Voice Assistant Tech for Spacecraft Anomaly Detection, Resolution Tested

Newsweek

11-07-2025

Based on facts, either observed and verified firsthand by the reporter, or reported and verified from knowledgeable sources.
A research team at Texas A&M University is studying the use of Siri-like virtual assistant technology for use in space. The technology, known as Daphne-AT, could be used to help drive decision-making that requires large amounts of real-time and previously learned data.
On Earth, virtual assistants (VA) are typically humans working remotely who provide administrative, technical or creative support to executives utilizing shared technology platforms. Daphne-AT allows for less human interaction.
"Daphne-AT uses a mix of logic and data-driven decision-making to help astronauts make informed decisions when a problem arises on a spacecraft," said Dr. Daniel Selva, an associate professor of aerospace engineering at Texas A&M's Department of Aerospace Engineering. "The VA uses spacecraft data in real time to detect anomalies, hypothesize what is causing the issue and give astronauts information on how to solve the problem."
The VA tool uses real-time data streams of a spacecraft's environmental controls and life support systems, determining what is within and outside of baseline tolerances. When there is a level that is outside the established tolerances, Daphne-AT could alert astronauts and command, giving them the information to solve the problem.
Researchers at the school used a mix of virtual reality (VR) and laboratory simulations to study the effectiveness of Daphne-AT. During VR trials, researchers tested the amount of time it took engineers to resolve five anomalies with and without Daphne-AT's assistance, finding that using the VA, participants were able to solve anomalies faster, with less mental workload.
Old Glory on the Red Planet, a close-up view of the United States flag plate on NASA's Perseverance was acquired on June 28, 2025 (the 1,548th day, or sol, of its mission to Mars), by...
Old Glory on the Red Planet, a close-up view of the United States flag plate on NASA's Perseverance was acquired on June 28, 2025 (the 1,548th day, or sol, of its mission to Mars), by the WATSON (Wide Angle Topographic Sensor for Operations and eNgineering) imager on the turret at the end of the rover's Mars robotic arm. More
NASA/JPL-Caltech/MSSS
Additionally, researchers took their trial on the road, testing NASA engineers at the Human Exploration Research Analog (HERA) facility at NASA's Johnson Space Center. Eight crews of four participants stayed in HERA for 45 days each and the results differed from those at the university, showing that Daphne-AT's assistance did not significantly affect the participants' timing in resolving anomalies.
"Part of the difference is likely due to the differences in training and experience levels of participants," said Selva. "There are only so many anomalies we can present to participants, and the trained professionals had more time with the VA and more opportunities to resolve different anomalies."
The technology is still in development, the result of Selva's lifelong fascination with space. "For the longest time, I have been fascinated by two things: space exploration, and intelligence (both human and artificial)," he told Newsweek.
"I became particularly interested in human-AI teaming during my dissertation (2008-2012), when I worked on intelligent decision support systems to help NASA engineers design and plan space missions. I continued with that idea as an Assistant Professor at Cornell, where we developed our own language models to serve as natural language interfaces for our AI assistants. These were ridiculously small models by today's standards, with hundreds to thousands of parameters instead of hundreds of billions. They were very reliable for our particular problem but limited in scope and with a mechanistic feel in their responses, due to the use of pre-coded answer templates," he continued.
Selva is working on the project with Dr. Ana Diaz Artiles, his wife: "We wrote the original proposal to NASA to study AI assistants to support astronauts with spacecraft anomaly resolution. Both Ana and I worked for Arianespace for several years as members of the Ariane 5 launch campaign – that is when we started dating! So we were very familiar with anomaly resolution in high-stakes operational environments.
"Moreover, the environment at Texas A&M University was perfect to develop this project, as it enabled us to leverage the amazing experience of former astronaut and Professor Dr. Bonnie J. Dunbar, a co-I in the project. We were also able to recruit a top-notch statistician to help us make sure we are analyzing our data correctly. It was one of those proposals where all the pieces fell together nicely. And luckily, it was funded. We started the project in 2019, and the field has evolved tremendously since then, particularly with the disruption of large language models, which have enabled immensely better interfaces for our AI assistants. The experience we have gathered over the last 6 years will help NASA develop guidelines for future development of similar AI assistants."
The results of the initial Daphne-AT study were published in the Journal of Aerospace Information Systems by Selva and Diaz Artiles. Co-authors include Dr. Bonnie J. Dunbar, an aerospace engineering professor at Texas A&M and former NASA astronaut and Dr. Raymond Wong, an associate professor of statistics at Texas A&M.