Seminar: Humanoid Robotics

Quick Facts

Lecturer:Oleg Arenz
Meetings:Over Zoom
Language:English
Questions: oleg.arenz@tu-darmstadt.de
TUCaN:20-00-1125-se Humanoid Robotics
Presentation of Topics:30.10.2023 11:00 (Slides with topics)
Deadline for Topic Preferences:04.11.2024 23:59
Topic Assignment:06.11.2024 23:59
Deadline for Report:11.03.2025 23:59
Deadline for Poster:15.03.2025 23:59
Poster Session:18.03.2025 10:00
Credits:3,0

Introduction

Robots are on the rise and will increasingly support us during our daily lives. While today's personal robots, such as robotic vacuum cleaners or lawn mowers, rely on a task-specific hardware design, future personal robots will need to be more flexible, so they can assist us with a variety of tasks. In particular, anthropomorphic robots (humanoids), i.e., robots that roughly follow the kinematic structure of humans, are of interest because they will be able to operate in a human environment. However, today's humanoids are not yet viable for personal assistance due to limitations in both hardware (e.g. in terms of robustness, and energy- and cost-efficiency) and software (e.g. in terms of control, reasoning, and the generalizability and transferability of skills). In this seminar, we will discuss the latter problems: You will select one of the many methodological challenges; through a review of the literature, supported by one of our research assistants, you will gain an understanding of the respective state-of-art approaches and their limitations. At the end of the seminar, you will share your acquired knowledge with other students in a peer teaching event (a poster session) and by writing a short article.

From left to right: NASA VALKYRIE, DLR TORO, PAL Robotics TALOS, IIT WALKMAN.
Images taken from 1) NASA, 2) J. Englsberger u. a., „Overview of the torque-controlled humanoid robot TORO“, Humanoids 2014. 3) O. Stasse u. a., „TALOS: A new humanoid research platform targeted for industrial applications“, Humanoids. 2017. 4) N. G. Tsagarakis u. a., „WALK-MAN: A High-Performance Humanoid Platform for Realistic Environments“. Journal of Field Robotics, 34(7). 2017.

Scope

Possible topics will be presented during our kick-off meeting. However, we encourage you to propose a topic related to the theme of the seminar, that you are particularly interested in. Please send suggestions to oleg.arenz@tu-darmstadt.de.

Both more general topics (e.g., review of bipedal locomotion) and more specific topics (e.g., current methods for centroidal trajectory optimization) can be explored. The following is a non-exhaustive list that sketches the scope of this seminar.

  • Bipedal Locomotion
  • Whole-Body Control
  • Sim2Real
  • Manipulation
  • Robot Perception
  • Scene Understanding
  • Trajectory Optimization
  • Human-Humanoid Interactions
  • Learning from Demonstrations
  • Footstep Planning

Requirements

Concurrent or prior enrollment in the course "Foundations of Robotics" and/or "Robot Learning" is recommended.

Format

During the kick-off meeting (30.10.) we will present topics. Shortly after, you can select topics of interest and get a topic assigned to you. You will regularly meet with your assigned teaching assistant to discuss the literature, striving to either gain expert knowledge in a narrow topic or general knowledge in a broader topic. At the end of the term, you will summarize your findings in an article of 6 to 8 pages. Furthermore, a poster session will be organized where you present your insights to fellow students. The grade is based on the report (75%) and poster (25%).

Topic Assignments

Student NameTopicAdvisor
Ardavan RoozkoshHumanoid Ball Catching and JugglingKai Ploeger
Keagan HolmesHierarchical Control for HumanoidsSiwei Ju
Dustin GoreckiBimanual Dexterous ManipulationCristiana Mirandade Farias
Christian BialasReward Shaping for Learning LocomotionSiwei Ju
Kilian FeesCombining Learning-based Methods with Modeling in LocomotionLucas Schulze
Lai Duc BachHuman to Humanoid Motion Imitation and RetargetingNicolas Hahn
Jonas WolfLearning Humanoid Loco-ManipulationNico Bohlinger
Yuddhish ChooahDiffusion Models for Grasping and ManipulationYichen Cai
Noah FarrSafe Exploration in Reinforcement Learning for HumanoidsPuze Liu
Kevin GerhardtImitation Learning for Humanoid Loco ManipulationMichael Drolet
Fabian HahneLearning Bimanual ManipulationAlap Kshirsagar and Vignesh Prasad

Templates

For the report, we recommend the ICRA template.
For the poster, we provide an SVG template (open with Inkscape) as an inspiration.