142512531 / Artificial Intelligence Space © Siarhei Yurchanka | Dreamstime.com
Space Ai Promo 61c20e8aeeb73

A Promising Future for AI and Autonomy in Space

Jan. 6, 2022
Machine learning and deep learning are the next frontier in AI, and thus, in space applications. To quicken the integration process, engineers need software tools that they’re familiar with.

What you'll learn:

  • Making spacecraft more autonomous.
  • How machine learning and deep learning are driving autonomy-facilitating algorithms.
  • Leveraging software-design tools like MATLAB and Simulink for algorithm training and other aspects.

Artificial intelligence (AI) continues to dominate the news, but the reality hasn’t always lived up to the hype. Such is the case in the automotive industry, where the goal of fully autonomous vehicles for the general public remains elusive.

Surprisingly, the space industry, long a stalwart of conservative engineering, stands poised in the shadows to deploy machine learning in practical and pragmatic ways and once again become a leader in computing applications. Indeed, AI could be a competitive differentiator for the next generations of spacecraft.

AI and Its Potential in the Space Industry

The applications of AI in the space industry lend themselves well to incremental development. First, a spacecraft that can perform the analogous task of “driving down the empty road” has already solved the self-driving problem, to some extent. No vehicles, pedestrians, or bicycles in space can jump in front of a spacecraft—there’s mostly emptiness, operating within well-understood laws of physics and devoid of the ethical issues.

Second, whereas the primary purpose of a car is transportation from point A to B (which is what the AI is focused on enabling), a spacecraft often has many other purposes. In fact, the equivalent of the “transportation” function (called guidance, navigation, and control in the language of the space industry), while critical, is often merely the enabler of other functions, such as providing satellite imaging or performing science observations. This means that a multitude of functions can leverage AI, some more risky than others. Since AI is still largely unproven in space, the less risky functions are good candidates for its first applications.

The space industry also has a long history of gradually making spacecraft increasingly autonomous. Space vehicles that can make more decisions on their own are much more valuable, especially for space exploration.

For example, NASA’s Mars Curiosity Rover is armed with an instrument called ChemCam, which analyzes the composition of Martian rocks and soils. But to do this, ChemCam first must point itself at a target. Giving the pointing instructions from the ground is a cumbersome process, limited by whether the right communications satellites are within view of Curiosity and even by the length of time it takes commands and data to travel from Mars to Earth (known as the light-time constraint). For this reason, Curiosity uses an autonomous targeting algorithm to point its instrument during times that ground commanding isn’t available.

It represents a good example of a “low risk, high reward” application of AI in space. There’s also very little downside to this approach—the algorithm adds scientific value if it succeeds, and if it fails, little to nothing is lost. However, because adding autonomy necessarily also increases design complexity, such algorithms are difficult and expensive to create using traditional programming techniques. But expanding the use of such algorithms will result in a more capable, and competitive, industry landscape.

Machine Learning and Deep Learning Push Autonomy Forward

Recent advances in AI, driven by machine learning and deep learning, have made autonomy-facilitating algorithms not only more powerful, but also faster to create and accessible to a larger pool of companies and engineers. These advances also have happened at a time when the space industry is experiencing unprecedented growth in the number of new, private companies entering the market. 

Earth sensing is one of the space applications that’s seeing large growth, with a demand for satellites that can sense the Earth’s surface using cameras, radar, and even radio-frequency analytics. This data is used for a variety of purposes, from agriculture (assessing crop health) to finance (counting cars in shopping center parking lots to draw conclusions about the economy).

A common problem Earth-sensing satellites wrangle with is that the amount of data they can collect is often larger than the amount of data they can beam down to Earth. In addition, some of the collected data isn’t actually useful for its intended purpose (for example, did it capture a picture of the target, or were there clouds in the way?).

Sifting the useful data from the “wasted” data is done today by computers on the ground, increasingly using deep-learning techniques. It would be much more efficient if the useful data could be separated from unhelpful data by the satellite itself, which would then be able to send only useful data—and more of it—to customers on Earth. Deploying AI algorithms using deep learning aboard the satellite enables this data separation. Companies that are able to do so will have an advantage in the market, and engineers with these skills will be in high demand!

Given that there’s a long history of increasing spacecraft autonomy ever since the early days of spaceflight, and that machine learning and deep learning can facilitate solving some of the common problems in the space industry, the industry is well-positioned to succeed in applying AI. Many contend that the inherent conservatism and risk aversion of the industry will prevent AI from being adopted.

There is, however, a flip side to that argument. Those very traits can help the industry figure out the greatest returns on investment for AI, without substantially increasing the risk to expensive missions and space vehicles; for example, carefully selecting the proper functions to apply AI techniques, such as in the case of the Curiosity Rover. The industry culture also is changing, with new companies willing to try new things.

AI and Engineering Software

Once AI is adopted onboard spacecraft, its use may become more widely accepted. More critical use cases for deep learning include vision-based navigation for rendezvous and proximity operations as well as hazard detection and avoidance for lunar or planetary landings. In addition, engineering software will continue to play a key role in the advancement of AI. 

Setting aside machine learning and deep learning for a moment, even traditionally coded AI algorithms benefit from tools that manage the design of increasingly autonomous spacecraft. The trend is toward enabling spacecraft to own complex decision-making capabilities that were previously left to a human operator. One such example is whether to abort or continue with an autonomous docking between two spacecraft.

Modern engineering tools abstract the design by providing views that hide the complexity, allowing engineers to better understand the potential behaviors of the system. For instance, modeling decision-making logic as digital executable state charts rather than if-then-else statements in code enables reviewers to have a clearer understanding of the possible logic paths. This approach of design abstraction, provided by tools such as MATLAB and Simulink, helps engineers understand the impact of individual functions on system behaviors during the design phases. 

Despite improvements in design abstraction, design errors in these complicated AI systems can be subtle and hard to catch. For example, assessing the robustness of a vision-based sensing and perception algorithm in conditions involving lighting and perspective is difficult through review alone and requires extensive simulation and testing.

Design tools such as MATLAB and Simulink are most effective when they provide capabilities not only for more easily training algorithms through features such as automated labeling, but also for early verification and validation through simulation. This enables requirements-based testing well before algorithm deployment on hardware.

Machine learning and deep learning, the next frontier in AI, inherently require engineering software because the machine-learning and deep-learning models are generated by computers, not humans. However, software tools used by data scientists to prototype machine-learning and deep-learning techniques often have a steep learning curve and aren’t always designed for production deployment of algorithms, which can lead to inefficient development cycles.

Thus, there’s a growing need for aerospace engineers to have access to tools such as MATLAB and Simulink for machine-learning applications that they’re already familiar with, and that are designed for engineering, as opposed to data science, workflows.

Sponsored Recommendations

Comments

To join the conversation, and become an exclusive member of Electronic Design, create an account today!