Visual SLAM-Based Robotic Mapping Method for Planetary Construction

Area of Science:

• Planetary Science
• Robotics
• Computer Vision

Background:

• Recent discoveries of lunar and Martian resources like water-ice and lava tubes drive interest in in-situ resource utilization (ISRU) for long-term space exploration.
• Current 3D terrain mapping methods for planetary surfaces, primarily based on orbiter imagery, lack the necessary resolution for construction purposes.
• Developing high-resolution, localized 3D maps is crucial for robotic and human base construction in extraterrestrial environments.

Purpose of the Study:

• To introduce a novel visual simultaneous localization and mapping (SLAM)-based robotic mapping method for planetary rovers.
• To improve the resolution and accuracy of 3D terrain mapping for extraterrestrial construction applications.
• To address the limitations of current mapping techniques in homogeneous and unstructured planetary terrains.

Main Methods:

• The proposed method utilizes a stereo camera system mounted on a rover for visual SLAM.
• The S-PTAM (Small-Planet Augmented Tracking and Mapping) framework is employed as the core SLAM algorithm.
• Self-supervised deep learning is integrated to generate enhanced disparity maps, improving mapping in challenging environments.

Main Results:

• The visual SLAM method demonstrates enhanced mapping capabilities in simulated planetary terrain conditions.
• The integration of self-supervised deep learning significantly improves the generation of accurate disparity maps.
• The developed technique provides higher resolution 3D terrain maps compared to traditional orbiter-based methods.

Conclusions:

• The proposed visual SLAM-based robotic mapping method offers a viable solution for high-resolution 3D terrain mapping on planetary surfaces.
• This advancement is critical for enabling effective base construction and long-term human exploration on the Moon and Mars.
• The validated performance in emulated environments suggests strong potential for real-world application in future planetary missions.