Robotic Manipulation Seminar

Overview

The Robotic Manipulation Seminar is a weekly graduate reading group at the University of Illinois Urbana-Champaign, focused on recent advances in robotic manipulation. Each meeting will feature student presentations of recent research papers, followed by group discussion.

Students may participate in either of two ways:

• Register for the 1-credit course, in which case the presentation and attendance requirements below apply; or
• Attend the reading group informally without registering for credit.

Everyone interested in robotic manipulation is welcome to attend and participate in discussions. Students who are not registered for credit are also welcome to volunteer to present, subject to instructor approval. In general, PhD students working in robotics will be approved to present.

At each meeting, we plan to have two paper presentations. Each presentation consists of:

• 20 minutes for the presentation
• 10 minutes for questions and discussion

Presenters should focus on helping the group understand the paper’s main ideas, technical contributions, experimental results, limitations, and connections to the broader robotic manipulation literature. Students may not present their own papers.

Quick Links & Signup

Requirements for Course Credit

To receive credit for the 1-credit course, a student must:

1. Complete two paper presentations during the semester, or complete an instructor-approved reading report in place of the second presentation if there are not enough presentation slots; and
2. Attend the reading group regularly, with no more than three absences during the semester. No excuse or documentation is required for these three absences.

Active participation in paper discussions is strongly encouraged.

If enrollment or scheduling makes it impossible for every registered student to give a second presentation, students who do not have an opportunity to present their second paper will instead complete a reading report.

Guidelines for Paper Presentations

Each presentation is 20 minutes + 10 minutes for Q&A and discussion. Please plan carefully and do not go over time.

Choose a Good Paper

Choose your paper at least one week before your presentation and update the paper title in the presentation schedule sheet. Instructors may request that you change papers to present if the paper’s quality does not meet the reading group’s requirements or if it significantly overlaps with papers already discussed/scheduled.

Selecting a good paper could be about as much work as preparing to present it. See below.

The reading group is a collective way to keep up with the field. Hundreds of papers appear on these topics every month and we only have presentation slots for a few handful. So, please choose a good paper to present. A good paper can take many different forms. A non-exhaustive list of what a good paper could be doing:

• presenting a new way to think about an important existing problem.
• defining a new and important problem.
• presenting experiments that settle an important question in the field.
• presenting important new resources that are useful to tackle important problems in the field.

It takes time to find good papers. We assume that active researchers are scanning through several papers a week. Most may not be worthy of everyone’s time, but every once in a while you come across something that teaches you something new. Those are good candidates for papers to present. Note that not all papers that are going viral on Twitter may be good papers to present. Tech reports from high-profile companies may also not always be good candidates, particularly the ones with very few details and just a dazzling demo.

To allow for candid discussion, we also request that you not present your own papers.

Before Your Presentation

Please read the paper thoroughly and critically. You should be prepared to answer questions about details that may not be covered in your presentation. You should also make sure you understand the background and related work that the paper builds upon. Consider preparing a few backup slides for Q&A.

What to Cover

Twenty minutes is limited, so focus on the paper’s key ideas and takeaways:

• Problem: What problem or challenge does the paper address?
• Method: Give an overview of the approach, emphasizing what is new or different from previous work.
• Results: Show the main results and conclusions. Focus on the most important experiments, examples, and videos rather than covering every detail.
• Your assessment: What are the highlights, limitations, and open questions?

Presentation Style

Assume the audience has the background knowledge listed on the website, so you don’t need to introduce standard concepts. However, briefly explain any advanced or less common methods that are important for understanding the paper.

Use figures, diagrams, and videos whenever possible, and avoid slides with too much text. Paper project websites are often a good source of additional videos and figures.

Before the Meeting

Please test your computer connection, slides, and videos before the meeting starts.

Know the paper in depth, but focus the presentation on what the audience needs to understand its key contributions and results.

Background Expectations

This reading group is intended to focus on recent research rather than introductory material. Participants are expected to have general familiarity with topics such as:

• Robot kinematics and control
• Robot vision
• Machine learning
• Reinforcement learning for robotics
• Imitation learning and commonly used methods
• Vision-language models (VLMs)
• Vision-language-action models (VLAs)
• World/action models and related approaches

Presenters should not spend substantial presentation time introducing basic concepts in these areas. Instead, presentations should focus on understanding and critically discussing the selected research paper.