12 mins read 29 Jul 2021

Trellis Data Developing Australian AI Capabilities in Space

Artificial intelligence is finding an increasing number of applications in the space industry, including on NASA’s latest robotic Mars explorer. In Australia, it is Trellis Data that is leading our efforts in developing sovereign AI technologies.

The Mars2020 mission included the rover Perseverance which was lowered by sky crane to the surface of the red planet in February of this year. Artificial intelligence was critical to the success of the landing. Credit: NASA/JPL-Caltech

After a seven-month journey across the vastness of space, NASA’s Perseverance ('Percy') rover made a textbook landing in an inhospitable crater on Mars in February of this year. The landing required years of meticulous research and planning back on Earth, but even still, not every contingency can be catered for from 250-million kilometres away. So, NASA’s engineers gave the spacecraft the ability to take care of itself – using its own artificial intelligence.

Although NASA had landed car-sized rovers on Mars before, the level of difficulty for Percy was ramped up significantly. Curiosity’s landing in 2012 was aimed at the equivalent of a 440 sq-km vacant parking lot, but Percy targeted an area of just 160 sq-km littered with boulders, gullies and cliffs.

Viking, which landed on Mars in 1976, targeted a landing region encompassing an area of 90,000 sq-km.

With the enormous distance between Earth and Mars, there is just no way to safely control a spacecraft landing on the red planet in real-time. Percy had to scan the Martian terrain and pick out a safe place to touchdown on its own.

The whole process was reminiscent of the first Moon landing. During the landing of the Apollo 11 lunar module, Neil Armstrong was looking out of the window and dodging boulders and craters while he searched for a safe landing site. The spacecraft eventually touched down with only about 15 seconds of fuel remaining.

Like Armstrong, Perseverance had memorised maps of its landing site’s surface and navigated by following known landmarks. It also had the ability to react appropriately when faced with unexpected obstacles. And so, instead of being stuck in deep sand or dangling over a cliff, Percy landed bang on target in the Jezero Crater.

The use of artificial intelligence systems didn’t end with the rover’s landing. It’s also built into Perseverance’s precision X-ray beam, PIXL (Planetary Instrument for X-ray Lithochemistry), that will be searching for signs of past microbial life. The laser can identify features in rock and soil as small as a grain of salt, and AI will be used to guide it to get the most accurate aim.

But the US is not the only country that is using artificial intelligence in its space hardware. Trellis Data is an Australian owned and operated company that specialises in providing a platform for AI applications for the national government and enterprise, including in the space industry. We managed to organise a chat with co-founder and CEO Michael Gately who gave us some wonderful insight into how AI can be applied to solve real-world problems now, and in the future.

Michael Gately is the co-founder and CEO of Trellis Data, an Australian owned and operated company specialising in applications of Artificial Intelligence. Credit: Trellis Data

Michael, I think that when most people think of artificial intelligence, they visualise robots that look and act like humans. You and Trellis Data are specialists in artificial intelligence and machine learning. Can you tell us what AI really is and whether that is a realistic view?

Robots that act and look like humans are still very much in the realm of science fiction. While there are robots that can be moved as a result of machine learning, they are only performing the most basic functions. AI in today's world is really much more about Intelligent Insights and Intelligent Automation. An Intelligent Insight example includes being able to transcribe voice into text and then analyse the text to determine if a customer is getting agitated during a phone call. An Intelligent Automation example includes taking photos from space to automate the identification and tracking of space junk. These are currently available technologies that leverage advanced machine learning.

Interesting that you mention space junk which has had a lot of focus lately. We might come back to that a bit later. But first I wanted to ask you about machine learning. We often hear of machine learning mentioned together with artificial intelligence. Are they the same thing, or how do they fit together?

Think of AI being a large circle that includes many technologies, such as rules engines. It also includes Machine Learning (ML) technologies. ML is effectively a subset of AI and provides improved insights and automation over other AI technologies, particularly where the problem to solve is complex. Where rules engines leverage rules to make decisions, ML leverages data to 'learn' what decisions to make. The results are game-changing for every industry across the globe.

Great explanation, and I know there is a very accessible section on your website that talks a bit about that. Even though AI seems to have been around forever, many people probably don’t recognise it in their everyday lives. Is that because they are simply not aware of it? Where are people interacting with AI today?

AI is everywhere already. ML is in most places too. AI includes bank application rules engines and even the rules that govern central heating systems in people's homes. ML is in every phone, every search engine and many other areas of our lives.

And in space as well! We all celebrated Perseverance’s landing on Mars earlier this year, and with everything running so smoothly it is tempting to think that landing a robot on another planet has become routine. Can you tell us a bit about the role that AI played in ensuring Perseverance had a safe landing?

The biggest issue facing Perseverance, apart from the massive engineering feat achieved by NASA and its partners, was the delay in getting commands to the asset while it was landing. There was a 7-minute window where Earth could not send new commands because the commands would take too long to get to Perseverance. AI was required to enable the landing to be successful. Additionally, it is fitted with ML capability as well in order to help the rover drive better once landed.

What other roles in the space domain is AI being used for now, or could it be used for in the near future? For example, you’ve mentioned space junk already. Is AI able to help us deal with space junk, or to perform other tasks like improving our defences against hazardous asteroids and comets?

Absolutely AI can help with space junk. Not only can it identify space junk, but it can also read the writing on space junk, allowing it to identify the owner and notify them of the current status of the space junk. But AI can be used for so much more in space as well. It can help understand weaknesses in communication signals between satellites and optimise where best to send signals for the best comms globally. It can be used to help identify climate change and other activities on Earth. Not only this, it can look and listen outwards and tell us when it hears or sees something. Then taking that first insight, it can predict what might happen next (e.g. what's the most likely path of a comet, or if there is a regular signal coming from anywhere in space). It can do all of this in an automated fashion, allowing us to expand how much we look out to space and how much more we can help our own planet.

That certainly sounds incredible. Can you speculate for a minute, and tell us what you think the future holds for AI over the next 50 or 100 years? Is there any limit to what AI is capable of doing?

AI will (and should) always be limited to supporting human life. This is a fundamental principle that we in Trellis Data adhere to and we hope all others in our field do as well. Given that, there is a limit to what AI will be able to do. The future is exciting for AI. We see it converging to be able to increasingly make more complex decisions, moving from transactional and detection level decisions, to strategy and 'what next' level decisions. We will see the building blocks of AI used with AI to perform functions that humans couldn't safely perform at a speed that humans couldn't achieve. 

For example, the future could include AI robots leveraging 3D printing to build components in space while on the moon, test and deploy ready for human use. This will allow us to set up space bases for humans to travel to while robots (using AI) build and grow gardens in space to support ongoing human life). The other area where AI might be at, perhaps in the next 100 years, is the ability to find 'new knowledge'. This is something very specific to brains, particularly human brains. 100 years from now, I see AI being able to shed new light on black holes, traveling at the speed of light, and understanding the depths of our ocean floors.

Can you now tell us about Trellis Data? For those who haven’t heard of Trellis Data before, it is an Australian company that provides a Machine Learning platform for AI systems. Can you tell us a bit about how Trellis Data came about, and what you do today, particularly in relation to space applications?

Trellis Data came about to deliver real-world benefits using ML. We are passionate believers in Australia and that we as a nation are capable of competing in ML on the world stage. We have an incredible team of people that are dedicated to this cause and love all things ML. Our multi-award winning platform, the Trellis Intelligence Platform (TIP), provides our customers with the ability to solve their own problems using their own data. They don't even have to be ML experts to use the platform to achieve these results. 

It's been an exciting journey for us as we have helped deliver improved results in biosecurity, defence, government and logistics industries. As part of this, we are leveraging TIP to peer down from space to Earth to identify the flow of ocean traffic, and various patterns within that. We are also looking to leverage TIP's ability to read things in the real world (think OCR on steroids) to read what is written on space junk, and in doing so, automate the identification of these assets and alert those on Earth who need to know. But we are just getting started in Space. There are many many uses for TIP in Space... pretty much as far as the mind can imagine.

This all sounds incredibly exciting, and I am sure there are many people out there who would be interested in pursuing a career in AI or even with Trellis Data. How should they prepare themselves to enter the industry?

The best way to do this is to reach out to us directly. We are always happy to talk about ML and how AI is changing the world around us. We have meetups for people interested in the research and even if the mathematics is difficult for some, there are other positions in AI, including with us at Trellis Data, that could suit someone who is keen to be part of this revolution. There are also short courses and a plethora of literature on ML and AI publicly available to read to get a sense of it. At Trellis Data, we have an 'introduction to ML' pack with links to shortcut some of this process as well for anyone who is keen to dip their toes into this exciting field of technology.

Michael Gately, thank you very much for your time. You’ve given us an amazing insight into artificial intelligence and its applications in space. We look forward to seeing how things progress and wish you and Trellis Data the best for the future.

The PIXL instrument on NASA's Perseverance rover uses artificial intelligence to ensure the X-ray beam is guided precisely onto its target. Credit: NASA/JPL-Caltech

As our operations in space become more frequent and more complex, artificial intelligence is going to become an increasingly important part of the spacecraft, satellites and robots that operate beyond Earth’s atmosphere. In the relatively distant future, we are going to need advanced autonomous AI systems if we want to explore Solar systems other than our own.

Already artificial intelligence systems are being developed in Australia that seek to predict unknown and unforeseen risks in space to keep us safe, identify new modes of more efficient transport and better navigation systems, and watch over the Earth to help solve problems such as rising sea levels and climate change.

With tangible benefits like these for all of us, it seems as though we should be embracing the use of artificial intelligence in space. If development is limited to supporting humanity, AI systems are nothing to be feared. Indeed, for us to advance to the next stage of space exploration, we will need to trust AI systems.

This is why a locally owned and operated company like Trellis Data is so important. Part of our confidence in the use of artificial intelligence systems will come through developing and building them right here in Australia. Trellis Data’s own platform is already used and trusted across multiple sectors in government and industry.

Artificial intelligence might seem futuristic, but Perseverance’s flawless landing on a planet hundreds of millions of kilometres away showed us that the future is already here.