Learning How to Walk: Warm-starting Optimal Control Solver with Memory of Motion

Teguh Santoso Lembono; Carlos Mastalli; Pierre Fernbach; Nicolas  Mansard; Sylvain Calinon

doi:10.1109/ICRA40945.2020.9196727

Learning How to Walk: Warm-starting Optimal Control Solver with Memory of Motion

Teguh Santoso Lembono, Carlos Mastalli, Pierre Fernbach, Nicolas Mansard, Sylvain Calinon

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract

In this paper, we propose a framework to build a memory of motion for warm-starting an optimal control solver for the locomotion task of a humanoid robot. We use HPP Loco3D, a versatile locomotion planner, to generate offline a set of dynamically consistent whole-body trajectory to be stored as the memory of motion. The learning problem is formulated as a regression problem to predict a single-step motion given the desired contact locations, which is used as a building block for producing multi-step motions. The predicted motion is then used as a warm-start for the fast optimal control solver Crocoddyl. We have shown that the approach manages to reduce the required number of iterations to reach the convergence from ~9.5 to only ~3.0 iterations for the singlestep motion and from ~6.2 to ~4.5 iterations for the multi-step motion, while maintaining the solution’s quality.

Original language	English
Title of host publication	2020 IEEE International Conference on Robotics and Automation (ICRA)
Publisher	Institute of Electrical and Electronics Engineers
Pages	1357-1363
Number of pages	7
ISBN (Electronic)	978-1-7281-7395-5
ISBN (Print)	978-1-7281-7396-2
DOIs	https://doi.org/10.1109/ICRA40945.2020.9196727
Publication status	Published - 15 Sept 2020
Event	2020 International Conference on Robotics and Automation - Virtual conference, France Duration: 31 May 2020 → 31 Aug 2020 https://www.icra2020.org/

Publication series

Name
Publisher	IEEE
ISSN (Print)	1050-4729
ISSN (Electronic)	2577-087X

Conference

Conference	2020 International Conference on Robotics and Automation
Abbreviated title	ICRA 2020
Country/Territory	France
City	Virtual conference
Period	31/05/20 → 31/08/20
Internet address	https://www.icra2020.org/

Access to Document

10.1109/ICRA40945.2020.9196727Licence: All Rights Reserved

Learning How to Walk_LEMBONO_DOA31012020_AFVAccepted author manuscript, 2.34 MBLicence: All Rights Reserved

https://ieeexplore.ieee.org/document/9196727Licence: All Rights Reserved

MEMMO - Memory of Motion
Vijayakumar, S. (Principal Investigator)
EU government bodies
1/01/18 → 30/06/22
Project: Research

Cite this

@inproceedings{6d17682ff6104d769aab793bb0a2a6ce,

title = "Learning How to Walk: Warm-starting Optimal Control Solver with Memory of Motion",

abstract = "In this paper, we propose a framework to build a memory of motion for warm-starting an optimal control solver for the locomotion task of a humanoid robot. We use HPP Loco3D, a versatile locomotion planner, to generate offline a set of dynamically consistent whole-body trajectory to be stored as the memory of motion. The learning problem is formulated as a regression problem to predict a single-step motion given the desired contact locations, which is used as a building block for producing multi-step motions. The predicted motion is then used as a warm-start for the fast optimal control solver Crocoddyl. We have shown that the approach manages to reduce the required number of iterations to reach the convergence from \textasciitilde{}9.5 to only \textasciitilde{}3.0 iterations for the singlestep motion and from \textasciitilde{}6.2 to \textasciitilde{}4.5 iterations for the multi-step motion, while maintaining the solution{\textquoteright}s quality.",

author = "Lembono, \{Teguh Santoso\} and Carlos Mastalli and Pierre Fernbach and Nicolas Mansard and Sylvain Calinon",

year = "2020",

month = sep,

day = "15",

doi = "10.1109/ICRA40945.2020.9196727",

language = "English",

isbn = "978-1-7281-7396-2",

publisher = "Institute of Electrical and Electronics Engineers",

pages = "1357--1363",

booktitle = "2020 IEEE International Conference on Robotics and Automation (ICRA)",

address = "United States",

note = "2020 International Conference on Robotics and Automation, ICRA 2020 ; Conference date: 31-05-2020 Through 31-08-2020",

url = "https://www.icra2020.org/",

}

Lembono, TS, Mastalli, C, Fernbach, P, Mansard, N & Calinon, S 2020, Learning How to Walk: Warm-starting Optimal Control Solver with Memory of Motion. in 2020 IEEE International Conference on Robotics and Automation (ICRA). Institute of Electrical and Electronics Engineers, pp. 1357-1363, 2020 International Conference on Robotics and Automation, Virtual conference, France, 31/05/20. https://doi.org/10.1109/ICRA40945.2020.9196727

Learning How to Walk: Warm-starting Optimal Control Solver with Memory of Motion. / Lembono, Teguh Santoso; Mastalli, Carlos; Fernbach, Pierre et al.
2020 IEEE International Conference on Robotics and Automation (ICRA). Institute of Electrical and Electronics Engineers, 2020. p. 1357-1363.

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

TY - GEN

T1 - Learning How to Walk: Warm-starting Optimal Control Solver with Memory of Motion

AU - Lembono, Teguh Santoso

AU - Mastalli, Carlos

AU - Fernbach, Pierre

AU - Mansard, Nicolas

AU - Calinon, Sylvain

PY - 2020/9/15

Y1 - 2020/9/15

N2 - In this paper, we propose a framework to build a memory of motion for warm-starting an optimal control solver for the locomotion task of a humanoid robot. We use HPP Loco3D, a versatile locomotion planner, to generate offline a set of dynamically consistent whole-body trajectory to be stored as the memory of motion. The learning problem is formulated as a regression problem to predict a single-step motion given the desired contact locations, which is used as a building block for producing multi-step motions. The predicted motion is then used as a warm-start for the fast optimal control solver Crocoddyl. We have shown that the approach manages to reduce the required number of iterations to reach the convergence from ~9.5 to only ~3.0 iterations for the singlestep motion and from ~6.2 to ~4.5 iterations for the multi-step motion, while maintaining the solution’s quality.

AB - In this paper, we propose a framework to build a memory of motion for warm-starting an optimal control solver for the locomotion task of a humanoid robot. We use HPP Loco3D, a versatile locomotion planner, to generate offline a set of dynamically consistent whole-body trajectory to be stored as the memory of motion. The learning problem is formulated as a regression problem to predict a single-step motion given the desired contact locations, which is used as a building block for producing multi-step motions. The predicted motion is then used as a warm-start for the fast optimal control solver Crocoddyl. We have shown that the approach manages to reduce the required number of iterations to reach the convergence from ~9.5 to only ~3.0 iterations for the singlestep motion and from ~6.2 to ~4.5 iterations for the multi-step motion, while maintaining the solution’s quality.

U2 - 10.1109/ICRA40945.2020.9196727

DO - 10.1109/ICRA40945.2020.9196727

M3 - Conference contribution

SN - 978-1-7281-7396-2

SP - 1357

EP - 1363

BT - 2020 IEEE International Conference on Robotics and Automation (ICRA)

PB - Institute of Electrical and Electronics Engineers

T2 - 2020 International Conference on Robotics and Automation

Y2 - 31 May 2020 through 31 August 2020

ER -

Learning How to Walk: Warm-starting Optimal Control Solver with Memory of Motion

Abstract

Publication series

Conference

Access to Document

Fingerprint

Projects

MEMMO - Memory of Motion

Cite this