Autonomous mobile robots are getting more and more widespread around the world and they utilize more and better technologies, from robots walking on feet to self driving cars. This document presents the design of the head of the not so common - spherical - robot, it also presents a possible solution for the robot's navigation, the RGBD visual SLAM.
The document gives a brief summary of the mathematical foundation and algorithms needed for the solution of the SLAM problem, for example, it goes trough the different ways of describing affine transformations, the usecases of homogenous coordinate systems in image processing, the way quaternions can describe rotations in 3D space, the feature detection algorithms SIFT and ORB and the model estimating algorithm called RANSAC. It introduces the different libraries used in the implementation. After the document goes trough the implemented algorithm, it tests it's behaviur, and demonstrates the usefulness of the loopback in the SLAM algorithm.
At the planning of the headpiece it shows why certain peripheries were needed, and the reasons for the hardware choices we made. The document also peeks into the parametric design implemented in CAD software.