Self-adaptive Synchronous Localization and Mapping using Runtime Feature Models

Mobile autonomous robotic systems need to operate in unknown areas. For this, a plethora of simultaneous localization and mapping (SLAM) approaches has been proposed over the last decades. Although many of these existing approaches have been successfully applied even in real-world productive scenarios, they are typically designed for specific contexts (e.g., invs. outdoor, crowded vs. free areas, etc.). Thus, for different contexts, different SLAM algorithms should be used. In this paper, we propose a feature-based classification of SLAM algorithms and a reconfiguration approach to switch between existing SLAM implementations at runtime. By this, mobile robots are enabled to always use the most efficient implementation for their current contexts.