Course Number: 20-00-1170-pr
General Information: This course gives you a great opportunity of testing the waters of research. Through the semester you will see yourself re-implementing a research paper or extending it to newer domains.
Below is the list of this year’s course project topics. Each project stems from a recent research paper tackling applications of AI for robot manipulation. Particularly, each project will be associated to either one individual student or a group of two students that will be asked to re-implement the indicated paper either in simulation or on a real robot in the lab. The format of the list below is as follows
[<Nature of implementation>] <Paper Name> (<website link>) [<Advisor Name>]
- [Simulation] Active Exploration for System Identification in Robotic Manipulation (website) [Gabriele Tiboni]
- [Simulation] LLM-enhanced Scene Graph Learning for Household Rearrangement (website) [Sohan Rudra]
- [Real World] Human-Robot Cooperative Object Manipulation with Contact Changes (website) [Sophie Lueth]
- [Real World] Diffusion Co-Policy for Synergistic Human-Robot Collaborative Tasks (website) [Vignesh Prasad]
- [Real World] MOSAIC: A Modular System for Assistive and Interactive Cooking (website) [Vignesh Prasad]
- [Simulation] TokenHSI: Unified Synthesis of Physical Human-Scene Interactions through Task Tokenization (website) [Snehal Jauhri]
- [Real World] MimicPlay: Long-Horizon Imitation Learning by Watching Human Play (website) [Franziska Herbert]
- [Simulation] Applying Sampling-based MPC to generate Robust Object Placements (website) [Jiayun Li]
- [Real World] Deep Visual Constraints: Neural Implicit Models for Manipulation Planning from Visual Input (website) [Yuezhe Zhang]
- [Simulation] SAPG: Split and Aggregate Policy Gradients (website) [Zechu Li]
- [Simulation] Neural Visibility Field for Uncertainty-Driven Active Mapping (website) [Yufeng Jin]
Moodle page:https://moodle.tu-darmstadt.de/mod/page/view.php?id=1560273
Course Format:
Each project stems from a recent research paper tackling applications of AI for robot manipulation. Particularly, each project will be associated to either one individual student or a group of two students that will be asked to re-implement the indicated paper either in simulation or on a real robot in the lab
- To enrol in this course you first need to apply for a specific project. For this, please email the listed supervisor directly with your cover letter (optionally add your CV and transcript of records). Projects are allocated on a rolling basis until the application deadline (May 10th, 2025), so early contact is encouraged.
Requirements:
- Basic knowledge of robotics, machine learning and deep learning techniques
- Programming skills in Python
- Previous registration for the Project Lab 1 is helpful, but not mandatory