| Category | Difficulty |
|---|---|
| HW | 3 |
| Project | 3 |
Self Driving Cars: Perception & Control is a new class introduced to Robotics Department in Spring 2023. This class combines both the perception and control knowledge in the field of self driving car and gives students homework and project that solves problems using real work and simulation data samples. The course is split into two halves: the first half semester covers perception, and the second half covers control.
Perception
- Pose sensors
- Camera & Lidar geometry
- Localization & SLAM
- Machine Learning for Computer Vision
Control
- State space models
- Linear Quadratic Regulators (LQR)
- Vehicle Dynamics
- Model Predictive Controls (MPC)
- Trajectory Optimization
The class only has one lecture per week, and it is 2 hour and 50 minute long, from 4 pm to 6:50 pm. During the lecture period there will have two intermissions, so it won't be too exhausting. During the lecture, professors will cover how certain perception/control techniques were developed and how to deploy them in self driving car scenarios. The lectures assume students have some prior knowledge in linear algebra, so it is good to take some linear algebra courses before taking this class. However, you are welcome to ask questions or concepts that you have not heard of and both perception and control professors will happily answer you.
Homework is due biweekly. The difficulty is not too hard, so you have plenty of time to work on the assignment. Note that there is only one TA per half of the semester (so one perception TA and one control TA), you should work on the assignment ahead of time and ask questions during the limited office hour time. There is one assignment you are allowed to submit up to 48 hours late. Since most materials are related to linear algebra, the assignment will be submitted using MATLAB grader and you can test your submissions for unlimited number of times before the due. It's better if you have worked with MATLAB before this class. However, you can still take it without the prior knowledge as there will be a warmup assignment on MATLAB before everything.
The final project is about perceiving other vehicles and controling your own vehicle's movement in the simulation environment of GTA5 (Wow!). This final project is pretty cool and you are allowed to use any perception and control technique. The baseline for passing the final project is fair, and you can explore more techniques after getting a passing score. The final project is in groups.
There are no exams in this class.
- Ask questions about anything during the class time
- Start your homework early
- Ask your classmates or on Piazza since TAs have few office hour slots
- Have fun :)