Creating a RTAB-Map variation for implementation in a component-based SLAM framework

Finished: 2020-11-30

Individual assignment

SLAM (Simultaneous Localization And Mapping) provides one of the most important functions of autonomous robots, determining its location. RTAB-map (real-time appearance-based mapping) is a SLAM algorithm which is widely adopted and has proved to be a solid solution for mapping in systems which use a stereo- or RGB-D camera as its main source of measurements. The main problem with SLAM is that the algorithms are designed for specific tasks and sensor layouts and minor changes in these layouts results in large modifications of the algorithm. For this reason, the UT is working towards a generally applicable SLAM framework that allows composing SLAM variations from a library. The main goal of this project will be to implement RTAB-map in the SLAM framework that was previously proposed by Jeroen Minnema and described in his thesis ‘Towards component-based, sensor-independent and back-end independent SLAM’ to create a component based RTAB-map variation that can form a basis for future modifications and improvements.

The scope of this project will be the following:

• A study to understand and determine the core components of the RTAB-map SLAM algorithm in order to specify which components have to be present for the algorithm to work correctly and how these components interact.
• The implementation of these components in the existing SLAM framework of Jeroen Minnema
• Validation of the component based algorithm and a comparison with the original RTAB-map algorithm to test performance and robustness of the newly created component based RTAB-map algorithm.