Mobile Robotics

Lecture (2h) and Exercise (2h) (5 ECTS)

Mobile robots have been sent to Mars, can vacuum clean our homes, mow the lawn, and are promised to carry us around in the form of self-driving cars or even flying taxis. In this module, you will learn about the different component of such mobile robots and their interactions: from perception to estimation, mapping, and control. As will also learn to work with different mathematical representations of robot states and maps. Since deployment in open-ended environments requires sophisticated perception, localisation, and mapping approaches, we will dedicate a substantial part of the course towards probabilistic multi-sensor-fusion and modern Simultaneous Localisation and Mapping (SLAM) and more general Spatial AI systems – including elements of Machine Learning. In the last part, you will then be learning about how to use these representations of robot state and surroundings for navigation and control.

I Foundations:

• Representations of States, the Environment, and Uncertainty
• Overview of Sensors and Actuators

II Multi-Sensor Estimation:

• Probabilistic Nonlinear Least Squares
• Kinematics and Temporal Models
• Recursive Estimation

III From SLAM to Spatial AI:

• Sparse Visual and Visual-Inertial SLAM
• Dense Mapping and SLAM
• Semantic, Object-level and Dynamic Mapping

IV Planning and Control:

• Feedback Control, Model-Based, and Model-Predictive Control
• Motion Planning and Exploration
• Reinforcement Learning: Relation to Models, Control, and Perception

Passion for mobile robots and drones, as well as solid mathematical foundations regarding analysis, linear algebra, and probability theory. In parallel to this course, “Mobile Robotics Practicals” is offered, the participation in which is highly recommended. Additionally, participation in the following courses is highly recommended:

• Computer Vision II: Multi-View Geometry (IN2228),
• Robotics (IN2067),
• Robot Motion Planning (IN2138),
• Introduction to Deep Learning (IN2346)
• Motion planning for autonomous vehicles (IN2106, IN0012, IN4221)

The exercises will require solid knowledge of Python.

• Lecture slides
• Probabilistic Robotics by Thrun, Burgard, Fox, MIT Press 2006

• Lecturer: Prof. Dr. Stefan Leutenegger
• Teaching Assistants: Dr. Xingxing Zuo, Simon Boche, Simon Schaefer