Legged robots

MICRO-507

Media

Media

This file is part of the content downloaded from Legged robots.
Course summary

The course presents the design, control, and applications of legged robots. It gives a review of different types of legged robots (including two-, four- and multi-legged robots) and different control approaches for legged locomotion. A few selected control approaches will be analyzed more in-depth. The course also trains students in making critical analysis of key articles in the field, and in designing their own legged robot models and locomotion controllers in simulation.

Grading: Student presentations of articles 20% of grade, projects 50%, and written exam 30%. Work for the presentations and the projects will be done by groups of 3 students.

Presentations of articles: Students will chose one article among a list of key articles in legged robots. The purpose is to carefully read the article, critically analyze its pros and cons, and present it to the class. The presentation will be graded by the professor and teaching assistants (TAs). The list of articles will soon be provided, and a schedule will be organized after 2-3 weeks. Each student in the 3-person team should participate to the presentation.

Projects: The two projects will be done in Python. The first project will involve developing a planner/controller based on the Divergent Component of Motion for a simulated biped (Atlas) in PyBullet. The second project will involve developing controllers for a quadruped robot (go1), first with Central Pattern Generators and then with Deep Reinforcement Learning. For both projects, student teams will need to hand in a single scientific report, their programming code, and movies of their results.

Written exam: The exam will cover the theoretical aspects of the course. It will cover material from the lectures (by Auke Ijspeert and guest lecturers), and the student presentations of articles. It will take place on December 10, 10:15-12:00 in INF2 and INF3.

Assistants: Guillaume.Bellegarda@epfl.ch (Postdoc, main assistant), Astha.Gupta@epfl.ch, Chuanfang.Ning@epfl.ch,   Milad.Shafiee@epfl.ch, Lixuan.Tang@epfl.ch (PhD student teaching assistants) and Serge.Elasmar@epfl.ch, Tim.Lucking@epfl.ch, and Jad.Bhamdouni@epfl.ch (MSc student teaching assistants).

The lectures and practicals will be given in INF2Lectures are on Tuesdays from 10:15 to 12:00, and practicals from 13:15 to 15:00.


September 10, Introduction, lecture 1, and mini-project

This first lecture gives a short introduction about the content and organization of the course (including the miniprojects). It also covers the history of legged robots, mechanical design choices, and pros of cons of legged versus wheeled robots. The lecture is given from 10:15 to 12:00. And is followed by the practical session from 13:15 to 15:00.


September 17, Lecture 2 and mini-project

Overview of gait definitions, pendulum models, stability criteria and different locomotion metrics.


Sept 24, Lecture 3 and mini-project

Overview of different control approaches. Note: Auke Ijspeert is away at a conference. The lecture is therefore the recorded version of last year, no need to come to the lecture room in the morning. Note: the recording mentions a guest lecture by Sangbae Kim, but that will not happen this year (we will only have Prof. Marco Hutter and Dr Guillaume Bellegarda).


October 1, Lecture 4 and mini-project

Overview of bio-inspired control approaches in the Biorobotics lab.

October 8, Lecture 5 by Dr Guillaume Bellegarda and mini-project

Guest lecture by Dr Guillaume Bellegarda.


October 15, Lecture 6 (recorded lecture by Prof. Sangbae Kim) and mini-project.

Auke Ijspeert is at a conference this week. Instead of the lecture in the morning (no need to come from 10:15-12:00), we provide here the recorded lecture of Prof. Sangbae Kim of last year. Sangbae Kim is famous for his work on the MIT Cheetah robots. That lecture provides interesting insights into the concept of physical intelligence, a type of intelligence that relies strongly on the embodiment and the physical interactions with the environment.

The practicals will be like normal from 13:15 to 15:00 (Guillaume and Milad will be absent, but the other TAs will be there).

Possible exam questions:
  • What are the challenges of physical intelligence (as opposed to "traditional" Artificial intelligence that is more focused on data than physical interactions)?
  • what could be good components for achieving physical intelligence, both in terms of hardware and software?
  • Is our language adapted for describing physical intelligence? if not, what is missing?



October 29, Lecture 7 by Prof. Marco Hutter, and mini-project

Prof. Marco Hutter will give this lecture in person in INF2 at the usual lecture time (10:15). Do not miss this opportunity to hear one of the leading researchers in legged robots!!

Note that the possible exam questions are on the p 93 of the pdf of the slides (not the last page).


November 5, Student presentations and mini-project

Iscen et al. - 2018 - Policies Modulating Trajectory GeneratorsGroup 22
Peng_et_al_RSS2020_Learning_Agile_Robotic_LocomotionGroup 15
Siekmann_ICRA_2021_Sim-to-Real Learning_Bipedal Gaits.pdfGroup 06
Siekmann et al_RSS_2021_Blind Bipedal Stair Traversal_Learning.pdfGroup 02
Radosavovic et al 2024_ScienceRobotics_Realworld_humanoid_locomotion_RL.pdfGroup 28
Cheng et al 2024_RSS_Learning_Whole-Body_Control_Humanoid_Robots.pdfGroup 12
Steingrube et al. - 2010 - Self-organized adaptation of a simple neural circu.pdfGroup 16



November 12, Student presentations and mini-project

Margolis et al. - 2022 - Learning to Jump from PixelsGroup 19
Agarwal et al. 2022 Legged_Locomotion_Egocentric_VisionGroup 10
Zhuang et al. 2023 Robot_Parkour_Learning.pdfGroup 05
Cheng_et_al_ICRA2024_Extreme_Parkour_Legged_RobotsGroup 18
Hoeller et al 2024-ScienceRobotics_Learning_ANYmal_ParkourGroup 27
Vollenweider et al. 2023 Advanced_Skills_through_Multiple_AMPGroup 20
Lee_et_al_SciRob2024_Learning_Navigation_Locomotion_Wheeled_Legged_Robots.pdfGroup 08


November 19, Student presentations and mini-project

Kajita_et_al_ICRA2003_ZMP.pdfGroup 07
Collins_et_al_2005_Dynamic_walkers.pdfGroup 25
Hof_2005_JOB_XCoM_Capture_point_balance.pdfGroup 26
Geng_et_al_IJRR_2006_Runbot.pdfGroup 13
Sentis_Khatib_ICRA2006_Whole_body_control.pdfGroup 21
Bhounsule_et_al_IJRR_2014_Ranger_reflex_based_control.pdfGroup 17


November 26, Student presentations and mini-project

Hyun_et_al_IJRR_2014_MIT_Cheetah.pdfGroup 24
Kenneally_et_al_2016_RAL_DirectDrive.pdfGroup 32
Grimminger et al. - 2020 - An Open Torque-Controlled Modular RobotGroup 33
Coros_et_al_2010_Generalized_walking_control.pdfGroup 09
Gehring_et_al_ICRA2013_Control_dynamic_gaits_quadrupedsGroup 23
Winkler_et_al_RAL2018_Gait_and_TrajOpt_for_Legged_Systems_TOWRGroup 30
Kim_et_al_2019_MiniCheetah_Impulse_MPC.pdfGroup 03


December 3, Student presentations and Mini-project

Lee_et_al_2020_Learning_Quadrupedal_Locomotion_Challenging.pdfGroup 31
Owaki_et_al_Interface_2012_CPG_Phys_Communication.pdfGroup 11
Van der Noot et al. - 2018 - Bio-inspired controller achieving forward speed moGroup 04
Suzuki_et_al_2019_Decentralized_control_salamanderGroup 14
Yasui_et_al_2019_centipede.pdfGroup 29
Badri-Spröwitz et al. - 2022 - BirdBot achieves energy-efficient gait with minimaGroup 01


December 10, Exam (INF2 + INF3) and Mini-project

The exam will take place from 10:15 to 12:00 in INF2 and INF 3.   Normal lab session in the afternoon.


December 17, Mini-project and competition

Today we will start at 10:15 with an assessment of the course evaluation followed by the Legged robot competition.The rest of the time is for the practicals.