Introduction to Sensor Fusion and Scene Understanding in AD/ADAS

Course Overview

This course introduces the fundamental concepts, architectures, and practical challenges of sensor fusion and scene understanding for autonomous driving and ADAS systems. It is delivered as a lectio magistralis at graduate engineering schools.

Lecture agenda:

1. CONTEXT — ADAS/AD Functional Structure
2. Perception — Sensors and detection pipelines (camera, LiDAR, radar, ultrasonic)
3. Data Fusion — Probabilistic frameworks, Kalman filters, multi-sensor architectures
4. High-level fusion example: Objects — Multi-object tracking and data association
5. High-level fusion example: Road — Road scene representation and fusion
6. Scene Understanding — Semantic perception, environment modelling

Theory — Lecture Slides

Practice — Python Lab

The hands-on lab is a Jupyter notebook guiding students through key sensor fusion algorithms implemented in Python.

Lab exercises include:

• Kalman filter — implementation for linear state estimation and tracking
• Sensor fusion — combining measurements from multiple sensors (camera, LiDAR, radar)
• Path prediction & collision detection — predicting future trajectories and detecting potential collisions

• Download Notebook (.ipynb)