Workshop on Humanoid Soccer Robots 2024

Humanoids 2024 IEEE-RAS International Conference on Humanoid Robots
November 22, 2024 - Nancy, France

Presentation

Humanoid robotics has seen significant advancements in recent years, with notable improvements in both hardware and software capabilities. Off-the-shelf actuators, components, and robotic platforms like Unitree H1, Nao, NimBro, and Robotis-OP3 are becoming increasingly accessible to researchers. Despite these advancements, many challenges remain in achieving robust locomotion, localization, and navigation in complex environments. The dynamic and adversarial nature of soccer makes it an ideal testbed for these challenges.

This workshop aims to gather the humanoid robotics research community, particularly those involved in the RoboCup competition, which aspires to field a robotic team capable of competing with human soccer players by 2050. The RoboCup challenge has fostered a dynamic, multidisciplinary community.

Topics

Humanoid robots prototyping
Motion planning and control
Perception
Localization
Bipedal locomotion
3D simulation of humanoid robots
Team coordination
Skill and strategy learning
Dynamic motions
Heterogeneous humanoid robots
Autonomous decisions
Machine learning for robotics

Call for contributions

We encourage researchers in humanoid robotics to present ongoing work, recent results, and future directions. We invite to submit a 2-page extended abstract in PDF format, following the IEEE conference template. The 2-page limit don’t include references.

Make your submission on EasyChair

Key dates

Program

Time Event
08:30 Registration opens
Session 1: Robot modelling & control
09:30 Organizer talk From Zero to Robot: Rhoban Pipeline for Designing and Prototyping Legged Robots
Grégoire Passault
Slides
09:50 Maximum Impulse Approach to Soccer Kicking for Humanoid Robots
Grzegorz Ficht
Extended abstract
10:10 Group discussion
10:30 Coffee break
Session 2: Machine learning for humanoid locomotion (1/2)
11:00 Invited talk LeRobot: Lowering the entry barrier to AI for robotics
Rémi Cadene
11:45 Learning Footstep Planning using Deep Reinforcement Learning for Dynamic Walking
Jasper Güldenstein
Extended abstract
12:05 Introducing FootstepNet: an Efficient Actor-Critic Method for Fast On-line Bipedal Footstep Planning and Forecasting
Clément Gaspard
Extended abstract - Repository
12:30 Lunch break
Session 3: Machine learning for humanoid locomotion (2/2)
13:30 Invited talk Tips for Learning Locomotion Directly on Real Hardware
Antonin Raffin
Slides
14:15 Introducing FRASA: An End-to-End Reinforcement Learning Agent for Fall Recovery and Stand Up of Humanoid Robots
Marc Duclusaud
Extended abstract - Repository
14:35 Towards the ultimate goal of RoboCup with a standard humanoid robot platform
Mingguo Zhao
Extended abstract
15:00 Coffee break
Session 4: Localization, perception and multi-robots coordination
15:30 Invited talk Generalization in Robotic Vision: From Efficient Algorithms to Foundation Models.
Jan-Nico Zaech
16:00 Anticipatory Approach to Combining Local and Global Perception for Stable Decision-Making
Heinrich Mellmann
Extended abstract
16:20 Kinematic State Estimation for Humanoid Robots in the Robocup Competition
Thomas O’Brien
Extended abstract - Slides - Repository
16:40 A Sampling-Based Algorithm for Multi-Agent Coordination in RoboCup Soccer
Filippo Spinelli
Extended abstract - Repository
17:00 Organizer talk NimbRo Humanoid Robots Winning RoboCup AdultSize Soccer Competitions: Mechatronics, Perception, Control, and Learning
Sven Behnke
Slides
17:45 End of the workshop

Invited speakers

Antonin Raffin (homepage)
German Aerospace Center (DLR) Presentation: Tips for Learning Locomotion Directly on Real Hardware

Antonin Raffin, a Research Engineer at the German Aerospace Center (DLR), specializes in reinforcement learning (RL). He is the principal developer of Stable-Baselines3, an open-source library implementing deep RL algorithms. Raffin’s work focuses on improving RL reproducibility and its applications in robotics. In particular, he aims at learning directly on the real robots

Rémi Cadene (homepage)
Hugging Face

Presentation: LeRobot: Lowering the entry barrier to AI for robotics

Rémi Cadene, a former Tesla researcher, is currently leading the development of LeRobot at Hugging Face. This open-source robotics project aims to integrate AI with physical systems, leveraging Cadene’s expertise in autonomous technologies. He is actively recruiting engineers to advance the initiative in Paris, focusing on creating accessible robotic solutions that utilize deep learning and embodied AI principles.

Location and venue

The Workshop takes place at the Nancy Congress Center (Centre Prouvé, 1 Place de la République, 54000 Nancy, France)

See the Humanoids 2024 website for more information about accomodation

Organizers