Autonomous satellites will be able to restore lost data about their movement even after failures and interference. The Moscow Aviation Institute (MAI) has created a prototype artificial intelligence-based system that analyzes the trajectory of a spacecraft and fills in lost coordinate sections.
The development was presented by MAI students. Project author Marta Artyomova said that the algorithm studies the satellite's movement before and after a failure, then restores the missing data taking into account the entire flight trajectory.
Real data on the movement of spacecraft was used to train the system. The model was loaded with the coordinates of several satellites for five days, recorded every minute. In total, specialists processed about 288 thousand records.
During testing, interference, gaps, and coordinate shifts were intentionally added to the data to check the algorithm's stability. According to the test results, the average recovery error was about 33 centimeters.
The developers believe that the technology will be useful not only for satellite constellations but also for unmanned transport, where precise navigation is critically important when signal is lost.
Currently, the project is at the prototype stage. The team plans to expand the amount of data for system training, conduct additional tests, and integrate the solution into existing navigation complexes.
