The Reaction Mass Biped: Geometric Mechanics and Control

Avinash Siravuru, Sasi P. Viswanathan, Koushil Sreenath, Amit Sanyal

Research output: Contribution to journalArticle

Abstract

Inverted Pendulum based reduced order models offer many valuable insights into the much harder problem of bipedal locomotion. While they help in understanding leg behavior during walking, they fail to capture the natural balancing ability of humans that stems from the variable rotational inertia on the torso. In an attempt to overcome this limitation, the proposed work introduces a Reaction Mass Biped (RMB). It is a generalization of the previously introduced Reaction Mass Pendulum (RMP), which is a multi-body inverted pendulum model with an extensible leg and a variable rotational inertia torso. The dynamical model for the RMB is hybrid in nature, with the roles of stance leg and swing leg switching after each cycle. It is derived using a variational mechanics approach, and is therefore coordinate-free. The RMB model has thirteen degrees of freedom, all of which are considered to be actuated. A set of desired state trajectories that can enable bipedal walking in straight and curved paths are generated. A control scheme is then designed for asymptotically tracking this set of trajectories with an almost global domain-of-attraction. Numerical simulation results confirm the stability of this tracking control scheme for different walking paths of the RMB. Additionally, a discrete dynamical model is also provided along-with an appropriate Geometric Variational Integrator (GVI). In contrast to non-variational integrators, GVIs can better preserve energy terms for conservative mechanical systems and stability properties (achieved through energy-like lyapunov functions) for actuated systems.

Original languageEnglish (US)
Pages (from-to)155-173
Number of pages19
JournalJournal of Intelligent and Robotic Systems: Theory and Applications
Volume89
Issue number1-2
DOIs
StatePublished - Jan 1 2018

Fingerprint

Mechanics
Pendulums
Trajectories
Lyapunov functions
Computer simulation

Keywords

  • Discrete mechanics
  • Geometric control
  • Legged robots
  • Non-linear control

ASJC Scopus subject areas

  • Software
  • Control and Systems Engineering
  • Mechanical Engineering
  • Industrial and Manufacturing Engineering
  • Artificial Intelligence
  • Electrical and Electronic Engineering

Cite this

The Reaction Mass Biped : Geometric Mechanics and Control. / Siravuru, Avinash; Viswanathan, Sasi P.; Sreenath, Koushil; Sanyal, Amit.

In: Journal of Intelligent and Robotic Systems: Theory and Applications, Vol. 89, No. 1-2, 01.01.2018, p. 155-173.

Research output: Contribution to journalArticle

Siravuru, Avinash ; Viswanathan, Sasi P. ; Sreenath, Koushil ; Sanyal, Amit. / The Reaction Mass Biped : Geometric Mechanics and Control. In: Journal of Intelligent and Robotic Systems: Theory and Applications. 2018 ; Vol. 89, No. 1-2. pp. 155-173.
@article{fd2524f8ca734cbba7a55bf251c0156e,
title = "The Reaction Mass Biped: Geometric Mechanics and Control",
abstract = "Inverted Pendulum based reduced order models offer many valuable insights into the much harder problem of bipedal locomotion. While they help in understanding leg behavior during walking, they fail to capture the natural balancing ability of humans that stems from the variable rotational inertia on the torso. In an attempt to overcome this limitation, the proposed work introduces a Reaction Mass Biped (RMB). It is a generalization of the previously introduced Reaction Mass Pendulum (RMP), which is a multi-body inverted pendulum model with an extensible leg and a variable rotational inertia torso. The dynamical model for the RMB is hybrid in nature, with the roles of stance leg and swing leg switching after each cycle. It is derived using a variational mechanics approach, and is therefore coordinate-free. The RMB model has thirteen degrees of freedom, all of which are considered to be actuated. A set of desired state trajectories that can enable bipedal walking in straight and curved paths are generated. A control scheme is then designed for asymptotically tracking this set of trajectories with an almost global domain-of-attraction. Numerical simulation results confirm the stability of this tracking control scheme for different walking paths of the RMB. Additionally, a discrete dynamical model is also provided along-with an appropriate Geometric Variational Integrator (GVI). In contrast to non-variational integrators, GVIs can better preserve energy terms for conservative mechanical systems and stability properties (achieved through energy-like lyapunov functions) for actuated systems.",
keywords = "Discrete mechanics, Geometric control, Legged robots, Non-linear control",
author = "Avinash Siravuru and Viswanathan, {Sasi P.} and Koushil Sreenath and Amit Sanyal",
year = "2018",
month = "1",
day = "1",
doi = "10.1007/s10846-017-0508-7",
language = "English (US)",
volume = "89",
pages = "155--173",
journal = "Journal of Intelligent and Robotic Systems: Theory and Applications",
issn = "0921-0296",
publisher = "Springer Netherlands",
number = "1-2",

}

TY - JOUR

T1 - The Reaction Mass Biped

T2 - Geometric Mechanics and Control

AU - Siravuru, Avinash

AU - Viswanathan, Sasi P.

AU - Sreenath, Koushil

AU - Sanyal, Amit

PY - 2018/1/1

Y1 - 2018/1/1

N2 - Inverted Pendulum based reduced order models offer many valuable insights into the much harder problem of bipedal locomotion. While they help in understanding leg behavior during walking, they fail to capture the natural balancing ability of humans that stems from the variable rotational inertia on the torso. In an attempt to overcome this limitation, the proposed work introduces a Reaction Mass Biped (RMB). It is a generalization of the previously introduced Reaction Mass Pendulum (RMP), which is a multi-body inverted pendulum model with an extensible leg and a variable rotational inertia torso. The dynamical model for the RMB is hybrid in nature, with the roles of stance leg and swing leg switching after each cycle. It is derived using a variational mechanics approach, and is therefore coordinate-free. The RMB model has thirteen degrees of freedom, all of which are considered to be actuated. A set of desired state trajectories that can enable bipedal walking in straight and curved paths are generated. A control scheme is then designed for asymptotically tracking this set of trajectories with an almost global domain-of-attraction. Numerical simulation results confirm the stability of this tracking control scheme for different walking paths of the RMB. Additionally, a discrete dynamical model is also provided along-with an appropriate Geometric Variational Integrator (GVI). In contrast to non-variational integrators, GVIs can better preserve energy terms for conservative mechanical systems and stability properties (achieved through energy-like lyapunov functions) for actuated systems.

AB - Inverted Pendulum based reduced order models offer many valuable insights into the much harder problem of bipedal locomotion. While they help in understanding leg behavior during walking, they fail to capture the natural balancing ability of humans that stems from the variable rotational inertia on the torso. In an attempt to overcome this limitation, the proposed work introduces a Reaction Mass Biped (RMB). It is a generalization of the previously introduced Reaction Mass Pendulum (RMP), which is a multi-body inverted pendulum model with an extensible leg and a variable rotational inertia torso. The dynamical model for the RMB is hybrid in nature, with the roles of stance leg and swing leg switching after each cycle. It is derived using a variational mechanics approach, and is therefore coordinate-free. The RMB model has thirteen degrees of freedom, all of which are considered to be actuated. A set of desired state trajectories that can enable bipedal walking in straight and curved paths are generated. A control scheme is then designed for asymptotically tracking this set of trajectories with an almost global domain-of-attraction. Numerical simulation results confirm the stability of this tracking control scheme for different walking paths of the RMB. Additionally, a discrete dynamical model is also provided along-with an appropriate Geometric Variational Integrator (GVI). In contrast to non-variational integrators, GVIs can better preserve energy terms for conservative mechanical systems and stability properties (achieved through energy-like lyapunov functions) for actuated systems.

KW - Discrete mechanics

KW - Geometric control

KW - Legged robots

KW - Non-linear control

UR - http://www.scopus.com/inward/record.url?scp=85014001686&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85014001686&partnerID=8YFLogxK

U2 - 10.1007/s10846-017-0508-7

DO - 10.1007/s10846-017-0508-7

M3 - Article

AN - SCOPUS:85014001686

VL - 89

SP - 155

EP - 173

JO - Journal of Intelligent and Robotic Systems: Theory and Applications

JF - Journal of Intelligent and Robotic Systems: Theory and Applications

SN - 0921-0296

IS - 1-2

ER -