ELC + ABS

This talk will introduce the problem of 3D positioning for a drone especially in GPS-denied environments. The talk will explain why usual odometry or any position algorithm based on standard consumer-level acceleration sensors can't really work. The talk will first focus on external sensors such as Kinect or UWB ulta-wide-band signal, which can be used to provide an accurate 3D position of a drone. The talk will then highlight computer-vision solutions that can be implemented on the drone itself: LKT optical flow, face and body tracking, SLAM algorithms. The limitations and implementation challenges will be highlighted and a brief overview of the hardware and software architecture will be presented. Last, a short demo of a flying nano-drone illustrating those solutions will be offered.