Efficient On-Board Autonomy Using RFS-SLAM for Robust Lunar Surface Mobility

Navigation on the lunar surface is difficult due to a lack of terrestrial equivalent infrastructure and landmarks. This Phase I program will develop a solution for robust lunar surface mobility operations of autonomous robotic agents within this sparse feature environment by using Random Finite Set (RFS) theory integrated with Simultaneous Localization and Mapping (SLAM) techniques. By using RFS-SLAM performance is improved by reducing the number of particles used internally in the particle filters allowing for reduced computational operating costs necessary for accurate results. The RFS-SLAM can handle additional clutter allowing for a wider range of sensors usable for SLAM. Phase I will focus on evaluating RFS-SLAM and its performance, employing flight ready sensors and processors as models using established robotic software frameworks and simulation tools. Point cloud generating sensors to be evaluated with this Lunar RFS-SLAM include stereo optical+thermal cameras and LIDAR. Odometry generating wheel odometers, inertial measurement units, and sun trackers will be evaluated for further improvements. In order to support near term operations, the RFS-SLAM algorithms will be implemented in Python and integrated into ASTER Labs’ SWARM Toolset. Additional software frameworks and tools that offer ease of integration into NASA Lunar missions will be architected, including ROS, along with its safety optimized derivative SpaceROS. The SWARM Toolset with the RFS-SLAM module will be used in simulations to evaluate system performance within a reduced development time. Python prototype software will serve as validation tests for future optimized implementations. The software developed under this project will apply to commercial markets of automated navigation on resource and infrastructure constrained platforms such as self-driving vehicles and indoor robots.