The course aims at introducing, with a systemic approach, students to the problems of autonomous and connected driving in order to present the technological challenges that this revolution is posing. In particular, following an overview of the evolution of ADAS, the course will be divided into three parts: the sensors necessary for autonomous driving, the control approaches with the open challenges, the interaction of the vehicle with the surrounding environment. Besides the lessons, students will have the opportunity to practically verify the main aspects of the course. There will be interventions from both academic and industrial experts in the sector and, where possible, guided tours.

Knowledge and understanding

At the end of the course, the student knows:

- the core features of an autonomous and connected vehicle
- which sensors are required and how to process data from these sensors
- which algorithms are used to drive an autonomous and connected vehicle
- how the autonomous and connected vehicle interacts with the environment and in particular with other vehicles (V2V) and with the infrastructure (V2I)

Ability to apply knowledge and understanding

At the end of the course, the student is able to:

- post-process sensor data through state of the art algorithms (such as CNN) to extract relevant information about the environment surrounding the autonomous and connected vehicle
- implement state of the art algorithms for driving autonomous and connected vehicles in a complex environment
- communicate relevant information with other vehicles/the infrastructure for implementing highly complex ADAS

The course describes the potentialities of autonomous and connected vehicles and provides the student with all the necessary tools to end up with a working autonomous and connected vehicle. In particular, the sensory part, the control part and the interaction part will be assessed giving students ready to use algorithms and asking them to customize/improve these algorithms for specific purposes.

In detail, the topics dealt with within the course are:

Introduction to autonomous and connected vehicles

Robotic sensors & introduction to computer vision
Camera models & camera calibration
Stereo vision
Image processing, feature detection & description
Information extraction & classic visual recognition
Modern computer vision techniques

Trajectory optimization
Trajectory tracking & closed loop control
Motion planning
State machines
Decision making under uncertainty
Reinforcement learning

Introduction to localization & filtering theory
Parametric filtering (KF, EKF, UKF)
Non-parametric filtering (PF)
Monte Carlo localization and particle filter SLAM
Multi-sensor perception & sensor fusion

Suggested references

R. Siegwart, I. R. Nourbakhsh, D. Scaramuzza. Introduction to Autonomous Mobile Robots
S. LaValle. Planning Algorithms
D. A. Forsyth, J. Ponce. Computer Vision: A Modern Approach
S. Thrun, W. Burgard, D. Fox. Probabilistic robotics
F. Gustafsson. Statistical Sensor Fusion
D. Simon. Optimal State Estimation: Kalman, Hinf, and Nonlinear Approaches
D. Bertsekas, Reinforcement Learning and Optimal Control

• SDG9 - INDUSTRY, INNOVATION AND INFRASTRUCTURE
• SDG11 - SUSTAINABLE CITIES AND COMMUNITIES

A thorough knowledge of the modelling of mechanical systems as well as of classical control theory is required.

The exam is based on an oral discussion about the topics dealt with within the course with the aim of verifying both the knowledge achievements and the related application abilities.