I took a summer internship on Graph-based SLAM at the CIRS research center in Girona last summer, implementing custom features to boost graph optimization.

The features are the ArUco markers that are placed on the cylindric structure. The markers’ poses may be represented in cylindric coordinates and, thus, there was a need to create new custom features which have the advantage of being represented by 2 degrees of freedom in 3D space. The position and radius of the cylinder are also optimized. This introduced an optimization boost as a result of the reduction of the dimensionality of the problem compared to 3D coordinates representation.

Optimized path (blue line), ArUco markers features poses (red spheres), estimated cylindric structure, and features observations (green spheres).

View from the robot’s camera: ArUco markers and a fragment of the cylindric structure.

Girona500 UAV platform.