceX, Mars & Beyond]. Within the next 5 to 10 years, Mars and the moon will increasingly have their own data processing needs. Edge computing could provide low latency processing over the Mars and lunar surface so astronauts can make key decisions quickly. It could also allow a scalable data processing network that can be used in the far reaches of the solar system. 

The cost of sending satellites to space is decreasing dramatically, from launch to manufacturing, the capital needed for a space mission is no longer a major hurdle. With this, the need for computation “at the edge” (in space) also grows more pronounced. This is why, here at Exo-Space, our mission is to enhance the infrastructure of space through orbital edge computing solutions. Our primary product is a compact data processing unit that connects to imaging satellites and runs the raw image data through machine vision algorithms before the data is sent to the ground. Edge computing has the potential to change the space industry for the better and create services that would have never been possible with the current computing infrastructure.