Autonomous Mobility on Demand systems based on self-driving cars will make a huge impact in the world. This class describes the basics of modeling, perception, planning, control and learning for self-driving cars. The focus is on integration and co-design of components and behaviors. The course has a heavy experimental component based on the Duckietown platform.
Lernziel
The students will learn how to design and implement all parts of an architecture for a complex multi-robot system performing nontrivial tasks.
Inhalt
Development tools and best practices for software development of open source projects; single autonomous car functionalities (perception, planning, modeling and control, based on vision data, complemented by learning based approaches); Multi agent behaviors (platooning, coordination, fleet-level policy optimization) focus in group projects.
Skript
Course notes will be provided for free in an electronic form.
Literatur
Course notes will be provided for free in an electronic form. These are some books that can be used to provide background information or consulted as references: (1) Siegwart, Nourbakhsh, Scaramuzza - Introduction to autonomous mobile robots; (2) Norvig, Russell - Artificial Intelligent, a modern approach. (3) Peter Corke - Robotics Vision and Control (4) Oussama Khatib, Bruno Siciliano - Handbook of Robotics
Voraussetzungen / Besonderes
This course is also known as Duckietown. Students should have taken a basic course in probability theory, computer vision, control systems, and should be familiar with basic programming (Python) and Linux use.
