Realization of vision-based navigation and object recognition algorithms for the sample return challenge

Abstract

We present the improvements to AERO, the Autonomous Exploration Rover, developed for the 2014 NASA Sample Return Robot competition with the intent of enabling more robust and reliable autonomous operation for sample return rovers. The competition requires the robot to navigate a large outdoor area, find and collect various geologic samples, and return to the starting platform all autonomously and utilizing only space compatible technologies. We highlight improvements made in the implementation and deployment of the vision, navigation, and planning systems. We describe the process of modifying the software to be closer aligned with ROS standard practices, resulting in more predictable and stable operation. We conclude by providing a roadmap for the integration of multiple heterogeneous systems in a shared control framework to enable efficient exploration of large unknown environments.