Professional Documents
Culture Documents
A Passive-Biped Model With Multiple Routes To Chao
A Passive-Biped Model With Multiple Routes To Chao
net/publication/226758151
CITATIONS READS
17 330
2 authors, including:
Mahyar Naraghi
Amirkabir University of Technology
76 PUBLICATIONS 601 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Mahyar Naraghi on 05 October 2015.
RESEARCH PAPER
F. Farshimi · M. Naraghi
Abstract This paper presents a new passive-biped model shallow slopes. His biped model was then simplified to a
consisting of a simplest walking model beneath an upper three mass version, two at the feet and one at hip joint, con-
body, with no kinematic constraint. The upper body is at- nected by two links, as shown in Fig. 1. Garcia also made a
tached to the legs with a linear torsional spring. The model further simplification by assuming that hip mass is negligible
is a passive dynamic walker, so it walks down a slope without compared to the feet masses. Equations of motion were de-
energy input. The governing equations of motion are derived rived and integrated simultaneously with collision equations
and simulated for the parameter analysis purposes. Simula- to find the period-1 gait cycle. Since his model possessed
tion results reveal some different routes to chaos that have only a single parameter, the slope angle, the parametric study
not been observed in previous models. was also restricted to a single bifurcation diagram.
1 Introduction
a kinematic coupling to keep the body midway between the When a foot hits the ground at heelstrikes, it acts as a hinge
legs. Considering this new limb, the model seems to have and remains at the ground until the other foot (swing leg)
three degrees of freedom, though by applying the kinematic reaches heelstrikes. At any time during locomotion, only
constrain in equations, the model still has two degree of free- one foot is in contact with the ground, and supporting on
dom, just like the simplest walking model. both feet occurs instantaneously.
where
[(α + 1) cos ϕ − α cos(θ − ψ) cos(θ + ψ)] Fig. 4 Stable period-1 cycle. γ = 0.004 5 rad
A= , (6)
{β sin2 ϕ − α[cos ϕ − cos(θ − ψ) cos(θ + ψ)] + 1}
B = A(1 − cos ϕ), (7) Table 1 A set of initial conditions and parameter values
cos(θ − ψ) for stable walking
C = 1+ (1 − A). (8)
ν
Initial conditions Parameters
Finally, the end of every sequence can be detected by mon-
θ 0.186 3 α 0.05
itoring foot clearance function. In particular, when the fol-
lowing conditions are met, a heel strike is detected: (a) Foot θ̇ −0.152 9 β 0
clearance is crossing zero; (b) Derivative of foot clearance is ϕ 0.382 6 κ 0.091 7
negative. ϕ̇ −0.010 5 υ 0.4
In a similar approach to Ref. [3], equations are simpli- ψ −0.365 9 η 0.02
fied by assuming β → 0, denoting that the hip mass is negli-
ψ̇ 1.792 5 γ 0.004 5
gible compared to the foot mass, and the simplified equations
can thus be derived as follows
(1 + α)θ̈ − αν cos(ψ + θ)ψ̈ + ανψ̇2 sin(ψ + θ) A linear approach is adapted to evaluate the stability. In
this approach, any arbitrary initial condition could be consid-
− sin(θ − γ) − α sin(θ − γ) = 0, (9) ered as a set of fixed points plus a small deviation.
By integrating Eqs. (9)–(11) and applying Eq. (5) in ev- q∗ + ∆qn+1 = p(q∗ + ∆qn ) ≈ p(q∗ ) + J∆qn , (13)
ery heel strike, a complete walking cycle will be obtained,
hence the entire calculation is performed as Poincare map- where J is Jacobian matrix and derived as J = ∂ p/∂q. Thus,
ping, transforming state vector before collision to state vec- the eigenvalues of Jacobian matrix can be used to judge the
tor after heel strike. This map is also referred to the Stride walking cycle stability [8]. As depicted in Fig. 5, all Jaco-
function p [1] in the previous works. We use Matlab ode 45 bian matrix’s eigenvalues lie inside the unit circle, indicating
with a tolerance of 10−12 for computational integration. the stability of relevant walking cycle.
By integrating governing equations, stable walking cycles The main focus of this paper is to study the effect of parame-
can be found for some sets of parameter values. Figure 4 ter variation on model behaviors; therefore, the essential def-
depicts stable period-1 cycle for a selected set of initial con- initions of routes (transients) to chaos are briefly presented
ditions and parameters that are presented in Table 1. for a better understanding of the next section.
280 F. Farshimi, M. Naraghi
Fig. 9 Fixed points of period-6 and period-12 motion. a γ = 0.010 16; b γ = 0.010 23
As the parameter α increases, a different bifurcation diagram Fig. 11 Phase portrait for chaotic motion. γ = 0.010 14
with a new route to chaos is produced. Figure 14 depicts the
bifurcation diagram for a range of α. As α is increased, the
period-1 fixed points suddenly expand to a larger region of
fixed points. By plotting the attractor of this area (Fig. 15), it
can be seen clearly that the attractor is a quasi periodic type,
implying that although we have infinite fixed points, only
two frequencies with irrational fractions exist. Study on this
physical model reveals that for small range of α, the differ-
ence between upper body harmonic motion frequency and
walking motion frequency is not large enough to separate
the system from period-1 motion. However, as α increases,
these two main frequencies of the biped diverge, resulting
in a quasi periodic motion. Figure 16 shows phase portrait
of θ, ϕ, ψ for quasi periodic motion, it is clear that as it is
expected for 3D quasi periodic motion the attractor lies in a
torus (or more precisely described in topology [9] “a shape
that is homeomorphic to torus”). Further increase of α will Fig. 12 Phase portrait for chaotic motion. γ = 0.010 32
282 F. Farshimi, M. Naraghi
5 Conclusion
thought that this parameter affects the system in the same of Biomechanical Engineering 120, 281–288 (1998)
way as parameter α, bifurcation diagram demonstrates an or- 4 Tehrani Safa, A., Ghaffari Saadat, M., Naraghi, M.: Passive dy-
dinary period-doubling sequence. By further decrease of this namic of the simplest walking model: Replacing ramps with
parameter, the system will enter into chaotic region. stairs. Mechanism and Machine Theory 42(10), 1314–1325
(2007)
5 Wisse, M., Schwab, A.L., Vander Helm, F.C.T.: Passive dy-
Acknowledgment Authors would like to thank Andy Ruina, Mar- namic walking model with upper body. Robotica 22, 681–688
tin Wisse, Manoj Srinivasan and Ali Tehrani Safa for their valuable (2004)
guidance and advices. 6 Gomes, M.W., Ruina, A.L.: A walking model with no en-
ergy cost. Department of Theoretical and Applied Mechan-
ics, Cornell University, 25 Aug. (2004) http://ruina.tam.cornell.
References edu/research/topics/locomotion and robotics/
7 Kurz, M.J., Stergiou, N., Heidel, J., et al.: A template for the
exploration of chaotic locomotive patterns. Chaos Solutions &
1 McGeer, T.: Passive dynamic walking. International Journal of Fractals 23(2), 485–493 (2005)
Robotics Research 9(5), 62–82 (1990) 8 Hilborn, R.C.: Chaos and Nonlinear Dynamics, An Introduc-
2 McGeer, T.: Dynamics and control of bipedal locomotion. tion for Scientists and Engineers. 2nd edn. Oxford University
Journal of Theoretical Biology 163(3), 277–314 (1993) Press, Oxford (2000)
3 Garcia, M., Chatterjee, A., Ruina, A., et al.: The simplest walk- 9 Stillwell, J.: Classical Topology and Combinatorial Group The-
ing model: Stability, complexity, and scaling. ASME Journal ory. 2nd edn. Springer Press, Berlin (1993)