Op-ed | The use of AI in space systems: opportunities for mission improvement
From analyzing the terrain on Mars to enhancing communications between satellites and ground communications, artificial intelligence (AI) is playing an increasing important role in space operations and exploration. It is a capability with numerous applications and vast promise for the data-rich and complex environment of space.
For example, many organizations with space operations are recognizing AI’s power to perform complex tasks quickly and accurately and enhance decision-making. Adoption of AI throughout the space domain can help improve mission effectiveness and resiliency.
Furthermore, today’s space environment is congested, complex, and contested—a warfighting domain that is no longer a sanctuary for US or allied space assets. AI has the potential to significantly improve domain awareness and command and control decision-making and increase the resilience of satellite and the networks that connect them.
For these potential advancements to reach their full potential, however, we must strengthen the security of, and trust in, AI technology. Consider the AI-generated analyses that aid human decision-making. Can commanders and operators trust that the algorithms behind these analyses were objectively formulated, with appropriate data and without bias? Can they be confident that the data being used hasn’t been corrupted or manipulated by adversaries? These are important questions to answer to ensure when lives and mission-critical assets are at risk.
Examples follow of ways to use AI to strengthen critical space missions and what’s needed for users to trust this technology.
Leveraging AI to improve space domain awareness
Space is getting more and more crowded. Orbiting the Earth today are over 2,600 active satellites, more than 34,000 objects of 10 centimeters or more, and over 900,000 pieces of space debris between 1 and 10 centimeters. All are moving in different orbits, across different planes, and at different speeds. Having a clear picture of this complex environment is an important first step for operating safely in space and protecting space assets.
Is an object space debris or a maneuvering satellite? What is its predicted path, and what are its capabilities?” AI operates on many levels to help operators answer questions like these and respond appropriately.
First, organizations can use available data and an AI system to generate a comprehensive catalog of known and observed Earth-orbiting objects. This same AI system could continuously monitor and assess the probability of collisions, alerting satellite and spacecraft operators in the event of heightened risk.
Here’s how such a scenario might play out. Once operators have identified a satellite at risk with the aid of their “space catalog,” AI can help them decide the best course of action for protecting that satellite. Such an AI/machine learning system would combine traditional modeling and simulation with a deep-learning network and collision avoidance algorithms to rapidly produce a list of potential maneuvers for avoiding the space object.
In space as on earth, each potential avoidance maneuver comes with a variety of pros, cons, and interrelated impacts. For example, one course of action may reduce fuel expenditures along with operational impact. Another may help operators “look ahead” to minimize downstream interference or collisions.
Organizations can program an AI/machine learning system to present the most appropriate avoidance maneuvers based on the most relevant criteria to the mission at hand. Users—the “humans in the loop”—can then use their judgment and mission knowledge to choose among the options and execute the most appropriate maneuver to keep valuable space assets out of harm’s way.
In time-sensitive situations, such an AI/machine learning system would deliver recommended solutions in minutes, versus the hours or days required with more traditional methods. This is the power of AI to accelerate domain awareness—and reduce costly collisions—in today’s increasingly crowded space environment.
Harnessing AI to accelerate command-and-control decision-making
Another area where AI offers great potential is in command-and-control decision-making, particularly when assets come under threat with very little time to react.
Consider a scenario in which an operator must protect space assets against a direct ascent anti-satellite (ASAT) attack. In such a situation, the operator may have only minutes to decide what to do. AI and data analytics put a previously near-impossible task within reach: helping decision-makers efficiently analyze vast amounts of data and swiftly arrive at a set of potential actions.
The AI system absorbs ASAT trajectory data to identify possible targets. It then rapidly develops multiple courses of action, which could include maneuvering, countermeasures, or engaging in offensive or defensive activities. Using machine learning, the system sifts through many possible courses of action, taking into account interrelated consequences and downstream implications. Operators and commanders then receive a timely menu of optimized choices, which accelerates command-and-control decision-making and strengthens space defense in mission-critical situations.
Strengthening resilience through machine learning and automation
In response to commercial demands for global communication and data transport, satellite constellations and the networks that connect them are becoming larger and more complex. These networks are also becoming increasingly vulnerable to increasingly sophisticated kinetic and non-kinetic threats.
By adopting AI into space systems, operators can mitigate these threats and make space networks and constellations more resilient. Organizations can use AI to quickly scan through data to recognize network vulnerabilities. They can then apply AI algorithms to “heal” or self-adapt in response, to ensure all nodes within the network are reconnected. Organizations can also embed self-learning algorithms into the satellites themselves, to make them more self-sufficient and more resilient if up-link and down-link communications with ground operations are lost.
Furthermore, AI can automate the monitoring of a satellite’s “health status,” the resolution of anomalies, and the execution of defensive actions against threats. Automating such tasks on satellites themselves can accelerate these actions and free operators to concentrate on more complex, mission-critical work.
Building trust through algorithm development and operator training
As with any application of a new technology, in space or elsewhere, security and trust are paramount to adoption and effectiveness. AI security begins with the development of the AI algorithms. Organizations must ensure the pedigree of the data used to train the algorithms, ensure that algorithms are developed with as little bias as possible, and maintain security throughout the software development process and data storage.
Additionally, organizations with space assets and systems will need to train operators in AI and machine learning, which includes an understanding of how AI systems are built and designed. Operators must also have a complete understanding of the capabilities and limitations of their AI-powered solutions. Only through comprehensive training and education, as well as implementing secure processes, will operators and decision-makers trust AI systems enough to use them to enhance their missions.
As the space environment rapidly evolves and proliferates with new users, new capabilities, and increasingly sophisticated threats, deterring and defending our space assets has become both an imperative for national security and a far more complicated task. Through improving space domain awareness, accelerating command-and-control decisions, making satellites and their networks more resilient, and more, AI solutions offer a transformative opportunity for protecting, improving, and enhancing space missions and helping the United States maintain space dominance. To realize AI’s great promise, however, we must also make sure that AI systems are securely developed and maintained and that commanders and operators have the training and understanding required to trust this transformative technology.
Chris Bogdan is the former head of the F-35 program and a senior vice president in Booz Allen Hamilton’s aerospace business. Saurin Shah is an artificial intelligence (AI) leader in Booz Allen Hamilton’s national security business.