Jacques K.

En tant que membre de l'équipe de neurorobotique du Human Brain Project, j'ai participé au développement de la plateforme de neurorobotique : un outil basé sur le web pour concevoir et simuler des expériences impliquant des cerveaux artificiels reliés à des robots. J'ai ensuite utilisé cette plateforme pour mes recherches de doctorat sur l'apprentissage sensorimoteur.

Mes activités principales à FZI étaient:

- développement de robots autonomes avec ROS en python et C++ sous Linux

- formateur ROS en entreprise et enseignant à l'université de Karlsruhe (robotique et machine learning)

- chef de projet technique pour de petites équipes (3-4 personnes) - supervision d'étudiants de master d'informatique et électrotechnique

This internship is in the framwork of my second year of Master of Science in Informatics at Grenoble (MoSIG). It is a renown, international research master specialized in Graphics, Vision and Robotics. During this intern I was a member of the team CHROMA (previously known as e-Motion) specialized in autonomous robots. Under the supervisation of Agostino Martinelli, I worked on a novel Closed-Form Solution to fuse intertial and visual measurements. We see this method being particularly suited to drones since a popular choice of on board sensors are a camera along with an inertial measurement unit (IMU). It can be used as an initialization to more popular filter-based algorithms, such as the Extended Kalman Filter (EKF), which require to be provided with an initial state. I implemented the Closed-Form solution in PTAM, a ROS package that deals with pose estimation using visual information. Before, the initialization of PTAM was done through 8-point algorithm over a pair of stereo images. However, it is not possible to recover the absolute roll and pitch angles and the absolute scale from visual measurements alone. Now, combining visual and inertial measurements, we can initialize PTAM with the Closed-Form solution which provides these values.