EE259: Principles of Sensing for Autonomy

Reza Nasiri Mahalati, Stanford University, Spring Quarter 2022–23

Announcements

This is the website for EE259, Spring quarter 2022–23.
Course materials can be accessed via the class website.

About EE259

EE259 is a new class, taught for the first time in Spring quarter 2022–23. Sensing is at the core of any autonomous robotics application, and EE259 covers the fundamental principles of design and operation of sensors that enable autonomy in robotic systems such as autonomous vehicles. We start from the basic physics principles such as Newton's laws of motion, or Maxwell's equations of electromagnetism to understand the underlying operating principles of different sensors. We then use this knowledge the analyze the mechanical and electrical architecture of different sensors. Finally, we develop the algorithms required for processing the raw signals from different sensors to produce outputs required for robot perception and planning, such as ego trajectory and 3D local maps. Along the way we also discuss the performance of sensors under different operating conditions, and study practical system design tradeoffs.

In short, EE259 covers how acoustic waves, photons and mechanical forces are translated to position, odometry, trajectories, pointclouds and images by GPS, Inertial sensors, Ultrasonic sensors, Radars, Lidars and Cameras.

The course is suitable for any graduate or advanced undergraduate students with the prerequisites or equivalent background. Target audience are researchers and engineers interested in working on any part of autonomous systems, from sensor hardware design to AI/ML models for robotic perception, planning or control. Anyone up for it is welcome.

EE259 was developed by Reza Nasiri Mahalati in 2022.

The early days of autonomy

The 2005 DARPA Grand Challenge was a driverless car competition that is considered a turning point in autonomy by many experts in the field. The Stanford Racing Team won the challenge with their car Stanley which was later donated to Smithsonian's National Museum of American History in Washington, DC. While Stanley was designed almost two decades ago, the sensing architecture does not look that different from driverless vehicles on the roads today, and unconditional autonomy remains largely unsolved. Perhaps it's time for a new generation of robotics and machine learning engineers to tackle autonomy from a fresh perspective.