Averting global disaster using edge computing

Planet Earth from Space.
Co-founder Jeroen Cappaert is the CTO at Spire, focusing on ensuring future technology developments deliver what the business needs. With a background in electronics engineering and aerospace engineering, Jeroen’s previous work and research includes high-enthalpy flow simulation, computational mechanics and fluid dynamics, spacecraft avionics and payload design, geo-engineering, and low-thrust astrodynamics at NASA Ames, Von Karman Institute for Fluid Dynamics, International Space University. Jeroen’s education includes an MSc in Space Studies from the International Space University, a Master thesis on Uncertainty quantification for hypersonic flow simulation from Von Karman Institute for Fluid Dynamics, an MSc in Mechanical Engineering (Cum Laude) and BSc in Mechanical Engineering (with a minor in Electrical Engineering) from Katholieke Universiteit Leuven.

Governments and corporations use Earth Observation (EO) data gathered from space to detect abnormal activity – from illegal shipping and distress signals on vessels to understanding the risk of weather events like hurricanes, tsunamis and climate change. 

Satellites have been providing EO data for a long time but until recently, it was mainly governments and institutions such as NASA and the European Space Agency (ESA) who developed and supported dedicated EO  missions. Now we’re seeing the growing commercialisation of space-based services, known as New Space and the EO data becomes one of the key applications this growing space segment supports.

Advances in electronics miniaturization have paved the way for new generations of small satellites, with remote sensing payloads, which can provide unparalleled coverage of oceans and other hard-to-reach places. This gives us a new powerful means to develop EO data from space, and give it a new dimension. 

The granularity of the data, and ability to derive actionable insights from it, allows experts to understand and predict specific adverse events with greater precision. To give an  example, a combination of weather prediction and soil moisture data can be used to remotely track and mitigate the migration of locusts which had been destroying crops in Africa, the Arabian Peninsula, and South Asia.

New Space, and growth of small satellite technology, makes it more cost-effective for organisations to collect EO data, from its surface to near-space. 

Early warning systems

This data dramatically improves early warning systems because teams can monitor adverse weather patterns or see if a ship has gone off course in near real-time. Paired with predictive analytics and advanced forecasting tools, this data supports more accurate forecasts of future events and a better estimation of the long-term impacts of climate change. 

Yet while developments in space-based data technology present a world of opportunities for organisations, their own processing capabilities are often a barrier. Many lack the bandwidth and IT infrastructure needed to interpret and act on data which is flooding in 24/7, and are reliant on human analysts who cannot work at the speed and scale required. 

Helping organisations to make sense of their data has been one of the driving forces behind our Brain in Space project, launched at the end of last year in partnership with ESA’s Earth Observation Science for Society Programme and Φ-lab. Having already developed our own novel computing platforms suitable for AI processing, we’re now expanding our machine learning, neural network, and AI processing capabilities further.

As part of the project, launched at the end of last year, our team created an on-the-ground simulated testbed which replicates our LEMUR 3U platform, flagship of our global constellation of over 110 nanosatellites. The testbed includes embedded edge AI and ML modules, allowing users to schedule, upload and test AI/ML-powered applications that would enable them to process more satellite data than ever.

The testbed was set up at our facilities in Luxembourg but is available to our partners and customers from anywhere in the world. Within this simulated environment, we’re testing how well the new embedded edge AI/ML modules will support the development of advanced AI-enabled analytics and edge computing in space. These modules combine signal processing with a dedicated set of machine learning capabilities, so sensors can rapidly detect and isolate patterns in large volumes of collected data, without human intervention.

Advances in AI-assisted, on-board processing of space sensor data mean we’re able to create ‘digital twins’ of our planet, while reducing the burden on ground stations and other infrastructure. Satellites, with embedded edge AI/ML modules, would autonomously prioritise what data is downloaded first to reduce delays. They’ll also direct sensors and provide actionable insights far faster than ever before, even as the data becomes more complex.

We’re reaching the point where satellites are capable of performing time-critical missions and making decisions autonomously, without impacting bandwidth or on-the-ground resources. Depending on operational requirements, they may be programmed to collect data for specific regions and, where anomalies are detected, a data stream can be shared with on-the-ground teams, so they can quickly identify what’s causing it.

AI/ML marks a major leap forward in how small satellite constellations are operated and managed. We now have greater computing power in orbit, communication links that enable satellites to operate as a network, and edge computing to support real-time decision-making. 

Orbital process

Rather than collecting disparate datasets, which require more bandwidth to download and resources to model and analyse, the entire process is performed seamlessly in orbit. Bandwidth constraints are reduced because relevant and precise information is automatically sent to the right location.  

What happened in the mid-90s with networked PCs is only now happening in the space industry. And, as we continue to develop better AI/ML algorithms, satellite technology will also improve rapidly to support their application in space. 

As well as enhancing early warning systems, this would open up other new possibilities – from automated, real-time row-level irrigation in agriculture to being able to check the provenance and sustainability of fish at a local restaurant. We’re already seeing how space data is being used to monitor very specific events, like swarms of locusts, but there are many more applications. 

We hope the Brain in Space testbed will drive forward AI-assisted space sensor data processing. Using a proven sensor platform, we’ll see whether AI/ML algorithms could support on-board processing on a large-scale and reduce current limitations. We’re also trialling a number of AI frameworks and algorithms for space applications in order to develop new products for different markets. Space data has, until recently, been a relatively untapped resource but innovation in the field means it’s now within the grasp of more organisations. 

Want to learn more about topics like this from thought leaders in the space? Find out more about the Edge Computing Expo North America on September 29-30, a brand new, innovative event and conference exploring the edge computing ecosystem.

Tags: , , , ,

View Comments
Leave a comment

Leave a Reply