Tutorial on Challenge-driven Learning of Humanoid Robot Control in Virtual Environments

Purpose of this Tutorial

The objective of this tutorial is to provide an opportunity for learning motion planning and control methods of humanoid robots, from basics to recently developed techniques, by making use of simulation technology. This tutorial is linked with Humanoid Virtual Athletics Challenge, a technical challenge in which a robot performs various athletic tasks in virtual environments.

This event consists of the seminar part and the discussion part. In the seminar part, invited researchers working on cutting-edge research on physics simulation and optimization-based motion planning and control will provide lectures covering introductory topics for beginners to practical topics such as software implementation.

In the discussion part, introduction to Humanoid Virtual Athletics Challenge 2022 will be given by the workshop organizer, followed by several talks given by participants in this challenge.

The participants of this tutorial will be able to learn both fundamental knowledge on relevant topics and practical how-tos for developing controllers that enable humanoid robots to perform complex and dynamic tasks in simulation environments. In a round-table discussion, we will discuss future research directions in this field.

Organizers

Yuichi Tazaki, Kobe University
Tomomichi Sugihara, OMRON Corporation
Takumi Kamioka, OMRON Corporation

Date and Place

Fullday, November 28th, 2022. This tutorial will be held in a hybrid style. For online participation, join the Zoom meeting from the following URL. Talks marked "remote" will be presented online. Other talks will be presented physically in a seminar room of the conference venue.
The tutorial has ended with success. Thank you all for your participation!

Time Table

Schedule is subject to change. Time is shown in JST.

09:30-09:50	Opening Remark and Introduction to Humanoid Virtual Athletics Challenge by Yuichi Tazaki
09:50-10:30	Seminar 1: Yuichi Tazaki
10:40-11:20	Seminar 2: Tomomichi Sugihara
11:30-13:00	Lunch Break
13:00-13:40	Seminar 3: Youhei Kakiuchi
13:50-14:50	Talks from Challengers Part 1
13:50-14:10	Talk1: Naoki Hiraoka
14:10-14:30	Talk2: Taiki Ishigaki
14:30-14:50	Talk3: Yuki Onishi
14:50-15:10	Coffee Break
15:10-15:50	Talks from Challengers Part 2
15:10-15:30	Talk4: Masaki Murooka (remote)
15:30-15:50	Talk5: Sotaro Katayama (remote)
15:50-16:30	Seminar 4: Justin Carpentier (remote)
16:40-17:20	Seminar 5: Stephane Caron (remote)
17:20-17:50	Discussion
17:50-18:00	Closing Remark

Talks

Seminar 1: How to make your robot balance and walk in simulation

Speaker: Yuichi Tazaki (Kobe University)

This talk reviews basic building blocks that enable a humanoid robot walk robustly in a simulation environment. Theoretical basics of kinematics, DCM-based balance control, walking pattern generation, and step adaptation are described using actual implementation examples on Choreonoid.
slides

Seminar 2: An approach to build a locomotion system of a humanoid robot

Speaker: Tomomichi Sugihara (OMRON Corp.)

A biped locomotion control system of the speaker's robot is presented. The controller is designed based on the reduced-order dynamics (COM-ZMP) model, with which biped locomotion skill is encoded by a model predictive control scheme. The robot motion is synthesized in real-time by combining it with a motion resolution technique. The observer to estimate unmeasurable quantities such as COM is also presented.

Seminar 3: Robot programming, Simulation and Environment using Choreonoid for Humanoid Novice

Speaker: Yohei Kakiuchi (Toyohashi University of Technology)

What kind of environment would be best for a novice to learn about humanoid robots? I would like to discuss how to build basic system for humanoid robot using Choreonoid, which are considered to help transition from novice to expert as quickly as possible. I will also present how to use Choreonoid as an integrated environment that includes from the design of a robot to the control system for the real robot.
slides

Seminar 4: Revisiting Contact Dynamics within the lens of Proximal Algorithms

Speaker: Justin Carpentier (Inria, Ecole Normale Supérieure)

In this tutorial, I will mostly cover Contact Dynamics revisited under the lens of Proximal Algorithms. While the latter has been introduced more than half a century ago by Jean-Jacques Moreau to study nonsmooth contact dynamics problems (essentially granular mechanics), their use for solving rigid contact dynamic problems involving multi-body systems such as in Robotics has been minimally or even not yet covered in the literature. Through this talk, I will introduce our recent work on Proximal and Sparse Resolution of Constrained Dynamic Equations to efficiently and robustly solve constrained dynamics problems. While this work essentially covers bilateral contact problems, I will sketch the ongoing directions to extend these approaches for unilateral contacts involving friction and Signorini conditions. This is will be also the occasion to discuss the development of the Analytical derivatives of rigid body dynamics algorithms and their extension to constrained problems, and to introduce the versatile Pinocchio library which integrates all these ongoing developments, among other features.
slides

Seminar 5: Humanoid and wheeled-legged controllers in C++ and Python: balancing at different frequencies

Speaker: Stephane Caron (CNRS)

The physics of a task dictate how fast a controller should run to perform it. Balancing, in particular, is a rather low-frequency task: it was demonstrated (in theory and on real hardware) that a DCM-based controller can balance an adult-size humanoid while running at only 10 Hz. In this talk, we will discuss how this property allows us to write more control-critical code in a high-level language like Python. We will illustrate our points with practical sub-modules from two controllers: the LIPM walking controller for HRP humanoids, implemented with mc_rtc, and the Pink controller for the Upkie wheeled-biped, implemented with Vulp (🦊).
slides (tentative)
code
discussion log

Talk 1: Reactive whole-body humanoid system integrating vision-based footstep modification and teleoperation-based object manipulation

Speaker: Naoki Hiraoka (University of Tokyo)

Talk 2: Hybrid-link system integrating rigid- and soft-bodies for humanoid control and human motion analysis

Speaker: Taiki Ishigaki (University of Tokyo)

Talk 3: Choreonoid accelerating humanoid robotics research: Technical overview from the basics to the latest features

Speaker: Yuki Onishi (Tokyo Institute of Technology)

Talk 4: Humanoid controller for walking and jumping motion based on mc_rtc framework

Speaker: Masaki Murooka (CNRS-AIST JRL, IRL)

slides

Talk 5: Whole-body model predictive control of bipedal locomotion in Humanoid Virtual Athletic Challenges

Speaker: Sotaro Katayama (Kyoto University)