ROB 498 / 599 · Fall 2026 · Research project

Research project

Pick a 3D robot perception problem, choose the kind of contribution you want to make, and carry it from a proposal to a paper-style report.

Groups 3–4 encouraged, 1–5 allowed Proposal Fri Oct 23 Report + video Fri Dec 18 Submit Canvas

Kinds of project

Any of these is a good project. So is something else entirely.

01Find the gaps in existing methods

Run existing 3D perception algorithms, on their own or in a robotics task, compare them against comparable approaches, and do an error analysis.

02Use a new method in a robot task

Demonstrate a new 3D perception algorithm in a robotic task, in simulation or on a robot platform you personally have access to. Robot platforms are not reserved for the course.

03Propose a new method

Propose a new 3D perception algorithm, or a new 3D perception algorithm for robotics, and test it.

04Collect a dataset

Collect a useful dataset for 3D perception, on its own or in robotics.

05Build useful software

Build a tool for 3D perception in robotics: a visualization tool, for example.

06Your own idea

Bring it to office hours and we will shape it into a plan.

The proposal

A document written in an Overleaf template, in the form of a paper. It should cover:

The contribution. What your project will add. Is it a tool? An analysis? Be clear.
Related research. What have people published in the area, and what are the high-level groupings of research relevant to your problem?
The method.
The evaluation. How will you evaluate the results? What is your experiment plan?
Resources. What do you need to do this project, and do you have it? If you need GPUs, we can get you access to ARC, Michigan’s cluster.

The final report and presentation video

Both are due Fri Dec 18 on Canvas.

The report. A document written in an Overleaf template like a research paper, but now with results and a discussion of future work.
The presentation video. A recorded presentation of the project, in place of an in-class talk: the problem, what you built, and the result, with the method shown working. There is no in-class presentation.

Getting started

Think about the kind of project you would like to do, both the topic and the kind of work.
Form and register a group. Use the Piazza thread if you do not have one yet.
Come to office hours to discuss your project. I can help you brainstorm a concrete plan. If office hours are too short, send a private Piazza message to set up a meeting.
Bring a draft proposal to office hours before you submit it. Feedback before you start the experiments helps make sure they succeed. Again, a private Piazza message will get you a meeting if office hours are full.