AI in Space Tech

With the advent of commercial space exploration companies like SpaceX and Blue Origin and increasing expenditure of governments worldwide into space technology, it is only a matter of time before mankind becomes an extra-terrestrial species. In fact, many believe this to be the only logical next step for man, considering how our home planet has finite resources and an infinitely growing population. While this will require immense effort in the fields of rocket engineering, another slightly unexpected aspect of this space race interests me. The increasing use and future potential of Artificially Intelligent systems in space.

Machine learning has two broad applications in space:

  1. Analysis and Prediction:

            While “Big Data” has only recently become a buzzword in the tech circles due to the rapid and widespread use of the internet, this is not the case for space exploration. Be it a large radio telescope array gathering information about a distant galaxy, or a satellite orbiting Jupiter and transmitting these readings 24/7 to an operator on earth, Space exploration has always involved massive amounts of data. As an example, NASA collects approximately 2GB of data every 15 seconds from all its satellites. However, only a fraction of this data is analysed due to limitations in time, manpower and resources. Thanks to AI, this can be done much quicker, and to a decent level of accuracy. NASA annually conducts Frontier Development Lab (FDL) sessions for 8 weeks every summer where technology and space innovators are brought together to brainstorm and come up with code to solve a variety of problems. FDL also partners with big tech companies who provide the scientists with an advanced tech repository of sorts in the form of hardware, algorithms, super-computer resources, funding, facilities and subject-matter experts. Recently a team of scientists came up with a model to analyse the massive data generated by the Kepler mission to search for habitable exoplanets by analysing the emission spectrum of these systems. Since there have been thousands of exoplanets discovered so far, making quick decisions about which of them are the most likely to harbour life can help us narrow down candidates for further detailed (and costly!) exploration. Another interesting outcome of FDL was an algorithm that used radar data from nearby asteroids to model the important parameters like their shapes, sizes, and spin rates which are critical in NASA’s efforts to detect and deflect threatening asteroids from Earth. While traditional methods could take up to 3 months to model a single asteroid, this ML model can complete the task in just 4 days! Another great example of the use of ML was the development of a “virtual sensor” which could fill in missing data gathered from sensors aboard the Solar Dynamics Observatory (SDO) by analysing historical trends.

In the future, the predictive power of AI could be extensively used for tasks ranging from charting out interstellar travel courses to helping us analyse distant star systems with high accuracy

2. Real-Time Decision Making:

So far, we have seen how ML can be used to predict or analyse large volumes of data. This field of AI is often termed as “supervised learning”. However, there’s another area of ML termed as “reinforcement learning” that has great potential as well. Essentially, reinforcement learning aims to mimic biological learning by making a robot repeatedly try a certain task. If the robot succeeds, we reward it with a “cookie” else if it fails we “punish” it. Thus with time, the system learns how to execute a particular task.

 Presently there’s a whole host of dangerous operations like spacecraft repairs that astronauts may have to perform in worst-case scenarios. Reinforcement learning algorithms show great promise in these areas. By simulating scenarios artificially in labs, we can “teach” robots to carry out various critical tasks that are otherwise too risky for humans. Apart from this, a combination of RL and supervised ML could also be used to develop systems that help humans carry out mundane or time-consuming tasks which could arise especially in long term space travel. In 2018, the German Aerospace Center (DLR) launched an AI assistant to support its astronauts in their daily tasks onboard the International Space Station. CIMON (Crew Interactive MObile companioN) is fully voice-controlled and can see, speak, hear, understand and even fly! CIMON returned after 14 months, but CIMON-2 arrived in December 2019 to replace it. The Japanese Space Agency (JAXA) has also taken pioneering steps in this field by developing an intelligent system that is currently aboard the International Space Station taking pictures of experiments in the Japanese module, KIBO. JAXA’s Int-Ball operates autonomously and can take pictures and videos. It was developed to promote the autonomy of extra- and intra-vehicular experiments while seeking to acquire the robotics technology necessary for future exploration missions. Scientists believe such robotic systems could also help in interstellar travel to navigate potentially deadly spaces like asteroid belts.

That being said, we still have a long way to go before AI systems like TARS from interstellar are developed. The key issues that are faced by the industry today include:

  1. Expensive nature of space engineering: Before any RL/ML system is deployed there comes the monetary expenditure that could potentially be wasted if the mechanical systems onboard fail to operate as expected.

  2. Unreliable results: Since ultimately any AI is only as good as the data we feed it, there’s a decent chance that the predictions made are inconsistent with our expectations, especially if the input data is noisy. Thus more research needs to be done to develop robust systems capable of generating reliable output to make their use more widespread.

  3. Separation of academia: Traditionally, space has been the domain of engineers and physicists while the fields of AI have been home to computer geeks. Since the two fields are so widely separated, it is very unlikely that intra-disciplinary discoveries are made. With initiatives like FDL, this gap is being bridged steadily.

With cutting edge technology being developed this very instant by some of the largest tech companies on the planet like Google, SpaceX and NASA, I strongly believe that the time when mankind will have settlements on other planets will occur within our lifetimes and no doubt AI would have played a pivotal role in this expansion.

Till then, May the force be with us all!

— Shreyas Bhat

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: