IFAC PapersOnLine 50-1 (2017) 10456–10461
ScienceDirect
Available online at www.sciencedirect.com
2405-8963 © 2017, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved.
Peer review under responsibility of International Federation of Automatic Control.
10.1016/j.ifacol.2017.08.1975
© 2017, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved.
Keywords: optimal control, discrete mechanics, variational mechanics, Denavit-Hartenberg,
Legendre-Transformation, Furuta double pendulum, swing-up, trajectory planning
1. INTRODUCTION
For many years inverted pendulum has been associated as
the classical problem in many disciplines such as math-
ematics, physics, dynamics and control due to its chal-
lenging system properties like nonlinearity and underac-
tuation. Its practical correspondence to stabilization of
humanoid robot, vehicle during the rollover, rocket launch-
ing and a lot of others make the stabilization of inverted
pendulum widely used. However, despite the vast studies
on the swing-up maneuver from the downward position
to the upright position it is still a challenging problem.
The swing-up problem is approached either on a nonde-
terministic way or on a deterministic way. Energy based
techniques as in (
˚
Astr¨om and Furuta, 2000) and (Spong,
1995), fuzzy based techniques as in (Wang et al., 1996)
and (Tao et al., 2008), bang-bang based techniques as in
(Awrejcewicz et al., 2012) are nondeterministic, whereas
a proper trajectory planning as in (Gl¨ uck et al., 2013) are
deterministic. In (Graichen et al., 2007) a solution of a two-
point boundary problem with free parameters is proposed.
In (Jung and Wen, 2004) a nonlinear model predictive
control is deployed on a Furuta single pendulum. Up until
now, most of the existing techniques are applied on other
versions of pendulums like the inverted pendulum on a cart
or on a wheeled pendulum. However, due to its rotation na-
ture an inverted Furuta double pendulum introduces more
centrifugal and coriolis forces and thus raises the challenge
of the swing-up. In this study, the swing-up of the inverted
Furuta double pendulum has been deployed using the
efficiency feature of the Discrete Mechanics and Optimal
Control (DMOC) (Junge et al., 2006). In a merge of the op-
timal control, variational integrators (Hairer et al., 2006)
and discrete mechanics (Marsden and West, 2001) the
concept of DMOC offers a novel alternative for formulating
an optimization problem. Direct methods ((Hargraves and
Paris, 1987), (Von Stryk, 1993)) and indirect methods
(Rao, 2009) in both the objective function and the dif-
ferential equations are given in a continues form. Derived
from the variation of the Lagrange-d’Alembert principle
the Euler-Lagrange equations results in the differential
equations. Most of certain optimization applications are
based on either the direct method as in (Petropoulos et al.,
2002), (Somavarapu et al., 2016) and (Geiger et al., 2006)
or the indirect method as in (Yoshimura and Yamanaka,
1982) and (Ranieri and Ocampo, 2006). The direct method
is mostly used due to its convenient implementation and
the widely used optimization techniques. However, for
efficient optimization the indirect method seems to have
an advantage, since it deals with a less computational
cost for solving the canonical boundary value problem
resulting from the Pontryagin’s maximum principle. In
(Graichen, 2012) and (B¨ achle et al., 2013) the indirect
method is employed for efficient optimization in the con-
text of model predictive control. In contrast, DMOC pre-
sumes a discrete cost function and discrete Lagrange-
d’Alembert principle. Due to its variation, a discrete Euler-
Lagrange with discrete forces is obtained, which depends
only on the position q instead of both position q and
speed ˙ q like in the continues Euler-Lagrange. In fact, this
discretization occurs at an earlier stage than in the direct
and indirect methods. For the boundary conditions the
explicit dependency of the speed is however solved using
the Legendre-Transformation (Marsden and West, 2001).
As a result, the forced discrete Euler-Lagrange equations
together with the discrete boundary conditions are ob-
tained, which serve as equality constraints for a finite
dimensional nonlinear optimization problem (NLP). To
*
Institute of Control Systems, University of Kaiserslautern P.O. Box
3049, 67653 Kaiserslautern, Germany
(e-mail: ismail@eit.uni-kl.de, sliu@eit.uni-kl.de).
Abstract: This paper presents a more accurate model of a Furuta double pendulum for efficient
planning of optimal trajectories using the new approach of discrete mechanics and optimal
control (DMOC). Based on the variation of discrete mechanics a direct discretization of the
Lagrange-d’Alembert principle enables a novel formulation of the optimization problem. This
leads to less optimization parameters and thus more computational efficiency. To shed the
light on the advantages of the proposed method the swing-up maneuver of the Furuta double
pendulum is used to demonstrate the point-to-point trajectory planning. For this purpose,
a model of Furuta double pendulum based on Denavit-Hartenberg convention with higher
accuracy is presented additionally.
Jawad Ismail
*
Steven Liu
*
Efficient Planning of Optimal Trajectory
for a Furuta Double Pendulum Using
Discrete Mechanics and Optimal Control