Adaptive Operation Planning and Intelligent Decision-Making System For Lunar Soil Drilling Based on Reinforcement Learning

Autonomous systems play a critical role in the early planning stages of health management and prognostics to ensure reliable and efficient operation in space missions. Presently, there are several technical obstacles that might prevent the use of autonomous robotic drilling along with sampling on the moon. To ensure the successful and reliable operation of regolith sampling missions, it is necessary to study robotic self-adaptive drilling methods on the moon's surface. Previous work by the authors has connected ground drilling sample data with mechanical-soil interaction processes, investigated how lunar soil affects core drilling and load fluctuation states, and more. Afterwards, they offered a plausible assessment of the lunar soil sequence's condition during the drilling phase. The residual useful life (RUL) is predicted using the reliability-adaptive systems technique to avoid maintenance conflicts. In the proposed research, a lunar surface having many systems, such as one base and six rovers, is shown. The rovers drill into the moon's surface to bring back soil samples to Earth. While the lunar materials manufacturing facility is located within a level area, the rovers' duty is to collect dirt or rocks from areas surrounding small obstacles put on the edge of the flat surface. The best way to determine RUL is to monitor the drill's wear. When determining the drill failure rate, a lot of random variables are considered. Some examples of these include radiation, soil temperature relative to the rover, and soil density. Entire Assistance This study introduces a method for predicting drilling forces using machine learning. This study improves control settings and uses a novel method for real-time force forecasting to deliver intelligent lunar drilling that is both safe and efficient. Both the penetrating force and the spinning torque may be precisely predicted by the system. The study also reveals the spatial and temporal relationships between the different drilling state variables. The results of the case study demonstrate that (1) the proposed method of fusing spatiotemporal features has the potential to provide accurate real-time predictions (r = 0.903, mSE = 0.109).