The KAIST research team announced that the AI-designed Hall thruster created for CubeSats will be installed on the KAIST-Hall Effect Rocket Orbiter (K-HERO) CubeSat to demonstrate its in-orbit performance. The study was published in Advanced Intelligent Systems.
(From left) PhD candidate Youngho Kim, Professor Wonho Choe, and Ph.D candidate Jaehong Park from the Department of Nuclear and Quantum Engineering. Image Credit: KAIST
Hall thrusters are high-efficiency electric propulsion devices that use plasma technology. They are essential space technology for missions such as NASA's Psyche asteroid mission and SpaceX's Starlink constellation. The fourth launch of the Korean Launch Vehicle, the Nuri rocket (KSLV-2), is planned for November 2025.
Under the direction of Professor Wonho Choe, the research team from the Electric Propulsion Laboratory of the KAIST Department of Nuclear and Quantum Engineering announced on February 3rd that they had developed an AI-based method to precisely forecast the performance of Hall thrusters that power satellites and space probes.
Hall thrusters have a high fuel efficiency, using a small amount of propellant to accelerate satellites or spacecraft significantly while generating a large amount of thrust in relation to power consumption. Hall thrusters are widely used in various space missions because of these benefits, such as deep space missions like asteroid exploration, deorbiting maneuvers for space debris mitigation, and the formation flight of satellite constellations.
As the space industry expands in the NewSpace era, the need for Hall thrusters appropriate for various missions is growing. Accurately forecasting thruster performance from the design stage is crucial for quickly developing highly effective, mission-optimized Halls.
Conventional approaches, on the other hand, are limited in their ability to handle the intricate plasma phenomena found in Hall thrusters or are only applicable in certain situations, resulting in a lower prediction accuracy.
The research team greatly reduced the time and expense involved in the iterative design, fabrication, and testing of thrusters by creating a highly accurate AI-based performance prediction technique.
Professor Wonho Choe's group has spearheaded research on electric propulsion development in Korea since 2003. The team used 18,000 Hall thruster training data points produced by their proprietary numerical simulation tool and a neural network ensemble model to forecast thruster performance.
The in-house numerical simulation tool, created to simulate thrust performance and plasma physics, was essential in supplying high-quality training data. With an average prediction error of less than 10%, the simulation's accuracy was confirmed by comparisons with experimental data from ten KAIST in-house Hall thrusters.
Based on thruster design parameters, the trained neural network ensemble model functions as a digital twin, accurately forecasting the Hall thruster performance in seconds.
Notably, it thoroughly analyses performance parameters like thrust and discharge current while considering Hall thruster design variables like magnetic field and propellant flow rate, which are difficult to assess with conventional scaling laws.
This AI model showed an average prediction error of less than 5% for the in-house 700 W and 1 kW KAIST Hall thrusters. For a 5 kW high-power Hall thruster created by the US Air Force Research Laboratory and the University of Michigan, it showed an average prediction error of less than 9%. This demonstrates the AI prediction method's wide range of applications across Hall thruster power levels.
The AI-based prediction technique developed by our team is highly accurate and is already being utilized in the analysis of thrust performance and the development of highly efficient, low-power Hall thrusters for satellites and spacecraft. This AI approach can also be applied beyond Hall thrusters to various industries, including semiconductor manufacturing, surface processing, and coating, through ion beam sources.
Wonho Choe, Professor, Electric Propulsion Laboratory, Department of Nuclear and Quantum Engineering, KAIST
Professor Choe mentioned, “The CubeSat Hall thruster, developed using the AI technique in collaboration with our lab startup—Cosmo Bee, an electric propulsion company—will be tested in orbit this November aboard the K-HERO 3U (30 x 10 x 10 cm) CubeSat, scheduled for launch on the fourth flight of the KSLV-2 Nuri rocket.”
Ph.D candidate Jaehong Park was the first author of this study.
This study was funded by the Space Pioneer Program (200mN High Thrust Electric Propulsion System Development) of the Korean National Research Foundation.
Journal Reference:
Park, J., et al. (2025) Predicting Performance of Hall Effect Ion Source Using Machine Learning. Advanced Intelligent Systems. doi.org/10.1002/aisy.202400555