Incremental Bundle Adjustment Techniques Using Networked Overhead and Ground Imagery for Long-range Autonomous Mars Rover Localization

This paper presents an image network-based technology designed for long-range onboard rover localization. This integrated overhead and ground image network provides a global geometric configuration. Incorporation of this image network with our unique incremental bundle adjustment and visual odometry methodologies ensures optimal localization accuracy over the entire landing site area. Progresses in rover traverse design, accuracy analysis and automation of cross-site tie point selection are reported. We expect to achieve onboard real-time rover localization accuracy of one percent (1 m over a traverse of 100 m) and an overall accuracy of 0.2 percent (10 m for a traverse > 5 km) using Earth-based computations.