Navigation autonome pour robots volants

Aperçu

In recent years, flying robots such as miniature helicopters or quadrotors have received a large gain in popularity. Potential applications range from aerial filming over remote visual inspection of industrial sites to automatic 3D reconstruction of buildings.

Navigating a quadrotor manually requires a skilled pilot and constant concentration. Therefore, there is a strong scientific interest to develop solutions that enable quadrotors to fly autonomously and without constant human supervision.

This is a challenging research problem because the payload of a quadrotor is uttermost constrained and so both the quality of the onboard sensors and the available computing power is strongly limited. In this course, we will introduce the basic concepts for autonomous navigation for quadrotors.

The following topics will be covered:

3D geometry, probabilistic state estimation, visual odometry, SLAM, 3D mapping, linear control. In particular, you will learn how to infer the position of the quadrotor from its sensor readings and how to navigate it along a trajectory.

The course consists of a series of weekly lecture videos that we be interleaved by interactive quizzes and hands-on programming tasks. For the flight experiments, we provide a browser-based quadrotor simulator which requires the students to write small code snippets in Python.

This course is intended for undergraduate and graduate students in computer science, electrical engineering or mechanical engineering. This course has been offered by TUM for the first time in summer term 2014 on EdX with more than 20.000 registered students of which 1400 passed examination.

The MOOC is based on the previous TUM lecture “Visual Navigation for Flying Robots” which received the TUM TeachInf best lecture award in 2012 and 2013. FAQ Do I need to buy a textbook?

No, all required materials will be provided within the courseware. However, if you are interested, we recommend the following additional materials:

This course is based on the TUM lecture Visual Navigation for Flying Robots.

The course website contains lecture videos (from last year), additional exercises and the full syllabus:

http:

//vision.in.tum.de/teaching/ss2013/visnav2013 Probabilistic Robotics. Sebastian Thrun, Wolfram Burgard and Dieter Fox.

MIT Press, 2005. Computer Vision:

Algorithms and Applications.

Richard Szeliski. Springer, 2010.

Do I need to build/own a quadrotor? No, we provide a web-based quadrotor simulator that will allow you to test your solutions in simulation.

However, we took special care that the code you will be writing will be compatible with a real Parrot Ardrone quadrotor. So if you happen to have a Parrot Ardrone quadrotor, we encourage you to try out your solutions for real.

Programme

• **Introduction à la Navigation Autonome**
Aperçu des quadrotors et de leurs applications
Défis de la navigation autonome des quadrotors
• **Fondamentaux de la Géométrie 3D**
Transformations de coordonnées 3D
Modèles de caméras et projections
• **Estimation d'État Probabiliste**
Introduction à la probabilité et au filtrage bayésien
Filtres de Kalman et extensions (par exemple, EKF, UKF)
• **Odométrie Visuelle**
Détection et correspondance de caractéristiques
Estimation de la pose à partir des données visuelles
• **Localisation et Cartographie Simultanées (SLAM)**
Fondamentaux du SLAM
Techniques de SLAM visuel
• **Cartographie 3D**
Capteurs de profondeur et traitement des nuages de points
Génération de cartes 3D à partir de données de capteurs
• **Contrôle Linéaire pour Quadrotors**
Notions de base de la théorie du contrôle
Contrôleurs PID et dynamique des quadrotors
• **Implémentations Pratiques et Expériences de Vol**
Présentation du simulateur de quadrotor
Tâches de programmation pratiques en Python
Transition des solutions de la simulation aux applications réelles
• **Composants d'Apprentissage Interactif**
Vidéos de conférences hebdomadaires
Quiz interactifs pour la compréhension
Tâches de programmation pas à pas pour des connaissances pratiques
• **Ressources du Cours et Matériaux Supplémentaires**
Accès aux vidéos de conférences et exercices
Livres recommandés pour une compréhension approfondie
• **Évaluation et Certification**
Évaluation des performances via des quiz et des devoirs de programmation
Certificats remis aux étudiants qui réussissent le cours
• **FAQ et Informations Complémentaires**
Logistique du cours et attentes
Opportunités supplémentaires d'apprentissage et d'application réelle

---

Enseigné par

Jürgen Sturm, Daniel Cremers and Christian Kerl

---

Matières

Artificial Intelligence