Proc. R. Soc.

B (2006) 273, 2861–2867

doi:10.1098/rspb.2006.3637
Published online 11 August 2006

Compliant leg behaviour explains basic

dynamics of walking and running
Hartmut Geyer1,*, Andre Seyfarth1 and Reinhard Blickhan2
1
Locomotion Laboratory, Friedrich-Schiller University Jena, Dornburger Strasse 23, 07743 Jena, Germany
2
Science of Motion, Friedrich-Schiller University Jena, Seidelstrasse 20, 07749 Jena, Germany
The basic mechanics of human locomotion are associated with vaulting over stiff legs in walking and
rebounding on compliant legs in running. However, while rebounding legs well explain the stance
dynamics of running, stiff legs cannot reproduce that of walking. With a simple bipedal spring–mass
model, we show that not stiff but compliant legs are essential to obtain the basic walking mechanics;
incorporating the double support as an essential part of the walking motion, the model reproduces the
characteristic stance dynamics that result in the observed small vertical oscillation of the body and
the observed out-of-phase changes in forward kinetic and gravitational potential energies. Exploring the
parameter space of this model, we further show that it not only combines the basic dynamics of walking
and running in one mechanical system, but also reveals these gaits to be just two out of the many solutions
to legged locomotion offered by compliant leg behaviour and accessed by energy or speed.
Keywords: biomechanics; human gait; spring–mass model

1. INTRODUCTION is also reflected in the forces acting on the centre of mass

In his seventeenth century volume ‘De motu animalium’, (COM; Marey 1894; Fischer 1899). For example, in
Borelli discussed walking as vaulting over stiff legs using a stance, the vertical ground reaction force (GRF) is
pair of compasses and noted the importance of rebound- characteristically M-shaped: it has one early and one late
ing on compliant legs in running (Borelli 1685). From that peak separated by a minimum around midstance, which
early account up to the present, walking and running have the inverted pendulum model cannot reproduce (figure 1;
been treated as different mechanical paradigms, and the Pandy 2003).
two corresponding models, the inverted pendulum model To account for these differences, more complex models
for walking (Alexander 1976; Mochon & McMahon of walking use more detailed representations of the leg
1980) and the spring-mass model for running (Blickhan components, including springs and dampers, multi-
1989; McMahon & Cheng 1990), have developed into the segments or neuromuscular structures (Siegler et al.
conceptual basis for our understanding and technical 1982; Gurp et al. 1987; Pandy & Berme 1988; Neptune
realization of legged locomotion. The models motivate the et al. 2001; Pandy 2003; Zajac et al. 2003). Although these
changes of kinetic and potential energies that are observed models describe the dynamics of walking much closer than
in each gait (Cavagna et al. 1964, 1977; Dickinson et al. an inverted pendulum can, and indicate compliant leg
2000), give insights into the remarkable universal behaviour to be relevant in walking, they are too complex
speed dependency that describes the walk–run transition to serve as conceptual models. Hence, despite being
inaccurate, the stiff-legged motion remains the mechanical
(Alexander 1989; Kram et al. 1997; Minetti 2001) and
paradigm for the walking gait (Dickinson et al. 2000;
reveal the importance of self-stability for legged systems
Srinivasan & Ruina 2005).
(Garcia et al. 1998; Kuo 1999; Seyfarth et al. 2002;
We argue that not stiff but compliant legs are
Ghigliazza et al. 2003) inspiring the construction of
fundamental to the walking gait. Hereto, we first introduce
walking (McGeer 1990; Collins et al. 2005) and running
the bipedal spring–mass model, which adds a second leg to
machines (Raibert 1986; Saranli et al. 2001; Cham et al.
the known running model and represents the simplest
2004).
walking model using compliant legs. We then look for
However, these models also show that, whereas stable locomotion of this model and show that its resulting
rebounding on compliant legs explains well the basic steady-state patterns reproduce those observed in walking.
mechanics of running, vaulting over stiff legs cannot truly Finally, we generalize the model and suggest that walking
reproduce that of walking (Full & Koditschek 1999). For and running are just two out of the many solutions to
instance, instead of the large vertical amplitudes suggested legged locomotion offered by compliant leg behaviour and
by vaulting over stiff legs, the upper body shows accessed by energy or speed.
comparably small vertical amplitudes during walking
(Weber & Weber; Lee & Farley 1998). This discrepancy
2. THE BIPEDAL SPRING–MASS MODEL
The model represents the body as point mass m at the
* Author and address for correspondence: Biomechatronics Group, COM and describes the legs as two massless, linear
MIT Media Laboratory, Massachusetts Institute of Technology,
20 Ames Street, Cambridge, MA 02139, USA (hartmut.geyer@ springs of equal rest length [0 and stiffness k (figure 2).
uni-jena.de). Both springs act independently and influence the model

Received 31 January 2006 2861 q 2006 The Royal Society

Accepted 1 June 2006
2862 H. Geyer and others Compliant leg behaviour in walking/running

which the (iC1)th apex is reached when the upward COM

motion stops (vertical COM velocity yZ _ 0), completing the
step. Owing to the parametric symmetry between both
springs, one step represents a basic gait cycle and its
identical repetition, the steady-state locomotion.

3. WALKING SOLUTIONS REPRODUCING

EXPERIMENTAL DATA
Depending on the chosen parameter values, the model
may take off in single support or stumble and fall down;
however, it also shows, and converges to, steady-state
locomotion. By searching for stable locomotion using the
return map of a single step (for details on the stability
analysis see appendix Aa–c), we find three characteristic
steady-state solutions A–C of the model, which have in
common that their stance-phase patterns resemble those
observed in animal and human walking (figure 3). The
horizontal GRF shows the observed change from negative
to positive values and the vertical axis, the double peak
that distinguishes the walking gait (Fx and Fy, first row of
subplots). Correspondingly, as in animals and humans
Figure 1. Standard conceptual models of legged locomotion (Lee & Farley 1998; Gard et al. 2004), the COM oscillates
and their predictive power with respect to walking and running
around its landing height in the vertical GRF with a
dynamics. The inverted pendulum and the spring–mass
system are the standard models for walking and running. smaller increase in height during stance than that of the
The model-predicted stance dynamics (red lines) fit experi- inverted pendulum motion (Dy, second row). Moreover,
mental data (black traces recorded from human treadmill closer than the inverted pendulum, the bipedal spring–
walking at 1.2 m sK1 and running at 4.0 m sK1) only for the mass model describes the out-of-phase changes in the
spring–mass model for running. Note that, in the inverted forward kinetic and the gravitational potential energies
pendulum dynamics, delta functions appear at 0 and 100% that occur in walking (DEk,x and DEp, third row).
stance time if one adds collision and push-off models imitating The stance-phase patterns of the three example
double support. Fx, y , horizontal and vertical ground reaction solutions A–C not only share general features of animal
force (GRF) normalized to body weight (bw). and human walking, but also have differences that reflect
those observed in walking at different speeds. First, the
dynamics only during stance when the spring force
patterns differ in their symmetry with respect to midstance
opposes the gravitational force. On the contrary, in
(50% of stance time). They are symmetric in A and C, but
swing, the respective spring has no physical meaning,
asymmetric in B. For instance, in A and C, the vertical
but describes a kinematic touchdown condition
GRF has two equal peaks that lead to the known M-shape.
yTD Z [0 sin a0 , given by the rest length [0 and the fixed On the contrary, in B, the vertical force has a first peak that
leg orientation a0 with respect to gravity. The transition is clearly higher than the second one (Fy, first row).
from swing to stance occurs when the swing-leg strikes the Furthermore, the patterns differ in their amplitudes,
ground, whereas that from stance to swing occurs when which are large in A and B, but only small in C. For
the spring reaches its rest length during lengthening. To instance, although for all three solutions A–C, the vertical
compare the model dynamics with that of human walking, displacement of the COM is smaller than that of an
we fix the parameters mass, rest limb length and inverted pendulum, in C, the COM remains close to the
gravitational acceleration to mZ80 kg, [0 Z 1 m and landing height throughout stance (Dy, second row).
gZ9.81 m sK2, respectively. Similar differences in symmetry and amplitude can be
To obtain steady-state solutions of this model, we found in animal and human walking when considering
investigate a single step, which is defined as the interval different speeds (Keller et al. 1996). For slow walking,
between two subsequent apex events marking the highest symmetric stance-phase patterns with small amplitudes
points of the COM trajectory. For example, in figure 2, the are observed that compare to the patterns in C. For faster
model starts at the ith apex with the left spring (black) in walking, patterns with larger amplitudes are observed that
single support and the right spring (white) describing the compare to the patterns in A or B.
swing-leg. Since the gravitational force exceeds the To investigate how the solutions A–C depend on the
opposing spring force, the left spring shortens while specific parameters chosen, we scan the physiologically
rotating forward and the COM height decreases (dotted plausible range of angle of attack a0, spring stiffness k and
line). When the right leg touches the ground (‘right TD’), system energy Es for stable locomotion of the model.
the model enters the double-support phase. The additional (Dimensional analysis shows that the model has only three
push of the right stance-spring reverses the vertical and independent parameters: angle of attack a0, dimensionless
decelerates the horizontal COM motion. Owing to spring stiffness kZ~ k[0 =ðmgÞ and dimensionless system
sufficient momentum, the forward progression continuous energy E s Z Es =ðmg [0 Þ, where Es is the constant system
~
to extend the left spring until it reaches its rest length. At energy of the conservative model. Without loss of general-
this instant, the left spring takes off (‘left TO’) and the ity, we can use their dimensional counterparts a0, k and
system enters the single support phase of the right spring in Es, since we fixed the remaining parameters m, g and [0 .)

Proc. R. Soc. B (2006)

 Compliant leg behaviour in walking/running H. Geyer and others 2863

Figure 2. The bipedal spring–mass model. The model has two independent, massless spring legs attached to a point mass m.
Both springs have stiffness k, rest length [0 and, in their swing phases, a constant orientation a0 with respect to gravity
( g, gravitational acceleration). A single step is shown that starts at the highest COM position in left leg single support (apex i ),
includes the double support ranging from right leg touchdown (right TD) to left leg take-off (left TO), and ends at the next apex
in right leg single support (apex iC1). FP, foot point position in single support.

a0 =69° a 0 =73° a0 =76° exp

A B C
k=14 kN m–1 k=14 kNm–1 k=20 kN m–1
Es =816 J Es =816 J Es =816 J

1 Fy
GRF (bw)

0.5

0
Fx
inverted
landing pendulum
5 height
∆y (cm)

–5

30
∆Ek,x
∆E (J)

–30 ∆Ep

0 50 100 0 50 100 0 50 100 0 50 100

stance time (%)
Figure 3. Stance-phase patterns of walking at about 1.2 m s . (A–C) Examples of three characteristic steady-state solutions of
K1

the bipedal spring–mass model are compared with (exp) experimental results (mean and s.d. shown as line and shaded area) of
five subjects (meanGs.d. of mass: 81G3.5 kg, leg length: 1.07G0.03 m) walking on a treadmill (Adal3D, TecMachine, France;
with force sensors recording horizontal and vertical GRFs). The subplots show horizontal and vertical GRFs, Fx and Fy; vertical
displacement, Dy; and changes in forward kinetic and gravitational potential energies, DEk,x and DEp. The vertical displacement
is compared with that of an inverted pendulum (dashed line). The shaded segments at the time-scales denote double supports.
The depicted lengths of the time-scales reflect the absolute stance times.

We find that the parameter adjustments leading to indicated by double force-peak icon). This domain splits
solutions, which describe animal- and human-like pat- into three sub-domains (slice through this domain at
terns similar to A–C, form a whole domain in this three- EsZ816 J shown in figure 4b), which correspond to
dimensional parameter space (figure 4a, walking domain the three characteristic solutions A–C. Within these

Proc. R. Soc. B (2006)

2864 H. Geyer and others Compliant leg behaviour in walking/running

(a) (b)

10

running 9
4500 Es = 816 J (v~1.2 m s–1)
domain
4000 8 50

3500 7
3000 6
2500 40
5
2000 4
3
1500

v (m s–1)
30

k (kN m–1)
Es (J)

C
1.6
walking
20
domains A
870 1.4

850 C
1.2
B 10 B
830
A 1.0
810
0.8 0
790 50 60 70 80
80
a 0 (deg)
770 70
)
50
40 60 eg
30 (d
20 a0
k (kN 10 50
m –1) 0

Figure 4. Parameter domains for stable walking and running. (a) Combinations of angle of attack a0, spring stiffness k and
system energy Es leading to stable locomotion are shown. Related to Es, the locomotion speed v is shown, which is the average
speed of all solutions that belong to one system energy (maximum deviation 0.1 m sK1 at EsZ800 J ). The model finds stable
walking at low energies or slow speeds (walking domains): next to the domain with double-peak patterns of the vertical GRF,
domains with multi-peak patterns exist (small icons). Owing to the limited scan resolution, only domains with up to five peaks
are resolved, and the four- and five-peak domains seem to overlap. Circles indicate the parameter sets of the examples A–C
shown in figure 3. In addition to walking, the model finds stable running with single-peak vertical GRF above an energy or speed
gap of about 500 J or 1.5 m sK1 (running domain). Note the different scales of system energy at the walking domains and the
running domain. (b) A slice at EsZ816 J (vw1.2 m sK1) through the walking domain with double-peak patterns is shown. Three
sub-domains of parameters exist that lead to three qualitatively different steady-state patterns (small icons) exemplified by the
three solutions A–C (compare figure 3).

sub-domains, the model is robust with respect to changes in 4. WALKING SOLUTIONS UNKNOWN FROM
swing-leg orientation and spring stiffness. Although different EXPERIMENTS
angles a0 or stiffness values k lead to different steady-state For lower system energies or slower speeds, the model
solutions, their stance-phase patterns show the same discovers new domains of parameters for stable loco-
characteristics. Across the three sub-domains, the model motion, which lead to steady-state walking patterns
still tolerates changes in its leg parameters, but assumes unknown from experiments (figure 4a, walking domains
characteristically different steady-state solutions. For indicated by multiple force-peak icons). For instance, the
instance, at a system energy of EsZ816 J, which corresponds vertical GRF patterns show more than two force peaks. In
to an average walking speed of about 1.2 m sK1 in the fact, toward small system energies, the number of force
model, solutions with small amplitudes, such as in C, peaks increases with each new domain; and although our
are obtained for steep angles and stiff-leg springs limited scan resolution resolves only domains with up to
ða0 O 758; kO 17 kN mK1 Þ; solutions with larger amplitudes, the five-peak force pattern, their number grows
such as in A or B, result from flatter angles and more incessantly.
compliant leg springs ða0 ! 758; k! 17 kN mK1 Þ. Here, the The simple analogue of a vertical spring that is loaded
leg stiffness predicted for walking is so low that it approaches by a mass m helps us to understand these multi-peak
values reported for running (Farley et al. 1993). patterns. Owing to the mass, this spring has a compressed

Proc. R. Soc. B (2006)

 Compliant leg behaviour in walking/running H. Geyer and others 2865

rest position [
0 Z [0 Kðmg=kÞ, where k and [0 are the (Minetti et al. 1994), which more closely resembles the
spring’s stiffness and normal rest length. Any small natural situation.


deflection causes this system to oscillate aroundpffiffiffiffiffiffiffiffi [ 0 in the
vertical GRF with a defined period T Z 2p m=k, and the
oscillation results in the peaks and valleys of the GRF 6. DISCUSSION
measured underneath the spring. If the loaded spring Our results suggest that the two fundamental gaits of
remains in the vertical, the oscillation goes on incessantly walking and running are much less different than generally
and an unlimited number of force peaks is recorded. assumed; with the same compliant stance-leg behaviour
However, if the spring rotates forward additionally, the found in running, a bipedal spring–mass model could
oscillation is limited as it falls down; the faster it rotates, reproduce the stance dynamics observed in walking
the lesser are the number of force peaks recorded. (figure 3). Thus, it seems that, rather than a stiff-legged
The natural spring oscillation and its limitation by inverted pendulum gait, walking is, like running, a
forward rotation also apply to the bipedal spring–mass bouncing gait. Here, the sequence of single and double
model; however, its double support (which is neglected in support in walking replaces that of flight and stance in
the inverted pendulum model) adds a crucial component running. Moreover, the identified multi-peak patterns
to obtain multi-peak patterns in walking. The second leg show that walking and running are just two out of the
not only prevents the COM from falling during loco- many stable solutions to legged locomotion of the same
motion, but also gradually increases and decreases the mechanical system. Each of these multi-peak solutions
effective load that acts on the other spring leg in its early occupies a separate domain in the parameter space
and late stance. This ‘load sharing’ allows a stance spring (figure 4a). In particular, the walking and running
to start from its initial rest length [0 , oscillate around the domains are isolated by a gap in system energy or
length [
0 without completely relaxing in single support, locomotion speed, which could explain why both
and finally resume its rest length [0 at the end of stance. gaits are perceived as such distinct gaits in animal and
The stance spring oscillation and the gradual transition of human locomotion, even though they represent the
support by the opposite leg require a precise timing, same mechanical concept that is based on compliant
which is met only by distinct sets of parameters that form leg behaviour.
the separate parameter domains for stable locomotion The results also challenge the view that the walking
of the model. efficiency depends on how close the COM motion
resembles that of an inverted pendulum. Classically,
walking efficiency is quantified by the percentage recovery,
5. WALKING AND RUNNING MECHANICS a parameter that determines how much of the stride
COMBINED IN ONE MODEL energy is recovered using the inverted pendulum’s
Earlier, we mentioned that, the bipedal spring–mass compensating exchange of gravitational potential and
model may also take off in single support. This happens kinetic energies (Cavagna et al. 1977; Mochon &
if the stance spring produces too large a rebound in single- McMahon 1980; Dickinson et al. 2000). For an ideal
support and relaxes completely, even though it is fully stiff-legged walk, the percentage recovery would be 100%,
loaded by the mass. For our simple analogue of a vertical yet walking experiments show that it reaches at best 70%
spring–mass system, the same happens if the deflection in bipeds and only 35–50% in quadrupeds (Cavagna et al.
from the compressed rest position [
0 is too large and the 1977; Minetti et al. 1999). The difference is mostly related
system energy is too high. Although the spring still to the double supports in which reversing the COM in the
oscillates around [
0 in the vertical, it has intermittent vertical disturbs the motion of consecutive inverted
flight phases and, in the stance phases, produces only pendulum arcs (Alexander 1976, 1991; McGeer 1990).
a single rebound with one force peak of the vertical However, the bipedal spring–mass model shows that
GRF—a behaviour related to bouncing gaits. Conse- double supports are crucial to obtain the walking
quently, the bipedal spring–mass model cannot only walk, dynamics and also demonstrates that a resulting low
but also run (see also appendix Ad ). percentage recovery (15–35% for examples A–C) does not
As a result, the bipedal model unifies both gaits with imply inefficient walking. By transiently storing in the leg
one mechanical concept and extending the systematic springs the energy that would otherwise be lost during
parameter scan to higher system energies additionally double support, the model recovers 100% of the stride
reveals the domain of parameters that lead to stable energy. Recent experimental findings support such an
running of the model (figure 4a, running domain elastic contribution (Fukunaga et al. 2001). Thus, walking
indicated by single force-peak icon). This domain has efficiency seems to depend less on how close the COM
already been described in an earlier study on spring–mass trajectory follows inverted pendulum arcs, but more on
running (Seyfarth et al. 2002). It requires a minimum how much of the stride energy can be stored elastically
locomotion speed of about 3 m sK1, which introduces a when redirecting the COM in double support.
speed gap of about 1.5 m sK1 between the two gaits. A The bipedal spring–mass model is a very simple model
similar speed gap, albeit not as large, is found in of legged locomotion, which in contrast to more complex
experiments on the human gait transition. Even though representations (Pandy 2003; Zajac et al. 2003) includes
humans prefer to switch from walking to running at one only two essential features: bipedalism and leg compli-
speed of about 2.3 m sK1 when they are instructed to walk ance. Yet, this model explains the basic dynamics of
or run at different speeds on a treadmill (Thorstensson & walking and running, and unites both gaits within one
Roberthson 1987; Hreljac 1993), they immediately switch consistent mechanical framework. It may therefore serve
from walking at about 1.8 m sK1 to running at about in future as a general gait template (Full & Koditschek
2.3 m sK1 during spontaneous overground progression 1999) that guides more complex models in exploring, and

Proc. R. Soc. B (2006)

2866 H. Geyer and others Compliant leg behaviour in walking/running

technical systems in realizing, legged locomotion from (c) Numerical return map analysis
walking to the walk–run transition to running. We exploit the convergence to stable fixed points to find
the steady-state locomotion of the model using the apex
The authors thank A. Biewener for helpful comments on the
manuscript. This work was supported by an Emmy-Noether return map. Instead of computing R from single-step
Grant (SE1042/1-5) of the German Research Foundation simulations of an extensive grid of initial apex states, we
(DFG) to A.S. and by a Marie-Curie Outgoing International investigate only 50 such states (see below), but for each of
Fellowship (MOIF-CT-20052-022244) of the European them, iterate the step simulation. If any of these states lies
Union to H.G. in the basin of attraction of a stable fixed point, the model
converges to the steady-state trajectory during iteration.
For practical reasons, however, we stop after 99 steps and
APPENDIX A verify that an identified fixed point is a stable one by
(a) Model equations and simulation environment checking l1,2 in a small neighbourhood of it.
For a single step from one apex (i ) to the next (iC1, The 50 initial apex states are 50 equally distributed
compare figure 2), the equations governing the motion apex heights yi ranging from ymin Z [0 sin a0 (landing
of the COM are mx€ Z Px and my€ Z PyKmg in the condition of swing leg) to ymax Z [0 (take-off condition of
initial left leg single support, mx€ Z PxKQðd KxÞ and stance leg) with the initial horizontal position fixed to
my€ Z PyC Q yKmg in the intermittent double support, DxiZ0. This restriction to vertical spring positions as
and mx€ ZKQðd KxÞ and my€ Z Q yKmg pinffiffiffiffiffiffiffiffiffiffiffiffiffiffi the ffi final initial apex states assumes that they cover the basin of
right leg p single support,
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi where P Z kð [0 = x 2
C y2 K1Þ, attraction of stable solutions. The assumption is justified
QZ kð[0 = ðd KxÞ C y K1Þ and d Z FPiC1;x KFPi;x with
2 2 for symmetric steady-state trajectories (examples A and C
FP denoting the foot point of a stance spring. The in figure 3) as they correspond to fixed points with
model is implemented in the M ATLAB /S IMULINK DxapexZ0, but it may be wrong for asymmetric steady-
environment (Rel. 14, Mathworks, Inc., Natick, MA, state trajectories (example B in figure 3) as they
USA) and simulations are run using the embedded correspond to fixed points with Dxapexs0. However,
variable step integrator ode113 (maximum step size: preliminary tests that compared this simplified search
10K3, absolute and relative error tolerance: 10K6). algorithm with extensive return map scans, including
asymmetric start positions xD,is0, showed no substantial
(b) Apex return map differences in finding stable asymmetric, as well as
During a walking step that starts at apex i, the simulation symmetric, steady-state solutions.
aborts if the model turns backward, takes off in single
support or stumbles and falls down ( yiC1!yTD). The (d) Extension to running
simulation otherwise stops if the model reaches the next Running needs no different model, but two formal
apex iC1. The relationship between the initial system changes in the simulation. First, we enable flight phases.
state at apex i and the resulting system state at apex iC1 Starting at apex i, the former sequence of single–
defines the apex return map, whose fixed points identify double–single support phase now can be paralleled by a
the steady-state locomotion of the model. At any apex sequence of flight–stance–flight phase. Reaching apex iC1
(index ‘apex’), the system state simplifies to only two in the second flight phase completes a step. The apex
independent variables: the horizontal distance between return map simplifies to R:yi/yiC1 (Seyfarth et al. 2002).
COM and foot point of the stance spring (FP; figure 2), Its analysis remains the same as described earlier. Second,
Dxapex Z xapex KFPapex;x , and the apex height, y apex . to avoid hopping with two parallel legs, we ensure that
This simplification holds since Dxapex accounts for the only one leg can land at a time.
influence of x apex on the dynamicsffi of a step,
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
x_apex Z 2Es =mKk=m ð[0 K[apex Þ2 K2gyapex , where Es is
the constant system energy of the conservative spring–
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi REFERENCES
mass model and [apex Z Dx2apex C y2apex ,
and y_apex Z 0 by Alexander, R. 1976 Mechanics of bipedal locomotion. In
definition of the apex event. Consequently, the apex Perspectives in experimental biology (ed. P. S. Davies),
return map is defined as R:(Dx, y)i/(Dx, y)iC1. A fixed pp. 493–504. Oxford, UK: Pergamon Press.
Alexander, R. 1989 Optimization and gaits in the locomotion
point R satisfies the condition (Dx, y)iC1Z(Dx, y)i and
of vertebrates. Physiol. Rev. 69, 1199–1227.
corresponds to steady-state locomotion of the walking Alexander, R. 1991 Energy-saving mechanisms in walking
model (we investigate only period-one limit cycles). and running. J. Exp. Biol. 160, 55–69.
A stable fixed point additionally satisfies that, in its Blickhan, R. 1989 The spring–mass model for running and
neighbourhood, the eigenvalues l1,2 of the Jacobian matrix hopping. J. Biomech. 22, 1217–1227. (doi:10.1016/0021-
0 1 9290(89)90224-8)
vDxiC1 vDxiC1
Borelli, G. 1685. De motu animalium, vol. 1. Leiden, The
B vDxi vyi C
B C Netherlands: Lugduni.
DRab Z B B C
vy vy C Cavagna, G., Saibene, F. & Margaria, R. 1964 Mechanical
@ iC1 iC1 A
work in running. J. Appl. Physiol. 19, 249–256.
vDxi vyi
Cavagna, G., Heglund, N. & Taylor, C. 1977 Mechanical
work in terrestrial locomotion: two basic mechanisms for
lie within the unit circle (jl1,2j!1) (Strogatz 2001). A minimizing energy expenditure. Am. J. Physiol. 233,
stable fixed point does not require that the fixed-point 243–261.
condition be met exactly when computing R, as the model Cham, J., Karpick, J. K. & Cutkosky, M. R. 2004 Stride
converges to the steady-state trajectory also from slightly period adaptation for a biomimetic running hexapod. Int.
disturbed apex states. J. Robotics Res. 23, 141–153.

Proc. R. Soc. B (2006)

 Compliant leg behaviour in walking/running H. Geyer and others 2867

Collins, S., Ruina, A., Tedrake, R. & Wisse, M. 2005 Efficient McMahon, T. & Cheng, G. 1990 The mechanism of running:
bipedal robots based on passive–dynamic walkers. Science how does stiffness couple with speed? J. Biomech. 23,
307, 1082–1085. (doi:10.1126/science.1107799) 65–78. (doi:10.1016/0021-9290(90)90042-2)
Dickinson, M., Farley, C., Full, R., Koehl, M., Kram, R. & Minetti, A. 2001 Walking on other planets. Nature 409,
Lehman, S. 2000 How animals move: an integrative view. 467–468. (doi:10.1038/35054166)
Science 288, 100–106. (doi:10.1126/science.288.5463.100) Minetti, A., Ardigo, L. & Saibene, F. 1994 The transition
Farley, C., Glasheen, J. & McMahon, T. 1993 Running between walking and running in humans: metabolic and
springs: speed and animal size. J. Exp. Biol. 185, 71–86. mechanical aspects at different gradients. Acta Physiol.
Fischer, O. 1899. Der Gang des Menschen 2. Teil: Die Bewegung Scand. 150, 315–323.
des Gesamtschwerpunktes und die äußeren Kräfte. Abhan- Minetti, A., Ardigo, L., Reinach, E. & Saibene, F. 1999 The
dlungen der mathematisch–physischen Klasse der Königl, relationship between mechanical work and energy expen-
vol. 25. Leipzig, Germany: Sächsischen Gesellschaft der diture of locomotion in horses. J. Exp. Biol. 202,
Wissenschaften. 2329–2338.
Fukunaga, T., Kubo, K., Kawakami, Y., Fukashiro, S., Mochon, S. & McMahon, T. 1980 Ballistic walking.
Kanehisa, H. & Maganaris, C. 2001 In vivo behaviour of J. Biomech. 13, 49–57. (doi:10.1016/0021-9290(80)
human muscle tendon during walking. Proc. R. Soc. B 268, 90007-X)
229–233. (doi:10.1098/rspb.2000.1361) Neptune, R., Kautz, S. & Zajac, F. 2001 Contribution of the
Full, R. & Koditschek, D. 1999 Templates and anchors: individual ankle plantar flexors to support, forward
neuromechanical hypotheses of legged locomotion on progression and swing initiation during walking.
land. J. Exp. Biol. 202, 3325–3332. J. Biomech. 34, 1387–1398. (doi:10.1016/S0021-9290
Garcia, M., Chatterjee, A., Ruina, A. & Coleman, M. 1998 (01)00105-1)
The simplest walking model: stability, complexity, and Pandy, M. 2003 Simple and complex models for studying
scaling. J. Biomech. Eng. 120, 281–288. muscle function in walking. Phil. Trans. R. Soc. B 358,
1501–1509. (doi:10.1098/rstb.2003.1338)
Gard, S., Miff, S. & Kuo, A. 2004 Comparison of kinematic
Pandy, M. & Berme, N. 1988 Synthesis of human walking: a
and kinetic methods for computing the vertical motion of
planar model for single support. J. Biomech. 21,
the body center of mass during walking. Hum. Mov. Sci.
1053–1060. (doi:10.1016/0021-9290(88)90251-5)
22, 597–610. (doi:10.1016/j.humov.2003.11.002)
Raibert, M. 1986 Legged robots that balance. Cambridge, UK:
Ghigliazza, R., Altendorfer, R., Holmes, P. & Koditschek, D.
MIT Press.
2003 A simply stabilized running model. SIAM J. Appl.
Saranli, U., Buehler, M. & Koditschek, D. 2001 Rhex: a
Dyn. Syst. 2, 187–218. (doi:10.1137/S111111110240
simple and highly mobile hexapod robot. Int. J. Rob. Res.
8311)
20, 616–631. (doi:10.1177/02783640122067570)
Gurp, M. V., Schamhardt, H. & Crowe, A. 1987 The ground Seyfarth, A., Geyer, H., Günther, M. & Blickhan, R. 2002 A
reaction force pattern from the hindlimb of the horse movement criterion for running. J. Biomech. 35, 649–655.
simulated by a spring model. Acta Anat. 129, 31–33. (doi:10.1016/S0021-9290(01)00245-7)
Hreljac, A. 1993 Preferred and energetically optimal gait Siegler, S., Seliktar, R. & Hyman, W. 1982 Simulation of
transition speeds in human locomotion. Med. Sci. Sports human gait with the aid of a simple mechanical model.
Exerc. 25, 1158–1162. J. Biomech. 15, 415–425. (doi:10.1016/0021-9290(82)
Keller, T., Weisberger, A., Ray, J., Hasan, S., Shiavi, R. & 90078-1)
Spengler, D. 1996 Relationship between vertical ground Srinivasan, M. & Ruina, A. 2005 Computer optimization of a
reaction force and speed during walking, slow jogging, and minimal biped model discovers walking and running.
running. Clin. Biomech. 11, 253–259. (doi:10.1016/0268- Nature 439, 72–75. (doi:10.1038/nature04113)
0033(95)00068-2) Strogatz, S. 2001 Nonlinear dynamics and chaos. New York,
Kram, R., Domingo, A. & Ferris, P. 1997 Effect of reduced NY: Perseus.
gravity on the preferred walk–run transition speed. J. Exp. Thorstensson, A. & Roberthson, H. 1987 Adaptations to
Biol. 200, 821–826. changing speed in human locomotion: speed of transition
Kuo, A. 1999 Stabilization of lateral motion in passive between walking and running. Acta Physiol. Scand. 131,
dynamic walking. Int. J. Rob. Res. 18, 917–930. (doi:10. 211–214.
1177/02783649922066655) Weber, W. & Weber, E. 1836 Mechanik der menschlichen
Lee, C. & Farley, C. 1998 Determinants of the center of mass Gehwerkzeuge. Göttingen, Germany: Dietrichsche
trajectory in human walking and running. J. Exp. Biol. Buchhandlungen.
201, 2935–2944. Zajac, F., Neptune, R. & Kautz, S. 2003 Biomechanics and
Marey, E. 1894 Le mouvement. Paris, France: Masson. muscle coordination of human walking part II: lessons
McGeer, T. 1990 Passive dynamic walking. Int. J. Rob. Res. 9, from dynamical simulations and clinical implications. Gait
62–82. Posture 17, 1–17. (doi:10.1016/S0966-6362(02)00069-3)

Proc. R. Soc. B (2006)