The Veterinary Journal 198 (2013) e3–e8

Contents lists available at ScienceDirect

The Veterinary Journal

journal homepage: www.elsevier.com/locate/tvjl

Gait as solution, but what is the problem? Exploring cost, economy

and compromise in locomotion
John E.A. Bertram ⇑
Faculty of Medicine, University of Calgary, 3330 Hospital Drive NW, Calgary, Alberta T2N 1N4, Canada

a r t i c l e i n f o a b s t r a c t

Keywords: Many studies have examined how legged mammals move, defining ‘what’ happens in locomotion. How-
Locomotion ever, few ask ‘why’ those motions occur as they do. The energetic and functional constraints acting on an
Gait animal require that locomotion should be metabolically ‘cost effective’ and this in large part determines
the strategies available to accomplish the task. Understanding the gaits utilised, within the spectrum of
gaits possible, and determination of the value of specific relationships among speed, stride length, stride
frequency and morphology, depends on identifying the fundamental costs involved and the effects of
different movement strategies on those costs. It is argued here that a fundamental loss associated with
moving on limbs (centre of mass momentum and energy loss) and two costs involved with controlling
and replacing that loss (muscular work of the supporting limb during stance and muscular work of
repositioning the limbs during swing) interact to determine the cost trade-offs involved and the optimi-
sation strategies available for each species and speed. These optimisation strategies are what has been
observed and characterised as gait.
Ó 2013 Elsevier Ltd. All rights reserved.

Introduction in which the animal operates. For example, each of the preferred
gaits of horses is selected and utilised to provide the least cost of
Gait is a pattern of movement that ultimately allows the animal transport over a certain range of speed, with the walk being used
to move effectively across a substrate. A wide variety of movement at slow speeds, the trot at intermediate speeds and the gallop at
patterns could result in locomotion, but a few distinctive patterns high speeds (Hoyt and Taylor, 1981). The operational conditions
(and variations on each of these main themes) are almost univer- of locomotion and the appropriate movement strategies available
sally used by quadrupeds. Historically from as early as Aristotle are predominantly influenced by the speed of progression,
(ca. 350 BC) (Goiffon and Vincent, 1779; Marey, 1873; Hildebrand, although other factors from the physical properties of the sub-
1965; Abourachid, 2003; Farquharson, 2007) and by tradition in strate, such as stiffness (Wilson and Pardoe, 2001) and slope (Self
equine practice, gait has been characterised largely on the basis et al., 2012), also have an influence.
of footfall sequence. This is a relatively easily observed and quan- It is generally assumed that the gait pattern observed under
tified aspect of animal locomotion, especially with modern analyt- normal circumstances is the only one available, potentially pro-
ical technology, such as high speed video, and newly emerging grammed by a neural control system originating in ancient coordi-
mathematical techniques, such as linear discriminant analysis nation centres of the brain and spinal cord (Grillner, 1975; McCrae
(Robilliard et al., 2007). and Rybak, 2008). However, evidence from analysis of human gait
Although detailed descriptions of footfall patterns can be used selection (humans are a reasonably tractable species sometimes
to define gaits and to document differences among patterns used willing to adapt to useful experimental conditions that other spe-
by different species in various circumstances (or even different cies resist), indicate that the control of key gait parameters, such
individuals), such characterisation does not inform us of the as the relationship between speed, stride frequency and stride
functional purpose of each gait or provide information on why length, are largely determined by the metabolic cost associated
each exists as it does. with each option (Bertram and Ruina, 2001). This results in the
The commonly observed, or ‘normal’, gaits are the ones that normal and spontaneous selection of quite unusual gait parame-
emerge from the range of possible gaits because they allow for ters under experimentally imposed circumstances, such as moving
metabolically cost-effective movement under the circumstances to a specified frequency or stride length.
These gait parameters are selected in a manner that minimises
the cost of locomotion under each set of specific conditions
⇑ Tel.: +1 403 2109857.
(Bertram, 2005; Snaterse et al., 2011; Fig. 1). The brain and spinal
E-mail address: jbertram@ucalgary.ca

1090-0233/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.tvjl.2013.09.025
e4 J.E.A. Bertram / The Veterinary Journal 198 (2013) e3–e8

the animal by producing positive work. Negative work, however,

turns out to be just one of many mechanisms through which
mechanical energy can be lost from the system (where the organ-
ism is the ‘system’ of interest). The process of energy loss actually
depends on a far more fundamental, but largely neglected, feature
of legged locomotion: collision-based loss caused by the interac-
tion of the moving mass and its solid support.
When the term ‘collision’ is mentioned, it is common to think of
violent crashes between objects such as automobiles. However, in
dynamics, a collision is defined simply as a discontinuity of the
velocity (magnitude or direction) of a travelling mass. Any such
discontinuity requires the application of force and will involve
the transfer or loss of energy. Within the precision of dynamics,
the term ‘discontinuity’ also has a specific meaning, where it is lim-
ited to an instantaneous event. In the real world, instantaneous
events require infinite forces, so they do not exist. However,
although such events occur more slowly in reality than in theory,
there are some aspects of collision theory that do translate to ener-
getic consequences in the real world of locomotion. As a simple
Fig. 1. Diagram showing how gait parameters, such as speed, step length and step example that most of us can relate to, consider a playground swing.
frequency, are selected in human walking to minimise energetic expenditure under
A common playground swing is an example of an active pendu-
the circumstances the system operates within. The light blue contours represent the
metabolic cost surface, where each contour has greater magnitude than the one lum. Much of the swinging motion is maintained through the spon-
inside it (minimum cost occurs at the central circle where the three linear taneous (and passive) exchange between potential energy (at the
relationships cross). For an individual walking at a determined speed (constant highest point of the swing) and translational kinetic energy (at
level on the y-axis), such as on a treadmill, minimum cost will occur using the step
the lowest point of the swing). Children learn early that they are
frequency where a horizontal line from the speed axis touches the innermost cost
contour it can. The speed–frequency relationship for walking on a treadmill at
able to contort their bodies in registry with the swinging motion
various speeds (red line) will be a series of such horizontal contours. Step frequency to add momentum (and energy) to the swing; this is usually called
(dark blue line) and step length (green line) cost minimisations will be determined ‘pumping’ (Wirkus et al., 1998). However, the height of the swing
in a similar manner for restricted frequency or step length conditions, except that can be increased in this way only so far. As the swing moves be-
the line determining each will intersect the cost surface in a different location (and
yond horizontal, a small jolt is felt on the down swing, following
therefore determines a different optimisation solution).
which the height of the next swing is much reduced (indicating
centres play a critical role in operating the locomotory apparatus, that energy has been extracted from the system). The jolt that is
but this role should be considered in the same way as the critical felt indicates a small collision caused by the mass of the individual
role of the driver in operating a motor vehicle. The driver is capable falling within the arc of the swing and being forced to change
of steering the vehicle anywhere, but a good driver remains on the direction (a discontinuity in its path) as the supporting cables of
roadway and operates according to the traffic regulations. To do the swing become taut. The energy loss occurs because the radial
otherwise results in less than optimum performance. When component of the falling centre of mass (CoM) velocity is directed
considering the best operation of the limbs in animal locomotion, along the supporting cables and is lost to the collision interaction
a major challenge is to discover the ‘road map’ and ‘driving (Fig. 2).
regulations’ that the central control centre must integrate with; A common acrobatic circus act involves demonstrating that a
these will be dependent on the physical interaction of the full 360° swing revolving completely around the rotational axis
organism with its environment and the energetic consequences can be accomplished. Understanding how this can be done in a cir-
of that interaction. cus, when it cannot be done in the playground, indicates the value
Understanding why footfall patterns appear as they do requires of considering the phenomena responsible for energy loss, since it
an understanding of the energetic consequences of the motions affects functional capability so directly. Success in the full rotation
and comparison of these with available options. This may appear of the circus swing depends on a single adjustment of equipment
to be a difficult task, given that mammals are such complex sys- (a change of ‘morphology’), replacing the flexible supporting cables
tems, using complex biological actuators (muscles) organised in of- of the playground swing with rigid struts that do not flex when the
ten redundant functional networks (thus providing multiple individual swings higher than the horizontal. With rigid struts, the
control options for any given movement and apparent ‘cost’). How- CoM is prevented from falling vertically, so is forced to remain on
ever, substantial insight into the purpose of gaits can be gained by the swing arc regardless of position in the rotation. Without a ver-
identifying the main determinants of locomotion cost. tical fall, there is no jolt and therefore no collision loss, so the en-
ergy state of the system can increase continuously until enough
has accumulated to allow the mass to move completely around
Energetic consequences of gait the axis. In the normal playground swing, the higher the swing
above horizontal, the greater the jolt and the greater the loss.
As much as we are all familiar with the fact that locomotion
using legs involves substantial energetic cost, it is not as apparent
why this must be so. Flying and swimming require much less met- Deflecting the centre of mass in legged locomotion
abolic investment per distance travelled (Tucker, 1970). The high
cost of terrestrial locomotion is most commonly attributed to the Gaits differ, but there are features that all the gaits have in com-
negative work expended during the stride cycle when moving on mon. Most obviously, they all involve moving the mass of the ani-
a solid surface. It is possible to observe and verify that negative mal across the substrate. Cyclic vertical motions occur while the
work occurs in legged locomotion, where the kinetic energy of animal moves forward at speeds that vary systematically over
the organism’s moving mass is expended as work on the muscles the period of the stride. To a much lesser extent, lateral movements
of the animal’s limbs instead of the muscles moving the mass of also occur. The interaction of the forward and vertical movement is
 J.E.A. Bertram / The Veterinary Journal 198 (2013) e3–e8 e5

Fig. 2. Collision loss involved in a playground swing. If an individual on a playground swing swings above the horizontal, the centre of mass (CoM) will fall vertically. When
the supporting cables again become tight, the individual will experience a small jolt and the next swing will reach a much lower height. The small jolt extracts the radially
oriented velocity, and consequently that component of the kinetic energy. Deflecting the CoM when a limb contacts the ground surface results in a similar ‘collision’, with
losses determined by the geometry of the velocity and constraining force applied by the contact limb.

a key factor in determining the energetic cost of legged locomotion, sophisticated strategies to limit energy loss as much as possible,
so the remainder of this discussion will neglect the lateral motions. and these strategies camouflage many aspects of the processes in-
Although the mass of the animal continues travelling forward in volved. Recognising the factors involved, and the strategies avail-
locomotion, the cyclic changes in height of the CoM requires that able, to limit energy loss within the stride cycle provides
the vertical component of its direction change through the stride substantial insight into why many of the motions observed occur
cycle: from forward and upward to forward and downward, and as they do and why others that we might imagine do not.
from forward and downward to forward and upward. These The contact of each limb is much like the contact of an individ-
changes of CoM height are natural aspects of all gaits, but they ual spoke of a wheel that does not have a rim to roll on. The key
are key factors in determining the consequences of any gait loss occurs with each contact and is basically equivalent to the jolt
strategy. felt in the playground swing that travels too high; it is derived from
In steady state locomotion (constant average horizontal speed), the collision-like deflection of the CoM as the limb strut applies
changes in the height of the CoM as the animal moves forward force to the ground (along its axis), constrains the CoM path (to
mean that there are two important portions in each stride: (1) a vault over the contact point) and shifts the CoM direction from for-
portion of low energetic cost; and (2) a portion of high energetic ward and downward, prior to contact, to forward and upward. If
cost. The low cost portion occurs when the CoM direction changes the limb simply acted as a rigid strut, then locomotion would in-
from moving upward to moving downward. This direction change volve violent transitions and large amounts of energy would be lost
involves little energy cost to the animal because it occurs sponta- at each contact.
neously as a result of the passive action of gravity. The high cost Maintaining steady forward speed requires the replacement of
portion occurs when the CoM direction changes from moving all lost energy, so the cost of locomotion in this case would be very
downward to moving upward. This direction change must be med- high, involving both large energy losses and the cost of work re-
iated by the action of the limbs and involves high cost because it is quired to repay the loss. Instead, animals have developed strategies
during this phase that energy can be lost and/or work must be per- to limit this loss and to allow them to travel for relatively low over-
formed by the limb. all energetic investment. These strategies involve decreasing the
The footfall patterns that are recognised as gaits are employed loss involved in the high cost portion of the gait cycle, while
to allow for the downward to upward direction change of the increasing the travel accomplished in the low cost portion. There
CoM, while limiting the net work that is required to maintain con- is a trade-off between these strategies and the active work that
sistent average forward speed. Different footfall strategies (gaits) must be invested to implement them. An understanding of the
will suit different circumstances of locomotion, largely determined movement patterns that we define as gaits requires identifying
by the speed of forward travel. Understanding why each gait is the factors involved in the optimisation solution and recognising
used requires understanding how energy is lost from the system the features and consequences of the trade-offs involved.
and what strategies are available to minimise and replace that loss.
If animals could travel across the substrate with virtually no Reducing loss in the high cost portion of the cycle: Walking and
loss, like a wheel rolling, it would not be necessary to add much en- galloping
ergy to each step. It is easy to recognise that animals do not roll like
a wheel, but it is not as easy to identify why legged locomotion Although they appear to be quite different gaits, walking and
involves energy loss. This is because animals have developed galloping are both strategies that primarily limit the energy loss
e6 J.E.A. Bertram / The Veterinary Journal 198 (2013) e3–e8

associated with the transition in the CoM from moving forward but
downward to forward and upward (the high cost portion of the
stride cycle). Two aspects of this transition contribute to the ener-
getic cost. The first is the collision-based energy loss that occurs
when the limb contacts the substrate and deflects the animal’s
mass. This loss could be substantial, so it is not surprising that
strategies have been developed to limit the magnitude of loss.
Some of these strategies are subtle and non-intuitive, even though
they are effective.
In walking, collision-based loss occurs when the leading limb in
the transition process contacts the substrate and deflects the ani-
mal’s mass. This loss must be replaced in order to maintain con-
stant average speed; this requires thrust from the trailing limb.
However, the sequence of action of the limbs (deflection then
thrust vs. a pre-emptive thrust then deflection), substantially alters
the net energetic cost. This is because the geometry of the collision
event differs between the two alternatives, even if everything else
remains the same. This is most easily seen in the more straightfor-
ward geometry of bipedal walking (McGeer, 1990; Donelan et al.,
2002; Collins et al., 2005; Ruina et al., 2005; Kuo, 2007; Fig. 3A),
but the same process occurs in quadrupedal gaits (Birkemeier,
1998; Usherwood et al., 2007; Lee et al., 2011). The key factor to
recognise in understanding this strategy is that only that portion
of the CoM velocity vector aligned with the contact is involved in
the collision; for the playground swing, this was the radial portion
of the child’s total velocity, while in legged locomotion it is the
trigonometric portion aligned with the axis of the limb.
The collision loss associated with deflecting the CoM is largely
dependent on the geometric relationship between the two
Fig. 3. (A) A simple biped model indicates the value of considering collision energy
dynamic features of the motion that interact: the velocity vector loss. The transition from one support limb to the next requires a contact from the
representing motion of the animal’s mass and the vector represent- lead limb and a push-off thrust from the trailing limb. Centre of mass (CoM) velocity
ing the net ground reaction force (which contributes to the impulse vectors are shown, where V and V+ are just before and just after transition
responsible for altering the direction of travel of the CoM, as between support limbs, respectively. If contact deflection occurs first, followed by
push-off, CoM velocity will change as indicated by the grey vectors. The vector tips
described above). Since only the component of the two vectors that
will follow line a as the lead limb acts to change the CoM direction (note a is parallel
align can be involved in the collision interaction, keeping these to the lead limb), then line b as the previous stance limb adds thrust (b parallels the
vectors as close to perpendicular as possible limits collision loss effective previous stance limb, a direct line between the point of contact and the
(Lee et al., 2011). As a consequence, collision loss can also be CoM). If, instead, the previous limb adds thrust before the next limb makes contact,
reduced by making the deflection smaller. Interestingly, dividing the vector tips will follow paths d and e. Altering the sequence of contacts, and
nothing else, changes the momentum and energy loss because it alters the
a large deflection into a series of smaller sub-deflections, even relationship between the CoM velocity vector and the limbs at points in the cycle
when the sub-deflections sum to the magnitude of the large when substantial energy can be lost. (B) Dividing a single deflection (shown at left)
deflection, can substantially reduce overall loss. This is because into a series of sub-deflections results in an overall decrease in energy loss.
the relative energy loss is a function of the square of the deflection Asymmetric gaits, such as the equine canter and gallop, appear to use this strategy
to lower locomotion cost as higher speeds.
angle (Ruina et al., 2005; Fig. 3B). Even with more contacts
(collisions), the substantial reduction of loss in each contact
reduces the overall loss. This is the strategy used in the three-beat The CoM during running and trotting has substantial horizontal
canter and four-beat gallop (Bertram and Gutmann, 2009); the velocity. Much of this is in jeopardy of being lost to large, jolting
deflection in the canter requires three contacts (collisions), but collisions if the contact dynamics are not managed properly. One
each involves the cubed root of the energy of a single, large simple strategy employed in these gaits is to maintain the contact
collision. In the end, if x = the proportion of energy lost in a single limb(s) as upright as possible. If the limb made contact perpendic-
contact, using a sequenced footfall canter instead would require ular to the horizontal path of the CoM, then no collision loss would
only 3(x)1/3 = 1/3x. With an extremely large number of limbs (n) occur (the CoM velocity and ground reaction force vectors would
the proportional loss declines to zero (1/n)x; equivalent to a wheel be perpendicular and could not interact with each other in a colli-
with a rim. sion-like manner). However, this would limit the ability to apply
the upward thrust that allows the ballistic flight phase of the gait;
the flight phase is a very low cost portion of the stride cycle, so it is
Increasing the low cost portion of the cycle: Running, trotting advantageous to emphasise this phase. One of the trade-offs in
and hopping managing a cost effective trot is to decrease the contact loss, while
increasing the low cost flight phase. Of course, these are mutually
Bipedal running, kangaroo hopping and the trot used by horses exclusive options, but the gait observed emerges as a compromise
and dogs all employ a series of discrete contacts with an interven- that optimises the net cost.
ing non-contact, or ‘flight’, phase. Since there is virtually no overlap Trots, runs and high speed hopping are also characterised by a
between contacts as in the walk, there is no opportunity to distinctive bouncing motion. The supporting limb(s) deflect during
decrease loss by applying a pre-emptive thrust prior to the next contact and the CoM is at its lowest point at mid-contact, following
contact. However, there are strategies available that allow these which it accelerates upward into the non-contact phase of the cy-
gaits to be energetically cost effective (explaining why they so cle. The obvious bounce-like motion of these gaits, paralleling the
commonly appear). bouncing of a rubber ball, has led to the expectation that these
 J.E.A. Bertram / The Veterinary Journal 198 (2013) e3–e8 e7

gaits depend on elastic recoil (the storage and return of elastic loss (depending on the strategy implemented), or at the very least
strain energy in limb tendons, ligaments and muscular compo- must replace the loss in order to maintain motion. The work per-
nents; Cavagna et al., 1977). However, it turns out that the bounc- formed by the leg can be considered in terms of either stance leg
ing motion is effective at managing collision costs and bounce-like work, altering the travel of the CoM while the limb is in contact
running gaits are energetically cost-effective, even if all of the with the substrate, or swing limb work, adjusting the position of
forces involved are actively generated and so have an active cost limb when it is not in contact (to control step length and fre-
(Ruina et al., 2005; Srinivasan and Ruina, 2005). quency). Although many details of these trade-offs/optimisations
Passive elastic structures also contribute to reducing the cost of remain to be determined for specific species, it is valuable to have
running and trotting (Alexander, 1984) and they do so without a conceptual understanding of the system.
interfering with the collision loss saving. The motion required to As discussed above, several strategies are available to reduce
limit collision loss during contact, smooth reversal of vertical mo- collision-based loss (e.g. altering the sequence of contact and
tion, and that of passively deforming elastic tissues is the same thrust, as described above). For a gait such as the walk, collision
(Ruina et al., 2005). The way animals trot and humans run is of the CoM with the substrate (via contact of a stiff stance limb)
mechanically (and metabolically) cost effective, even without pas- could be eliminated by allowing the stance limb to bend and ex-
sive elasticity, but passive elasticity can be used to enhance the tend so the CoM travels on a path parallel to the substrate. How-
cost-effectiveness of the gait because the best leg deformation ever, such a strategy would require more work from the stance
strategy for avoiding loss is coincident with the best leg deforma- limb bending and extending under the load of the mass than would
tion strategy for strain energy recovery. It has always been prob- be saved by eliminating the collision-based loss (Srinivasan and
lematic to envision the adaptations required to produce the Ruina, 2007; Srinivasan, 2011). Both the excessive loss of a jolting
elegantly cursorial forms that exploit elastic recoil for effective contact using a stiff, strut-like limb (and the work required to re-
locomotion. If the running gait were primarily dependent on effec- place that loss) and the work required to eliminate that loss by
tive elastic recovery, we would wonder how the elastic mecha- bending and extending the limb (for a completely level motion of
nisms could be developed prior to the origin of a gait that could the CoM) are greater than a compromise that exists between these
exploit them. Instead, from understanding the energy changes two extremes. The compromise optimises the combined decrease
within a running stride, it turns out that the bouncing gait is a in loss with the cost of stance limb work to find a CoM path that
cost-effective option, even without passive elasticity. Because varies in height in a manner that requires the least energetic
these organisms already possessed connective tissues with proper- investment. The jolting ‘collision’ is diminished to the point where
ties that could augment these optimised motions, adaptive pro- it is difficult to detect, while the CoM vertical oscillations are rela-
cesses were then free to modify the passive components to work tively small, although not eliminated entirely.
with bouncing gaits to provide a true ‘bounce’, as suited to the cir- Collision-based losses could also be virtually eliminated by tak-
cumstances of each particular species. ing very small steps. If a step is small enough, the CoM falls very little
between contacts, so very little loss is involved. However, to travel at
a reasonable speed with such small steps would require substantial
Trade-offs that optimise gait active swinging of the non-stance limb, in order to reposition it
appropriately (Kuo, 2001). The numerous swing limb accelerations
Gaits and the manner that animals use their limbs for locomo- and decelerations would require greater energetic investment than
tion are best considered as an optimisation problem, where many is saved by totally eliminating the collision loss through small step
features of the motion patterns may result in either positive or length. Optimisation models indicate that a minimum cost strategy
negative effects. The trade-off described above between strategies exists somewhere between the long, jolting steps that involve little
that decrease the energetic loss in the high cost portion of the gait swing limb cost, but large collision loss, and short, swift steps that
cycle (a ‘positive’) being tied to a decrement of the low cost portion decrease collision loss, but require substantial swing limb work.
(a ‘negative’), is an example. The best strategy then, and the one
used by animals while moving across a surface, is a compromise
Conclusions
that results in the best overall performance. To understand and
appreciate how gaits and the pattern of limb movements are deter-
The normal gaits we observe in animals largely represent the
mined, it is necessary to identify the factors that influence the opti-
optimisation solution best suited to the circumstances the animal
misation process. A large number of factors will be involved, of
faces. The similarity of gaits between individuals of a given species
course, but substantial insight can be gained by identifying the
indicates simply that each individual is ‘solving’ the same
three most dominant factors; an important but neglected energy
functional problem. Various biological and control factors may
loss and two inter-dependent energetic costs associated with work
influence the details of these gaits, but it is useful to understand
performed by the limb.
the functional ‘purpose’ of the gait in order to interpret how such
The discussion can become complex, even when restricted to
factors affect locomotion and to anticipate the consequences of
the interplay between just three factors. It would be surprising if
variation among animals and circumstances. Although we are still
this were not the case, since locomotion is itself a complex dy-
a long way from quantifiably interpreting (and predicting) the
namic process. The challenge is to recognise that negative effects
energetic consequences of specific movement strategies or the
will be eliminated as much as possible and so will become difficult
influence of any morphological variation, eventual success at these
to observe and identify. Such a decrease in negative effects comes
challenges will depend on understanding the fundamental com-
at the expense of energetic cost (in the form of work performed)
promises that define the range of effective movement strategies
that is required to implement the available strategies. The gait
available to the organism.
and the details of its implementation will largely result from the
best trade-off (balance) between reducing the negative effects,
increasing the positive effects and generating the effort needed Conflict of interest statement
to accomplish these tasks.
The basic trade-offs involved in legged locomotion can be sum- The author of this paper has no financial or personal relation-
marised as (1) the collision-based energy loss caused by limb con- ship with other people or organisations that could inappropriately
tact; and (2) the cost associated with leg work that can reduce that influence or bias the content of the paper.
e8 J.E.A. Bertram / The Veterinary Journal 198 (2013) e3–e8

References Lee, D.V., Bertram, J.E.A., Anttonen, J., Ros, I.G., Harris, S.L., Biewener, A.A., 2011. A
collisional perspective on quadrupedal gait dynamics. Journal of the Royal
Society Interface 8, 1480–1486.
Abourachid, A., 2003. A new way of analysing symmetrical and asymmetrical gaits
Marey, E.-J., 1873. La Machine Animale: Locomotion Terrestre et Aeriemme.
in quadrupeds. Comptes Rendus Biologies 367, 625–630.
Baillière, Paris, France.
Alexander, R.M., 1984. Elastic energy stores in running vertebrates. American
McCrae, D.A., Rybak, I.A., 2008. Organization of mammalian locomotor rhythm and
Zoologist 24, 85–94.
pattern generation. Brain Research Reviews 57, 134–146.
Bertram, J.E.A., 2005. Constrained optimization in human walking: Cost
McGeer, T., 1990. Passive dynamic walking. International Journal of Robotics
minimization and gait plasticity. Journal of Experimental Biology 208, 979–991.
Research 9, 68–82.
Bertram, J.E.A., Gutmann, A., 2009. The running horse and cheetah revisited:
Robilliard, J.J., Pfau, T., Wilson, A.M., 2007. Gait characterization and classification in
Fundamental mechanics of two galloping gaits. Journal of the Royal Society
horses. Journal of Experimental Biology 210, 187–197.
Interface 6, 549–559.
Ruina, A., Bertram, J.E.A., Srinivasan, M., 2005. A collisional model of the energetic
Bertram, J.E.A., Ruina, A., 2001. Multiple walking speed–frequency relations are
cost of support work qualitatively explains leg sequencing in walking and
predicted by constrained optimization. Journal of Theoretical Biology 209, 445–
galloping, pseudo-elastic leg behavior in running and the walk-to-run
453.
transition. Journal of Theoretical Biology 237, 170–192.
Birkemeier, M.D., 1998. Modeling the dynamics of quadrupedal running.
Self, Z.T., Spence, A.J., Wilson, A.M., 2012. Speed and incline during Thoroughbred
International Journal of Robotics Research 17, 971–985.
horse racing: Racehorse speed supports a metabolic power constraint to incline
Cavagna, G.A., Heglund, N.C., Taylor, C.R., 1977. Mechanical work in terrestrial
running but not to decline running. Journal of Applied Physiology 113, 602–607.
locomotion: Two basic mechanisms for minimizing energy expenditure.
Snaterse, M., Ton, R., Kuo, A.D., Donelan, J.M., 2011. Distinct fast and slow processes
American Journal of Physiology: Regulatory, Integrative and Comparative
contribute to the selection of preferred step frequency during human walking.
Physiology 233, R243–R261.
Journal of Applied Physiology 110, 1682–1690.
Collins, S., Ruina, A., Tedrake, R., Wisse, M., 2005. Efficient bipedal robots based on
Srinivasan, M., 2011. Fifteen observations on the structure of energy-minimizing
passive-dynamic walkers. Science 307, 1082–1085.
gaits in many simple biped models. Journal of the Royal Society Interface 8, 74–
Donelan, J.M., Kram, R., Kuo, A.D., 2002. Mechanical work for step-to-step
98.
transitions is a major determinant of the metabolic cost of human walking.
Srinivasan, M., Ruina, A., 2005. Computer optimization of a minimal biped model
Journal of Experimental Biology 205, 3717–3727.
discovers walking and running. Nature 439, 72–79.
Farquharson, A.S.L., 2007. On the Gait of Animals by Aristotle. http://
Srinivasan, M., Ruina, A., 2007. Idealized walking and running gaits minimize work.
etext.library.adelaide.edu.au/a/aristotle/gait/complete.html (accessed 15 April
Proceedings of the Royal Society A: Mathematical, Physical and Engineering
2013).
Sciences 463, 2429–2446.
Goiffon, G.-C., Vincent, A.-F., 1779. Memoire Artificielle des Principes Relatifs a la
Tucker, V.A., 1970. Energetic cost of locomotion in animals. Comparative
Fidele Representation des Animaux Tant en Peinture, qu’en Scultpture. I Partie
Biochemistry and Physiology 34, 841–846.
Concernant le Cheval. Alfort, France.
Usherwood, J.R., Williams, S.B., Wilson, A.M., 2007. Mechanics of dog walking
Grillner, S., 1975. Locomotion in vertebrates: Central mechanisms and reflex
compared with a passive, stiff-limbed, 4-bar linkage model, and their collisional
interaction. Physiological Reviews 55, 247–304.
implications. Journal of Experimental Biology 210, 533–540.
Hildebrand, M., 1965. Symmetrical gaits of horses. Science 150, 701–708.
Wilson, A.M., Pardoe, C.H., 2001. Modification of a force plate system for equine gait
Hoyt, D.F., Taylor, C.R., 1981. Gait and the energetics of locomotion in horses. Nature
analysis on hard road surfaces: A technical note. Equine Veterinary Journal
292, 239–240.
Supplement 33, 67–69.
Kuo, A.D., 2001. A simple model of bipedal walking predicts the preferred speed-
Wirkus, S., Rand, R., Ruina, A., 1998. How to pump a swing. College Mathematics
step length relationship. Journal of Biomechanics 123, 264–269.
Journal 29, 266–275.
Kuo, A.D., 2007. The six determinants of gait and the inverted pendulum analogy: A
dynamic walking perspective. Human Movement Science 26, 617–656.