The focus of the assignment is the research gap regarding online path planning including multiple coverage of object surfaces to create and improve a reconstructed 3D model of an object of interest. This is of high importance towards the physical interaction with objects, as this requires a detailed representation of the model.
The algorithms are going to be implemented within the ROS framework and the image acquisition is planned to be done by using a stereo camera mounted on a UAV, a fully-actuated hexarotor, within a simulated environment.
Formulated as research questions:
• Which SLAM algorithm would be useful to create 3D visual models on-the-fly?
• Does multiple coverage of an object's surface lead to an improved 3D visual model of it?
• In order to autonomously improve the 3D visual model of an object, is it possible to determine suitable flight directions on-the-fly, which lead also to a reduction in flight time?