How AI and Simulation are Driving Space Innovation
Our Expert’s Insights on the Intersection of AI, Simulation Tools, and Space Technology
AI uses massive amounts of data to help us explain, predict, and optimize; modeling and simulation tools leverage data to gather evidence and test different scenarios. Integration of these two fields holds promising implications for space systems innovation.
Currently, these capabilities are used in space for more traditional purposes, e.g., finding novelties in data, helping space operators see something they may have missed, and conducting preventative maintenance on a spacecraft.
Future use cases for AI and simulation include enabling more efficient and precise spacecraft maneuvers, automatically identifying and tracking objects to ensure space situational awareness, and allowing robotic missions to make decisions about what to explore on other planets. In addition, AI and simulation can help improve the accuracy of weather simulations and provide predictive analytics for spacecraft maintenance and optimization.
At the SmallSat Symposium in Mountain View, February 7–9, 2023, a panel of experts moderated by Lori Gordon of Aerospace explored current and future opportunities to use AI and simulation in space applications. Panelists shared insights about emerging trends, including expanding beyond satellite data analysis to applications integrated with an operator’s experience.
We caught up with Lori Gordon to find out more about the interesting topics discussed on the panel.
How are AI and modeling simulation tools being used in space today? Who is using these tools?
Both commercial space and government are essentially operating robots in space right now — running AI/ML algorithms on spacecraft to anticipate potential conflicts, using AI to identify gaps in security, and optimizing systems to get more out of existing and future spacecraft.
Current use cases include leveraging the AI body of knowledge to make good design decisions, exploring the repair and refueling of spacecraft on-orbit, as well as run prognostics, and conducting predictive maintenance on spacecraft. AI and simulation tools are used to train operators to recognize adversarial activity and understand how our services will perform in special operating environments. These tools are also essential for analyzing and interpreting data from surveillance activities.
Are government and the commercial space industry collaborating on these systems?
Governments are realizing what seems like an infinite number of use cases for developing digital models, or ‘digital twins.’ For example, the EU’s European Centre for Medium-Range Weather Forecasts (ECMWF) is creating a high-precision digital model of the earth, known as Destination Earth (DestinE).
DestinE is a new Earth system simulation and observation capability in support of the EU’s green transition initiative. It leverages both weather prediction and artificial intelligence advances.
The integrated model will provide greater visibility into issues related to climate change, biodiversity, and deforestation, as well as food security, polar region changes, and sea level rise. The capability will be used to predict activities needed in the next several decades to support climate adaptation and mitigation policies.
What are some of the current challenges with integrating AI and modeling and simulation tools for space applications?
Whether AI will be able to have complete autonomy is an often-debated topic at cocktail parties. Developers are climbing a steep hill to fully trust AI. For now, it’s used as an assistant. Trust comes in training, verification and validation, and other aspects. While engineers go to great lengths to achieve trust goals and objectives, there are still significant challenges when using AI-based onboard autonomous systems.
Costs for obtaining training data are significant, so there may be a business model in sharing data sets. Limited computational power and memory storage in space could pose challenges to the use of AI algorithms in spacecraft. Additionally, adversarial attacks are a particular area of focus. Engineers must consider potential failure modes and how they can be identified and mitigated.
AI has a long way to go, particularly given the environmental constraints in space. In five to ten years, we will see new algorithms and computational power accelerate the trajectory of autonomous systems for space.
Are space applications exploiting the state of the art in AI and simulation? What advances in simulation and AI tools from other industries do you anticipate having impacts on space missions?
Space is a tough environment with different parameters. Due to cost and risk tolerances, other industries are ahead of the space industry in thinking about where and when humans no longer need to be in the loop. We take inspiration from other industries, including healthcare and banking, to learn what use cases could be extended to space. Developers rely on analogies a lot!
Planetary defense is one area rapidly incorporating AI and simulation. The European Space Agency’s Hera mission will use an approach similar to self-driving cars to maneuver itself through space toward an asteroid. While other deep-space missions have a driver back on Earth, Hera will integrate data from sensors and build a model of its surroundings to make autonomous onboard decisions.
Recent military exercises conducted by the U.S. Air Force Research Laboratory (AFRL) and the U.K. Defence Science and Technology Laboratory (DSTL) demonstrated AI technology in new operational environments. The two exercises were designed to meet constantly changing mission conditions and military needs.
The goals were to improve situational awareness, unify the inputs from sensors, networks, and systems with users, and accelerate the decision-making timeline. As space becomes a more contested environment, the use of AI will be critical to future needs.
What do future applications of this integration look like?
Future areas for exploration are across the satellite mission lifecycle — from design to testing to operations. These applications will connect data between different tools that haven’t talked to each other previously. While these tools do their jobs very well independently, the lack of interconnectivity leads to problems with scale when crossing lifecycle stages.
The ideal is a seamless, interconnected system of systems where all software tools used in system design, data management, and satellite operations are digitally integrated so operators can automate across an entire mission.
We may see business models that include developing and validating data sets needed to feed machine learning algorithms. We may apply AI to the problem of managing collision avoidance maneuvers in space, many of which still rely on manual interactions between humans. Ideally, AI-driven processes and automation could help close the gaps between operational needs, system design, tactics development, and mission execution.
Lori W. Gordon is Systems Director in the External Engagements Office of The Aerospace Corporation. She is a technology strategist in national and homeland security, cybersecurity, and infrastructure risk and resilience. She is an advisor to ISO, ANSI, and NIST technical working groups and is a fellow at the National Security Institute. She has also served on curriculum advisory boards in the areas of cybersecurity and infrastructure security, law and government, and resilient design. Lori has a bachelor’s degree in geography from the University of Maryland, and a master’s degree in public administration from the University of Massachusetts, Amherst.
