Cosmo Bee

A Hall thruster in operation. / Courtesy of KAIST

 

A Hall thruster for CubeSats, developed by KAIST researchers using artificial intelligence (AI), is scheduled to undergo in-space performance validation aboard a CubeSat on Nuri’s fourth launch in November 2025.

 

KAIST announced on February 3 that a team led by Professor Wonho Choe of the Department of Nuclear and Quantum Engineering had developed an AI method capable of predicting the thrust performance of Hall thrusters—the engines used in satellites and space probes—with high accuracy.

 

Hall thrusters are high-efficiency propulsion devices that use plasma. Their high propellant efficiency enables them to significantly accelerate satellites and spacecraft while using little propellant. They are widely used for missions ranging from maintaining formation flight in SpaceX’s Starlink constellation to providing propulsion for comet and Mars exploration.

Accurate performance prediction from the design stage is essential for developing high-efficiency Hall thrusters. Conventional methods, however, could not precisely model the complex plasma behavior inside Hall thrusters or were limited to specific operating conditions, resulting in low prediction accuracy.

 

To address these limitations, the research team developed its own AI-based performance-prediction method. This significantly reduced the time and cost required for repeated Hall-thruster design, fabrication and testing. The team also used its numerical analysis tool to generate 18,000 training data points, which were applied to the AI model to improve prediction accuracy.

 

Prediction errors were also low. Comparisons with experimental data from 10 Hall thrusters developed in Korea showed an average error of less than 10%. The average error was below 5% for the team’s 700 W and 1 kW Hall thrusters and below 9% for a 5 kW high-power Hall thruster developed by the U.S. Air Force Research Laboratory.

 

The study demonstrated that the AI prediction method could be broadly applied to Hall thrusters with different power levels.

 

“This AI technique can be applied not only to Hall thrusters but also to research and development of ion-beam sources used in various industries, including semiconductors, surface treatment and coating,” Professor Choe said.

 

 

Original article : KAIST Develops AI-Based Technology to Predict CubeSat Engine Performance