The AMADEE-20 Robotic Exploration Cascade: An Experience Report

As part of the AMADEE-20 Mars analog mission, an exploration cascade for the remote sensing of extraterrestrial terrain was implemented. For this purpose autonomous mobile as well as aerial robots were conceptualized and used in an iterative exploration cascade. First, satellite imagery was analyzed using Deep Learning methods. This large but low resolution image was used as a basis for mission planning as well as for an initial traversability analysis for the ground robot. High resolution low orbit aerial images of a defined area of interest were then recorded, and utilized together with the low resolution map were in a path planning tool for the mobile robot. The mobile robot uses a navigation concept of global and local planning as well as local perception to navigate to the area of interest. Experiments during the AMADEE-20 mission in the Israeli Negev Desert validated the exploration cascade.