96994 - Smart Vehicular-Systems

Learning outcomes

The course provides advanced knowledge of the principles and technologies underlying autonomous and connected vehicles, with particular focus on communication, mobility, and artificial intelligence challenges. Students will learn how to analyze real-world application scenarios, design system architectures, and use simulation and development tools for autonomous vehicular systems. Laboratory activities will enable students to gain practical experience in modeling and validating solutions for mobility.

Course contents

Autonomy, mobility, and artificial intelligence are among the technological innovations that will most deeply shape society in the coming years.
This course provides a comprehensive survey of connected, mobile, and autonomous vehicular systems, with the goal of understanding their fundamental building blocks and principles of operation.

Students will explore a wide range of scenarios—from Advanced Driver Assistance Systems (ADAS) to fully autonomous driving—showing how current technology is still in an early stage and how no universal solution exists. Topics will cover sensing, perception, decision-making, networking, system design, and Digital Twin–based approaches for modeling and simulation.

Pre-requisites

Students are expected to have:

• A solid understanding of algorithms and basic AI concepts.

• Good programming skills and familiarity with Linux environments and real-time or embedded systems.

• Preferably, experience with AI frameworks and models (e.g., YOLO, TensorRT).

• Knowledge of ROS, CUDA/Jetson programming, and introductory control theory is a plus, though not mandatory.

Course Objectives

The course aims to provide students with a deep foundation for understanding how autonomous and connected vehicular systems are designed, deployed, and validated today—and how they are likely to evolve in the near future. Laboratory sessions based on the CARLA simulator will allow hands-on experience with perception, planning, control, and Digital Twin methodologies.

A successful student will be able to:

• Understand the challenges and opportunities arising from autonomous vehicles, intelligent mobility, and mobile robotics.

• Gain in-depth knowledge of advanced networking and distributed systems concepts applied to connected vehicles, including architectures beyond traditional Internet models.

• Analyze the requirements of an autonomous system and select appropriate sensing, perception, communication, and computation architectures.

• Acquire hands-on familiarity with simulation tools (e.g., CARLA) and experimental platforms for autonomous and connected systems.

Readings/Bibliography

The primary material for this course will be provided directly by the instructor and will consist of research papers, technical reports, and selected excerpts from classic and modern references in autonomous driving, mobile systems, vehicular networking, and AI for mobility.

Students will also be required to complete online tutorials and technical documentation for specific tools and frameworks used in the laboratory activities (e.g., CARLA, ROS2, AI toolkits, sensor processing libraries).

Recommended Reading List

Although not mandatory, the following books provide useful background and complementary perspectives:

• Joseph O’Rourke, Computational Geometry, Cambridge University Press.
  A classic text for understanding geometric algorithms involved in navigation, mapping, and spatial reasoning.

• Kevin M. Lynch & Frank C. Park, Modern Robotics: Mechanics, Planning, and Control, Cambridge University Press.
  A foundational reference for kinematics, motion planning, and control in autonomous systems.

• Andreas Lindholm, Niklas Wahlström, Fredrik Lindsten,
  Machine Learning: A First Course for Engineers and Scientists, Cambridge University Press.
  A practical introduction to machine learning methods relevant to perception, prediction, and decision-making.

• Sean Meyn, Control Systems and Reinforcement Learning, Cambridge University Press.
  A modern and rigorous bridge between classical control theory and reinforcement learning for autonomous behavior.

Teaching methods

The course is structured around two complementary teaching components:

1. Traditional Lectures (Theory)

The theoretical part of the course will be delivered through classic in-person lectures, introducing fundamental concepts in smart vehicular systems, autonomous driving, vehicular communications, perception, and Digital Twin methodologies. Lectures will provide the background needed to understand system architectures, algorithms, and design trade-offs.

2. Laboratory Sessions with the CARLA Simulator (Practice)

The practical component consists of hands-on laboratory sessions based on the CARLA autonomous driving simulator. Students will work individually or in small groups to implement, test, and analyze vehicular functions such as perception, sensor fusion, planning, control, and Digital Twin–based modeling.

Assessment methods

The assessment is based on a team project (up to 3 students per group), which includes a written report, source code, a final presentation, a live demo on the CARLA simulator, and an individual oral examination on topics covered during the theoretical lectures.

Project Submission

Each group will receive a project assignment during the course.

Students must submit the project report and the complete source code at least one week before the presentation date.

The project must already be integrated with the simulator controller used in the laboratory.

Grading Breakdown

The final grade (maximum 32 points) is composed as follows:

1. Project Report + Source Code — 16 points

Evaluation criteria include:

• Correctness and completeness of the implementation

• Quality of the architecture and design choices

• Originality and clarity of the technical report

• Functionality and robustness of the system

• Proper integration with the CARLA simulator

2. Presentation + Live Demo — 8 points

• Completeness of the presentation: 3 points

• Mastery and clarity of exposition: 3 points

• Compliance with time limits: 2 points

3. Individual Oral Examination — 8 points

Each student will be individually assessed through oral questions on topics covered during the theoretical lectures.

Evaluation criteria include:

• Understanding of the main theoretical concepts

• Ability to explain and connect the topics discussed during the course

• Appropriate use of technical terminology

• Ability to apply theoretical concepts to practical scenarios

Presentation Time

• 3-student team: 30 minutes (plus Q&A)

• 2-student team: 20 minutes (plus Q&A)

• Individual project: 10 minutes (plus Q&A)

Slides are required to accompany the presentation.

The live demo must be executed inside CARLA, demonstrating the functional integration of the developed system.

Teaching tools

The course makes use of a set of software tools, simulation environments, and development frameworks that support both the theoretical lectures and the laboratory activities. The main teaching tools include:

Simulation and Development Platforms

• CARLA Simulator for autonomous driving, sensor modeling, and scenario testing

• Python for scripting, data analysis, and prototyping

• Linux environment (Ubuntu recommended) for development and integration

• Visual Studio Code or similar IDEs for coding and debugging

• Git/GitHub for version control and project collaboration

AI, Perception and Robotics Tooling

• PyTorch / TensorFlow for machine learning and deep learning

• OpenCV for computer vision tasks

• ROS2 (optional) for modular robotic system development

• NVIDIA CUDA / Jetson support (optional) for hardware acceleration

Vehicle and Networking Tools

• Wireshark, tcpdump, and networking utilities for understanding vehicular communications

• MQTT clients for interfacing with message-oriented architectures (only if needed by project)

Office hours

See the website of Roberto Girau