No rover to date has traveled to or imaged the lunar poles. The lack of atmosphere on the Moon and low elevation of the sun at the lunar poles creates regions of brilliant illumination, absolutely black shadow, and scenes containing both extremes. Rover perception at the lunar poles must create accurate terrain models in these unprecedented circumstances. The research investigates stereo geometry, algorithms, lensing, exposures, material appearance, and terrain illumination unique to lunar poles. Additional considerations include processing on space-relevant computing and camera hardware and achieving the short cycle times essential for achieving continuous rover motion. This research formulates, implements, and evaluates an illuminated stereo system and software specialized for lunar polar perception and deployment to the Moon. The resulting perception system will guide Carnegie Mellon's autonomous MoonRanger micro-rover that will map ice on the lunar south pole in 2023.

The perception system generated in this work is being incorporated into MoonRanger, compelling distinct simplicity in design. Key features include eight infrared laser dot projectors to illuminate terrain, a fixed lens aperture size of F/4.0, a 16 millisecond exposure, and the semi-global matching stereo algorithm that enables continuous imaging and stereo depth perception on lunar polar terrain without switching camera hardware or software when imaging different extreme illumination conditions. The research additionally suggests an extension to the flight perception system that leverages High Dynamic Range techniques within the 16 millisecond exposure to minimize image whiteout in edge-case roving conditions.

65 pages

Thesis Committee:
William "Red" Whittaker (Chair)
David Wettergreen

Srinivasan Seshan, Head, Computer Science Department
Martial Hebert, Dean, School of Computer Science

Return to: SCS Technical Report Collection
School of Computer Science

This page maintained by reports@cs.cmu.edu