SLAM for SCOUT: A ROS2-Based Multi-Sensor System for Compliant Rovers

Abstract

This paper investigates a ROS2-based Simultaneous Localization and Mapping (SLAM) system for a ground-based multi-body planetary cave rover. Developed within the German Aerospace Center (DLR), the Space Cave explOration UniT (SCOUT) aims to autonomously explore volcanic cave systems on the Moon and Mars. A robust SLAM system for accurate environmental perception and mapping is therefore a crucial prerequisite for mission success. The presented system is based on visual perception, with data generated by two opposing RGB-D cameras, mounted inside the front and rear modules of the rover. Each of them is accompanied by an efficient LED matrix lighting system specifically designed for the mission environment. The unique architecture of the SCOUT rover, which consists of three flexibly connected aluminum modules with two rimless wheels each, presents novel challenges for SLAM deployment. In particular, its compliant structure, coupled with erratic motion during travel due to the wheel geometry and unknown surface conditions, makes accurate mapping difficult. In this context, the compliant connection between the two cameras requires an additional pose estimation system, based on multiple IMU sensors. To evaluate the developed SLAM system, a portable demonstrator with SCOUT-like characteristics was built. Tests conducted on this platform confirm the system's ability to produce high quality and accurate maps of a karst cave system. This work provides further insight into the specific requirements and challenges of a SLAM system for the SCOUT rover. It also demonstrates the system's effectiveness in mapping a cave system and provides valuable guidance for the future development of autonomous ground-based cave rovers.