A Crater-based Optical Navigation Approach for Precise Spacecraft Localization

Lunar exploration represents the launching pad to develop novel navigation approaches in preparation of future missions to Mars and beyond. Innovative navigation systems, which make use of machine learning techniques for image processing, are currently under development to aid in the execution of challenging tasks, including autonomous pinpoint landing operations. Due to the recent push to revisit the Moon both from a scientific and commercial perspective, this represents a key capability to support, for example, the exploration of key scientific sites (e.g., permanent shadowed regions) or the provisioning of resources to future human settlements on the lunar surface. In this work, a crater-based navigation system is presented that leverages a machine learning approach to detect craters in the images captured by the onboard navigation camera. By matching the craters detected the images with known craters from an onboard database, accurate localization is carried out, enabling trajectory adjustments to reach the target landing site. In addition to crater-based optical measurements, orientation estimates provided by star trackers are jointly processed with data from an onboard Inertial Measurement Unit (IMU) in a visual-inertial framework to enable a robust estimation of the spacecraft's orbital state and attitude throughout the descent and braking phase. Simulation campaigns based on the NASA VIPER mission are carried out in a realistic synthetic lunar environment to demonstrate the capabilities and performance of the multi-sensor navigation software.