Reinforcement Learning based GNC for a high-fideLity emUlator module of eXtraterrestrial body LANDER

Descent and landing (D&L) are arguably among the most critical phases for the success of upcoming space exploratory missions. They can face difficult challenges due to extremely uncertain and variable physical environments of the target planet or asteroid bodies. These challenges require fully autonomous Guidance, Navigation and Control (GNC) algorithms to achieve mission success. Reinforcement Learning (RL) is a promising technique that has the advantage over traditional engineering GNC systems to adapt to unknown situations and circumstances that are hard to account for manually. The integration of RL with the GNC subsystem in a realistic infrastructure (such as LUXLANDER) could be a game-changer for upcoming space-exploration missions. Indeed, using RL-based algorithms in the real world it is crucial to develop a suit of the trained control policies for the given problem in a highly accurate simulation environment and to demonstrate their robustness with a realistic Verification& Validation (V&V) test-bench.

Project Information

Duration: January 15, 2022 to January 15, 2025. 3 years

Funding Source
Principal Investigator(s)
Prof. Miguel Olivares-Mendez, SnT-UL
Matteo El Hariry, Dr. Baris Yalcin Can