CS545: Introduction to Robotics

Daniel Seita

Hejia Zhang (Teaching Assistant)

Shihan Zhao (Teaching Assistant)

Use Piazza to contact the course staff: https://piazza.com/usc/spring2024/csci545

When posting any code debugging questions on Piazza, you must declare (in the post) that you have read our class policies here.

Mondays and Wednesdays: 4:00 pm to 5:50 pm. Location: ZHS 352.

We will record lectures and share them on this website (see the "Recording" column below).

Daniel: Mondays, 5:50 pm to 7:00 pm, PHE 212.

Hejia: Thursdays, 3:00 pm to 4:00 pm, RTH 419.

Shihan: Fridays, 10:30 am to 11:30 am, RTH 419.

Exceptions: due to holidays on Jan 15 and Feb 19, Daniel will have office hours on Jan 16 and Feb 20, respectively, both from 4:30pm to 5:30pm.

This class will introduce students to the fundamental questions in robotics: what are good models of the world and how to integrate them reliably into the planning of deployed robotic systems physically interacting with the environment. All these problems arise from the uncertainty due to sensor noise, modeling limitations, approximations in algorithmic computations and inherent unpredictability of action outcomes. The course will explore probabilistic techniques that allow robots to act reliably and exhibit a variety of different behaviors in spite of different sources of uncertainty. We will first cover algorithms for state estimation in both known and unknown environments. We will then explore functional aspects of robot’s interaction with the world, such as the geometry of configuration spaces and manipulation planning in these spaces. We will wrap up the course by exploring the interplay of inference and planning and its applications in robot autonomy, especially for manipulation and mobility.

In this course, you will be introduced to probabilistic techniques that allow state estimation, manipulation and planning in robotics. By the end of this course you should be able to:

• demonstrate proficiency in the tools that support state estimation, manipulation and planning with sensor and modeling noise.
• implement these techniques and test them with real-world datasets.
• integrate your algorithms with state-of-the-art simulation environments.
• explain the computational and practical challenges of applying these techniques in real-world interaction settings and compare them in terms of robustness, scalability and performance.

Students should have a solid background in probability, linear algebra, and calculus. Students should also be comfortable programming in Python.

Much of the lecture material is based on readings from:

• Probabilistic Robotics, by Sebastian Thrun, Wolfram Burgard, and Dieter Fox. (TBF)
• Modern Robotics, by Lynch and Park. You can find the textbook here (use the 2019 version) with lecture videos. (LP)
• Robotic Manipulation: Perception Planning and Control, by Russ Tedrake. You can find the textbook here. (RT)
• Robotics: Modelling, Planning, and Control, by Bruno Siciliano, Lorenzo Sciavicco, Luigi Villani, and Giuseppe Oriolo. (SS)
• Principles of Robot Motion, by Howie Choset, Kevin M. Lynch, Seth Hutchinson, George Kantor, Wolfram Burgard, Lydia E. Kavraki, and Sebastian Thrun. (CL)
• Planning Algorithms, by Steven LaValle. You can find the textbook here. (LA)

Additionally, we will also have materials from these relevant courses:

• Introduction to Artificial Intelligence from Berkeley. You can find a recent course here with slides and videos. (CS188)
• Advanced Robotics from Berkeley (Pieter Abbeel). You can find the Fall 2019 course here with slides and videos. (CS287)
• Introduction to Robotics from USC (Stefan Schaal). You can find the Fall 2019 course here with slides and videos. (CS545)
• Introduction to Robotics from USC (Stefanos Nikolaidis). The most recent offering was in Fall 2023.

The overall course grade is based on:

• Homework assignments: 10%
• Lab assignments: 30%
• Midterm exam: 30%
• Final exam: 30%

We will share grades via Gradescope. We reserve the right to make tweaks to the grading as needed. When submitting assignments on Gradescope please follow the instructions carefully to make sure that the formatting is correct (e.g., if we ask for one page per answer, please do that). To encourage in-class attendance and productive discussions on Piazza, we may offer a small amount of extra credit points to students with excellent course participation or who have written exceptionally helpful or useful Piazza posts. In practice, this may help students who are just below the cutoff for a particular grade. We do not offer alternative times to take the midterm or final exam.

This course is mostly based on materials from Stefanos Nikolaidis' class at USC. We will cite other acknowledgements throughout the slides.

Syllabus

• Unless stated otherwise, all deadlines below are at 11:59PM (in Pacific Time) of the given day.
• See Piazza for links to the homework assignments and labs.