Simultaneous Localization And Mapping (SLAM) is a process in which sensor data is used to construct a map of the environment and for estimation of the current location in this environment. Many SLAM-implementations are developed for different platforms and applications. Often reusability is limited due to platform, algorithm and sensor-specific dependencies and optimisations.
Real-Time Appearance-Based mapping (RTAB-Map) and ORB-SLAM2 are existing open-source SLAM algorithms that can perform appearance-based mapping and localisation under real-time constraints.
This research investigates the factors that affect modularity and reusability for these systems and aims to develop a framework in which SLAM systems can be developed in a way that improves reusability for different projects. Such a modular framework can also be leveraged to use SLAM systems with different sensors and to support mechanisms such as map sharing and distributed SLAM.
To join the presentation via Microsoft Teams click here