A three-leg model producing tetrapod and tripod coordination

J Neurophysiol 115: 887–906, 2016.
First published November 18, 2015; doi:10.1152/jn.00693.2015.
A three-leg model producing tetrapod and tripod coordination patterns of
ipsilateral legs in the stick insect
T. I. Tóth and S. Daun-Gruhn
Department of Animal Physiology, Institute of Zoology, University of Cologne, Cologne, Germany
Submitted 14 July 2015; accepted in final form 10 November 2015
three-leg model; interleg coordination; locomotion; sensory feedback
LOCOMOTION OF INSECTS, IN
particular that of the stick insect, is
a well-studied field of animal physiology. In the studies, a
number of different approaches have been used. The spectrum
reaches from behavioral observations (e.g., Hughes 1952; Wilson 1966; Graham 1972, 1985; Delcomyn 1981), over electrophysiological methods (e.g., Bässler 1977, 1983; Laurent and
Burrows 1989a,b; Büschges 1995, 2005; Akay et al. 2007;
Büschges and Gruhn 2008; Borgmann et al. 2007, 2009, 2011;
Schmitz 1986a,b), to modeling work (e.g., Cruse 1980, 1990;
Cruse et al. 1998; Dürr et al. 2004; Ekeberg et al. 2004;
Daun-Gruhn 2011; Daun-Gruhn and Tóth 2011; Daun-Gruhn
et al. 2011; von Twickel et al. 2011, 2012; Schilling et al. 2013;
Tóth et al. 2015). The models cited here reflect various aspects
of insect locomotion and reach from one-leg models (Ekeberg
et al. 2004; Daun-Gruhn et al. 2011; von Twickel et al. 2011;
Tóth et al. 2012; Knops et al. 2013) to six-leg models (Cruse
1980 1990; Cruse et al. 1998; Dürr et al. 2004; von Twickel et
Address for reprint requests and other correspondence: S. Daun-Gruhn,
Dept. of Animal Physiology, Institute of Zoology, Univ. of Cologne Zülpicher
Strasse 47b, D-50674 Cologne, Germany (e-mail: [email protected]).
www.jn.org
al. 2012; Schilling et al. 2013). Thus Ekeberg et al. (2004), for
example, study and simulate the mechanisms of coordination
first in a restricted middle leg of the stick insect. They then
examine to what extent the mechanisms found in the middle
leg can, perhaps, be adapted to apply to the stepping of front
and the hind leg, too. On the other end of complexity, Cruse
(1980, 1990), Cruse et al. (1998), Dürr et al. (2004), and
Schilling et al. (2013) have created, in a steadily evolving
process, an artificial neural network that can emulate various
kinds of six-leg walking of insects.
Thus a large body of knowledge on this topic has been
created over the years. In spite of the progress made, a number
of important details of the intra- and interleg coordination
during locomotion (normal walking) remain unclear. Our aim
in the work presented here has been to introduce a modeling
approach that is capable of grasping complex interactions of
the local (neuronal) mechanisms of the legs during walking.
More precisely, we endeavored to establish an intersegmental
coupling mechanism between the ipsilateral legs that could
produce the patterns of normal walking ipsilaterally. To
achieve this, we made use of an approach that was different
from the aforementioned ones. Our models consisted of local
neuronal networks that controlled the function of an antagonistic muscle pair at a leg joint (e.g., protractor-retractor at the
thorax-coxa joint). Even more importantly, each such network
was constructed using anatomical data and experimental results, where they were available (Bässler 1983; Büschges 1995,
1998, 2005; Westmark et al. 2009), and physiologically reasonable assumptions to fill in the gaps, where data were not
available. Our model was thus created by connecting these
local control networks by excitatory or inhibitory sensory
pathways whose existence seems to have, at least indirectly,
been justified by experiments.
In our earlier work, we constructed models of the neuromuscular system of an insect leg, in particular of a (middle) leg of
the stick insect: Tóth et al. (2012), Knops et al. (2013), and
Tóth et al. (2013a, b). The models included two or all of the
three antagonistic muscle pairs (m. protractor and retractor
coxae, m. levator and depressor trochanteris, and m. extensor
and flexor tibiae) at the three main joints: the thorax-coxa
(ThC), the coxa-trochanter (CTr), and the femur-tibia (FTi)
joint. Each muscle pair in the models possessed its dedicated
local neuronal control network. The structure and function of
these local networks were based on experimental findings and
physiologically reasonable assumptions (Borgmann et al.
2011; Daun-Gruhn et al. 2011; Tóth et al. 2012; Knops et al.
2013). Using these one-leg models, we constructed a model of
three ipsilateral legs with suitable intersegmental coupling
between them as described in the present work. Note that the
0022-3077/16 Copyright © 2016 the American Physiological Society
887
Downloaded from http://jn.physiology.org/ by 10.220.32.247 on June 17, 2017
Tóth TI, Daun-Gruhn S. A three-leg model producing tetrapod
and tripod coordination patterns of ipsilateral legs in the stick insect.
J Neurophysiol 115: 887–906, 2016. First published November 18,
2015; doi:10.1152/jn.00693.2015.—Insect locomotion requires the
precise coordination of the movement of all six legs. Detailed investigations have revealed that the movement of the legs is controlled by
local dedicated neuronal networks, which interact to produce walking
of the animal. The stick insect is well suited to experimental investigations aimed at understanding the mechanisms of insect locomotion.
Beside the experimental approach, models have also been constructed
to elucidate those mechanisms. Here, we describe a model that
replicates both the tetrapod and tripod coordination pattern of three
ipsilateral legs. The model is based on an earlier insect leg model,
which includes the three main leg joints, three antagonistic muscle
pairs, and their local neuronal control networks. These networks are
coupled via angular signals to establish intraleg coordination of the
three neuromuscular systems during locomotion. In the present threeleg model, we coupled three such leg models, representing front,
middle, and hind leg, in this way. The coupling was between the
levator-depressor local control networks of the three legs. The model
could successfully simulate tetrapod and tripod coordination patterns,
as well as the transition between them. The simulations showed that
for the interleg coordination during tripod, the position signals of the
levator-depressor neuromuscular systems sent between the legs were
sufficient, while in tetrapod, additional information on the angular
velocities in the same system was necessary, and together with the
position information also sufficient. We therefore suggest that, during
stepping, the connections between the levator-depressor neuromuscular systems of the different legs are of primary importance.
888
INTERLEG COORDINATION
Fig. 1. Schematic illustration of the main joints (segments) of a stick insect leg.
The Greek letters ␣, ␤, and ␥ denote the retraction, the levation, and the flexion
angles, respectively. [Adapted with permission from M. Gruhn, unpublished
observations.].
model created thus is not simply a triplication of the previous
one-leg model but it reaches a higher level of complexity by
the intersegmental connections that represent excitatory or
inhibitory sensory pathways.
In this paper, we report simulation results with this three-leg
model. The model can successfully simulate the characteristic
coordination patterns that occur during walking of the stick
insect (so-called tripod and tetrapod coordination patterns) and
the transitions between them with satisfactory accuracy. The
biological relevance of intersegmental connections implemented in the model will be discussed. We will also discuss
some implications of our results for the locomotion of other
legged species.
METHODS
The three-leg model to be presented here is based on the one-leg
models (Daun-Gruhn et al. 2011; Tóth et al. 2012; Knops et al. 2013).
In Fig. 1, the schematic picture of a stick insect (middle) leg is
Meso−thorax (ML)
PR control network
g
g
d13
d14
LD control network
EF control network
g
g
MN
IN13
C49
IN14
C50
C7
C8
g
MN7P
C25
MN8R
C26
MN
g
MN
IN17
C53
IN18
C54
C9
C10
MN10L
C28
MN
g
MN
IN15
C51
β
IN21
C57
IN22
C58
C11
C12
MN12E
C30
MN
CPG
Lev. m.
IN19
C55
d22
g
MN11F
C29
CPG
Dep. m.
Ret. m.
g
d21
g app11 g app12
g
MN9D
C27
CPG
Pro. m.
IN16
C52
d18
g app9 g app10
g app7 g app8
g
g
d17
IN20
C56
Flex. m.
Ext. m.
γ
IN23
C59
IN24
C60
g d20
Fig. 2. The neuromuscular model of 1 (middle) leg. Local networks (PR, LD, and EF as indicated) controlling the activity of three antagonistic muscle pairs:
m. protractor and retractor coxae (Pro. m. and Ret. m.), m. levator and depressor trochanteris (Lev. m. and Dep. m.), and m. extensor and flexor tibiae (Ext. m.
and Flex. m.). Central pattern generator (CPG) in each of these networks: central pattern generator consisting of a pair of mutually inhibitory nonspiking neurons
(e.g., C7–C8 in the PR control network); gapp,7, gapp,8, etc.: (central) input to the CPG neurons (individually variable). MN7P, MN8R, etc.: motoneurons driving
the corresponding muscles; gMN: uniform excitatory input to the motoneurons. IN13-IN14, etc.: premotor interneurons; gd13, gd14, etc.: (individually variable)
inhibitory inputs to these interneurons. IN15-IN16, etc.: interneurons conveying sensory signals to the corresponding CPG neurons from the other local networks.
Hexagons with ␤ or ␥ in them: sources of sensory signals encoding position, load, and ground contact corresponding to a critical value of ␤ (see text). Other
symbols: empty triangles: excitatory synapses; filled circles: inhibitory synapses. Arrows from muscles to the hexagons symbolize that the sensory signals arise
because of mechanical movement due to muscle activity. Note that the neurons in this network are numbered such that they fit in the larger model that contains
all 3 ipsilateral legs.
J Neurophysiol • doi:10.1152/jn.00693.2015 • www.jn.org
Downloaded from http://jn.physiology.org/ by 10.220.32.247 on June 17, 2017
displayed. It shows the three most important leg joints and segments: the thorax-coxa joint that enables pro- and retraction of the
leg in the horizontal plane, the coxa-trochanter joint for leg
movements in the vertical plane, and the femur-tibia joint for
flexion and extension of the tibia of the leg. The three angles ␣, ␤,
and ␥ that quantitatively describe these movements are also shown
in this figure. Thus ␣ is associated with the protraction-retraction
movements (increasing with the retraction of the leg), ␤ with the
levation-depression movements of the leg (increasing with levation),
and ␥ with the extension-flexion movements of the leg (increasing
with flexion). In Fig. 2, the neuromuscular model of one (middle) leg
is displayed. It consists of three interconnected local networks, as
indicated, that control the activity of three antagonistic muscle pairs:
m. protractor and retractor coxae (Pro. m. and Ret. m.), m. levator and
depressor trochanteris (Lev. m. and Dep. m.), and m. extensor and
flexor tibiae (Ext. m. and Flex. m.). Accordingly, the corresponding
control networks are called PR, LD, and EF control network. The core
unit of each local control network is a central pattern generator
(denoted as CPG). The existence of these local control networks is
experimentally well established (Büssler 1983; Büschges 1995, 2005).
Each CPG consists of a pair of mutually inhibitory nonspiking
neurons (e.g., C7–C8 in the PR control network). The CPG neurons
receive (central) excitation, denoted by the conductances of the
excitatory currents (gapp,7, gapp,8, etc.). These inputs are individually
variable for each CPG neuron. Occasionally, strong inhibitory synapses suppressing the excitatory effect may become active at these site
for short periods (see below). The CPG neurons also receive input
from one of the other local control networks conveyed by excitatory
(IN16, IN20, IN24) and inhibitory (IN15, IN19, IN23) interneurons.
Motoneurons (represented by boxes) drive the corresponding muscles,
e.g., MN7P means protractor motoneuron, and it drives the m.
protractor coxae. The motoneurons receive uniform central drive
denoted gMN by its conductance. The CPG and the motoneurons are
connected via inhibitory premotor interneurons (IN13, IN14, etc.)
(Büschges 1998, 2005; Westmark et al. 2009), which in turn receive
individually variable central inhibitory inputs (gd13, gd14, etc.). The
local control networks are connected by means of sensory signals
expressing the vertical position of the femur, thus capable of signaling
INTERLEG COORDINATION
ground contact, and the position of the tibia relative to the femur. The
former is represented by the vertical angle ␤, the latter by the angle ␥
(hexagons with ␤ or ␥ in them). At critical values of ␤ (ground
contact) and ␥, the conductances of IN16, IN24, and IN20, respectively, are strongly reduced or restored to their normal (high) values.
This property of the model ensures that it can exhibit intrinsic
oscillations, if the critical values of the ␤ and ␥ are chosen appropriately (see RESULTS). Here, intrinsic means that the oscillations are
generated by the interaction of the intraleg LD and EF local networks.
Now, taking three copies of this one-leg model, we obtain a
three-leg model, for the time being with no interleg connections (Fig.
3). Note that the EF local control network of the hind leg is different
from those of the other two legs: the CPG neurons (C17–C18) and
premotor interneurons (IN33-IN34) are “cross-connected” compared
with the connection between the same type of neurons in the other two
EF local control networks. This is necessary to account for the fact
that the hind leg extends during the stance phase in contrast to the
other two legs, which carry out flexions during the stance phase.
PR control network
LD control network
EF control network
g
g
g
g
d1
d2
MN
IN1
C37
IN2
C38
C1
C2
MN2R
C20
MN
g
MN
IN5
C41
IN6
C42
C3
C4
MN
β
IN3
C39
MN
C5
C6
g
MN5F
C23
MN6E
C24
Flex. m.
IN7
C43
IN11
C47
IN12
C48
LD control network
EF control network
g
g
g
g
d13
d14
g
d17
d18
g app9 g app10
IN13
C49
IN14
C50
C7
C8
MN8R
C26
MN
g
MN
IN17
C53
IN18
C54
C9
C10
MN10L
C28
MN
g
MN
IN22
C58
C11
C12
g
MN11F
C29
IN16
C52
Flex. m.
Lev. m.
β
IN15
C51
MN12E
C30
MN
CPG
Dep. m.
Ret. m.
d22
IN21
C57
CPG
CPG
Pro. m.
g
d21
g app11 g app12
g
MN9D
C27
MN
Ext. m.
γ
IN8
C44
g d8
PR control network
g
MN7P
C25
g
Lev. m.
g app7 g app8
g
MN
IN10
C46
CPG
Dep. m.
Ret. m.
IN4
C40
Meso−thorax (ML)
MN4L
C22
IN9
C45
CPG
CPG
Pro. m.
d10
g app5 g app6
g
MN3D
C21
g
d9
g app3 g app4
g
MN1P
C19
d6
Ext. m.
γ
IN20
C56
IN19
C55
IN23
C59
IN24
C60
g d20
PR control network
g
g
d25
d26
g
MN
IN25
C61
IN26
C62
C13
C14
g
MN13P
C31
M14R
C32
MN
g
MN
g
g
d29
d30
IN29
C65
IN30
C66
C15
C16
MN16L
C34
MN
g
MN
IN27
C63
β
IN33
C69
IN34
C70
C17
C18
MN18E
C36
MN
CPG
Lev. m.
IN31
C67
d34
g
MN17F
C35
CPG
Dep. m.
Ret. m.
g
d33
g app17 g app18
g
MN15D
C33
CPG
Pro. m.
IN28
C64
EF control network
g
g app15 g app16
g app13 g app14
Meta−thorax (HL)
LD control network
IN32
C68
Flex. m.
Ext. m.
IN35
C71
IN36
C72
γ
g d32
Fig. 3. The 3-leg model still with no inter-leg connections in it. Note the special connection between the CPG and premotor neurons in the EF local control
network of the hind leg. This connection ensures that the hind leg extends during the stance phase in contrast to the other two legs in which flexion takes place
during the same phase of the stepping movement.
J Neurophysiol • doi:10.1152/jn.00693.2015 • www.jn.org
Downloaded from http://jn.physiology.org/ by 10.220.32.247 on June 17, 2017
g
g
d5
g app1 g app2
Pro−thorax (FL)
889
890
INTERLEG COORDINATION
In the following, we will show how the three copies of the one-leg
model can be connected to obtain the characteristic coordination
patterns of stick insect walking (tripod and tetrapod). For the sake of
simplicity, we did not make any adjustment of the mechanical properties of the legs. Thus all of them uniformly had those of a middle
leg. Also, the angular ranges of the protractor-retractor movements
were left to be the same for all legs. Finally, we want to emphasize
that our model is a kinematic one. Gravitation therefore does not play
any role in the further considerations, and the load signals to which we
will occasionally refer remain abstract. Although this seems a substantial constraint on the model, several studies (e.g., Schmitz et al.
2008; Schilling et al. 2013; Knops et al. 2013) have demonstrated that
a kinematic simulation of insect walking is a good approximation of
the real one.
RESULTS
A Short Description of the Tripod and Tetrapod
Coordination Patterns
Figure 4 shows the tetrapod and the tripod coordination
patterns as found in the experiments (Graham 1972). The
diagrams show whether an individual leg (L1–R3) is lifted at
any given time, time being the horizontal axis. The black spots
are the swing phases, that is when the corresponding legs are
lifted. During tetrapod two legs are lifted at the same time
while the other four are on the ground. From the three ipsilateral legs, only one is lifted at any time. For example, the right
hind leg (R3) and the left middle leg (L2) are simultaneously
lifted. In another variant of tetrapod, the right hind leg (R3) is
lifted together with the left front leg (L1). The phase differences between the six legs, within a step cycle, are illustrated
in the in Fig. 4, right, for the former case. The legs are
represented by circles, and legs having the same phase are
connected. Note a characteristic property of the tetrapod coor-
Intraleg Coordination During Stepping of a Single Leg
As a first result, we show the angular movements of the
femur and tibia in each of the three legs when there is no
interleg connection between them but the intraleg connections
are intact. At the same time, this result will also indicate the
need for intersegmental coordination to produce walking. The
simulation results show that the movements are intrinsic periodic oscillations of the angles in each of the three legs, as
displayed in Fig. 5. In the absence of intersegmental coordination, all legs can, however, be simultaneously in the air,
which clearly does not happen during walking. On the other
hand, coordinated movements of the femur and the tibia bring
about periodic stepping of each individual (simulated) leg. The
special role of the hind leg can clearly be seen in Fig. 5. Here,
the extension-flexion movements are complementary to those
of the two other legs.
Simulation results in Fig. 6 demonstrate that a permanently
weak connection between the local control networks does not
admit intrinsic oscillations in the leg. Interestingly, a permanently strong intraleg coupling, too, has the same effect (Fig.
6). Taking the middle leg as an example, the strength of
coupling is expressed by the value of gd20, the conductance of
the excitatory synaptic current from the EF control network to
the LD control network of the mesothoracic ganglion. This
value is, during oscillations, determined by the actual value of
the angle (cf. Fig. 2). Setting gd20 permanently to a low value
Tetrapod
2/3
1/3
1/3
0
0
2/3
0
1/2
1/2
0
0
1/2
Tripod
Fig. 4. Experimental data on the tetrapod and tripod coordination patterns in the stick insect (Graham 1972). The horizontal axis is time. The 6 legs are listed
vertically. L1 denotes the left front leg, L2 the left middle leg, and L3 the left hind leg. The notation is analogous for the right legs (R1–R3). The diagrams show
the state (swing or stance phase) of the legs at any given time. The black spots represent the swing phases during walking, and the white spaces are stance phases.
The main tetrapod diagram shows that during this coordination pattern two contralateral legs are lifted at the same time. The panel actually shows 2 variants of
tetrapod. (They are divided by a vertical dashed line.) Note that in both variants, the swing phases of the 3 ipsilateral legs occur in the following order: hind,
middle and front leg. At right, the phase differences between the leg movements are illustrated for the first variant of tetrapod. The circles represent the legs,
and legs having the same phase are connected, which means that they are lifted simultaneously. The numbers are the phase differences within 1 step cycle with
respect to the right middle leg (R2). The length of the cycle is normalized to 1. The phase diagram of the second variant of tetrapod can be obtained from that
of the 1st variant by exchanging the sides left and right. In the tripod coordination pattern, 3 legs are simultaneously lifted at any instant of time, as it can be
seen in the main panel for tripod. Note also a characteristic property of the tripod coordination pattern: the front and the hind leg on the same side of the animal
are lifted simultaneously. The phase diagram is here simpler than in the tetrapod case: all phase differences are equal either to a half period of the step cycle
or are 0.
J Neurophysiol • doi:10.1152/jn.00693.2015 • www.jn.org
Downloaded from http://jn.physiology.org/ by 10.220.32.247 on June 17, 2017
First, we provide a short description of the characteristic
coordination patterns of stick insect walking: tetrapod and
tripod. Then, we will present our simulation results that mimic
them.
dination pattern: the sequence of stepping on one side of the
animal is hind leg, middle leg, and front leg within a step cycle.
The tripod coordination pattern is somewhat simpler to
describe. Here, three legs are lifted simultaneously at any time,
for example, the right middle leg (R2), the left hind leg (L3),
and the left front leg (L1). The phase differences between the
leg movements are either zero or a half of a full step cycle (see
Fig. 4, right). It is characteristic for the tripod coordination
pattern that the ipsilateral front and hind leg are always lifted
simultaneously and the middle leg always alone.
125
100
75
50
25
125
100
75
50
25
1000
891
Fig. 5. Autonomous oscillations of the 3 legs in the
model with no interleg connection but with intact
intraleg connections. FL, ML, HL: front, middle, hind
leg. Red line: protraction-retraction angle ␣ in range:
28° (anterior extremal position)-128° (posterior extremal position); green line: levation-depression angle
␤ in range: 30° (lowest position, ground contact)-60°
(highest position); blue line: extension-flexion angle ␥
in range: 45° (maximal extension)-110° (maximal
flexion). These coordinated oscillations produce periodic motion of the simulated leg. Note the difference
in the angular movement of the 1st 2 legs and of the
hind leg (bottom row).
2000
3000
4000
(gd20 ⱕ0.3 nS), independently of the value of ␥, stops the
oscillations in the middle leg (Fig. 6, top right). The same kind
of result is obtained when the strength of the LD¡EF connection is reduced (Fig. 6, top left). As mentioned, at a permanently strong (gd20 ⱖ4.0 nS) coupling, the oscillations also
cease (Fig. 6, bottom right and left). It is thus apparent that for
the intrinsic oscillations to take place in the leg, a switch
between weak and strong excitatory synaptic currents (between
low and high conductances) is necessary. Neither permanently
strong nor weak couplings between the LD and EF systems can
produce intrinsic oscillations, hence stepping movement, of the
leg. These considerations are valid for the other two legs, too
(with conductances gd8 and gd32, respectively). However, a
closer inspection of Fig. 6, top right (weak EF¡LD), and Fig.
6, bottom left (strong LD¡EF), reveals that after the oscillations cease, all three legs remain lifted: all three ␤-angles are at
their maximum (⬇60°). This is of course a state that does not
occur in stick insects. We will return to this problem later in
this paper.
Basic Principles of the Interleg Coordination
To carry out normal locomotion, i.e., walking, insects
must coordinate the movement of their legs. In the work
reported here, we studied the coordination of the ipsilateral
legs, only. This falls, of course, short of a full coordination
of all six legs but is a necessary condition for that. Accordingly, the coordinated movement of the ipsilateral legs must
be established as a precondition of the full coordination of
all legs during walking.
The basic constraint that determines the coordination is that
if one leg is lifted some other (one or two) legs must remain on
the ground. Thus sensing of ground contact, or the lack of it, in
each of the legs is crucial for the leg coordination. Sensing of
ground contact was implemented in our model implicitly, via
the usage of the angle ␤, which expresses the vertical position
of the femur. We defined a critical value of ␤ at which the
synaptic excitation to the “retractor” CPG neuron and the
inhibition of the “protractor” CPG neuron were greatly increased (cf. Fig. 2). Attaining this critical value of ␤ signaled
ground contact in the model. This started the retraction, i.e., the
5000
t (ms)
stance phase. We thus succeeded in integrating the load signal
into the angular signal ␤. Accordingly, the model will use this
angle as indicator to decide whether a leg is on the ground or
lifted and, depending on that, impose conditions on the ␤-angle
of the other legs. In this way, a primary role in the intersegmental coordination was assigned to ␤ in the model. Although
there is, at present, no direct experimental evidence for this in
the stick insect, strong evidence of this kind was found in the
lobster (Ayres and Davis 1977). We therefore deem the existence of such a system in the stick insect also possible, even
plausible.
Beside the position signals ␤, the corresponding angular
velocities d␤/dt may also play a part in shaping the stepping
movement. These signals provide information on whether the
leg is moving upwards (d␤/dt ⬎ 0) or downwards (d␤/dt ⬍ 0)
or is moving slowly (|d␤/dt| ⬍ ⑀, ⑀ being a small number). The
actual values of the angles ␤ and of their angular velocities
supply the information needed for the desired type of interleg
coordination.
The activity of the motoneurons in the LD local control
network, hence the muscle forces generated by it, determine
the actual values of the ␤-angles and their angular velocities. In
turn, these networks will be interconnected in a suitable way,
as the above considerations imply. Moreover, the CPG of the
LD local control network is the ultimate source of the periodic
vertical movement of the femur. It appears therefore to be
reasonable to have direct synaptic connections that convey the
aforementioned information on the angle ␤ and its angular
velocity d␤/dt to the CPG of the LD local control network of
another segment. The output ␤ (and occasionally d␤/dt) of the
LD local control network of one segment will therefore be
transmitted directly to the CPG neurons of the LD local control
network of another segment in form of a synaptic current. In
the behavioral studies (e.g., Graham 1972; Grabowska et al.
2012), it was found that the coordination patterns normally
propagate in the posterior-to-anterior direction. Hence, there
will be no intersegmental synaptic connections on the CPG of
the LD local control system of the metathoracic (HL) segment
(ganglion).
J Neurophysiol • doi:10.1152/jn.00693.2015 • www.jn.org
Downloaded from http://jn.physiology.org/ by 10.220.32.247 on June 17, 2017
HL
α β γ (o)
ML
125
100
75
50
25
α β γ (o)
FL
α β γ (o)
INTERLEG COORDINATION
892
INTERLEG COORDINATION
Weak EF −−> LD connection
3000
4000
5000
α β γ (o)
2000
α β γ (o)
125
100
75
50
25
1000
125
100
75
50
25
125
100
75
50
25
1000
t (ms)
2000
3000
4000
5000
t (ms)
α β γ (o)
125
100
75
50
25
1000
125
100
75
50
25
α β γ (o )
125
100
75
50
25
3000
4000
5000
t (ms)
Strong EF −−> LD connection
125
100
75
50
25
α β γ (o)
HL
α β γ (o)
ML
α β γ (o )
FL
125
100
75
50
25
α β γ (o)
Strong LD −−> EF connection
2000
125
100
75
50
25
1000
2000
3000
4000
5000
t (ms)
Fig. 6. Termination of the intrinsic oscillations of the joint angles by setting the coupling strength between the local control networks to a permanently low or
high value. The specific kind of the reduction or increase of the coupling strengths is indicated above each panel. Note that in 2 cases (Weak LD¡EF and Strong
EF¡LD), all legs grind to a halt on the ground after the oscillations cease, whereas in the 2 other cases (Weak EF¡LD and Strong LD¡EF), all three legs remain
lifted. The traces are color coded as indicated by the color of the angle symbols (␣, etc.) left of the vertical axes in each panel. The arrows in the individual panels
mark the begin of the change in the connection strength.
Interleg Coordination During Tripod
We start with the tripod coordination pattern that has turned
out to be the simpler case. Here, the ipsilateral front and hind
leg move simultaneously, i.e., lift off and touch the ground in
synchrony (Graham 1972). In the simulations, we found the
following simple conditions to suffice for the generation of this
coordination pattern. First, it is continuously monitored
whether the hind leg is in the air, i.e., the angle ␤hleg is greater
than a threshold value (␤th ⫽ 31°). If so, the CPG of the LD
local control network of the middle leg (mesothoracic ganglion) will be inhibited. This ensures that the middle leg stays
on the ground. If not, the middle leg is free to lift off. Second,
the same test is carried out for the middle leg. If it is lifted, the
CPG of the LD local control network of the front leg is
inhibited, otherwise the front leg can lift off. These simple
rules are formalized as a signal flow chart in Fig. 7. It is
noteworthy that these simple rules show good agreement with
rules 1 and 2 of the stepping movement by Cruse (e.g., Cruse
1980; Cruse et al. 1998; Dürr et al. 2004; Schilling et al. 2007,
2013). However, our model does not use the additional rules
(rules 3 and 4) by Cruse that are effective in the rostro-caudal
direction (Dürr et al. 2004; Schilling et al. 2013).
The corresponding synaptic pathways between the segmental LD local control networks are displayed in Fig. 8. Note that
there are no connections from other segments on the LD CPG
of the metathoracic (HL) segment. The actual value of the
conductance (ginh3, ginh9) of the inhibitory synaptic current
from the posterior segment to the next anterior one is completely determined by the actual value of ␤ of the posterior
segment. Whether ginh3 and ginh9 are large (inhibition is effective) or negligible (no inhibition) is decided according to the
rules in the signal flow chart in Fig. 7.
Having implemented the above conditions in the model, we
carried out simulations with it. The results for the tripod
coordination pattern are shown in Fig. 9. For the normal tripod,
the time evolution of all three main angles are displayed in Fig.
9. These are ␣ describing the protractor-retractor (horizontal)
movement of the femur, ␤ of the vertical movement of the
femur, and ␥ of the extension-flexion of the tibia. In this case,
the row for the angle ␤ clearly shows that the hind and front leg
J Neurophysiol • doi:10.1152/jn.00693.2015 • www.jn.org
Downloaded from http://jn.physiology.org/ by 10.220.32.247 on June 17, 2017
125
100
75
50
25
125
100
75
50
25
α β γ (o)
HL
α β γ (o)
ML
α β γ (o)
FL
125
100
75
50
25
α β γ (o)
Weak LD −−> EF connection
INTERLEG COORDINATION
At any instant of time t:
βhleg
Yes
hind leg lifted?
inhibit lev. CPG n.
in middle leg
No
normal drive to lev. CPG n.
in middle leg
No
normal drive to lev. CPG n.
in front leg
Fig. 7. Signal flow chart of the conditions for the tripod coordination pattern.
␤hleg, levation angle (vertical position) of the femur of the hind leg; ␤mleg,
levation angle (vertical position) of the femur of the middle leg; lev. CPG n.,
levator CPG neuron.
move, i.e., lift off and touch down synchronously. This synchrony can also be observed in the time evolution of the
␣-angles (Fig. 9, 1st row left). The extension-flexion movements of these two legs are, however, not synchronous but are
out of phase, since, as mentioned earlier, the hind leg carries
out extension during the stance phase. In addition to the
angular movements, the electrical activity of the CPG of the
PR local control network (Fig. 9, 4th row left) and that of
the CPG of the LD local control network (Fig. 9, 5th row left)
are also exhibited. As can clearly be seen in the 5th row left, the
middle and the front leg are periodically strongly inhibited, the
inhibition coming from the hind and the middle leg, respectively. However, as Fig. 9, 5th row left, shows, the hind leg
receives no inhibition from any of the other legs. Hence, its
oscillatory behavior is completely determined by the local
CPGs and the coupling between them. The period, 605 ms, of
the simulated tripod is in good agreement with values observed
in the experiments. Also, the length of the swing phase and
stance phase are about the same: about half of the whole
period, as it should be (Graham 1972).
In Fig. 9, right, a “failed” tripod is depicted. The failure is
obvious: the front leg is permanently in the air, in the anterior
extreme position, the leg maximally extended. We obtained
this result by making the connections between the LD local
control networks of the three segments cyclic, i.e., adding a
synaptic pathway from the prothoracic segment (front leg) to
the metathoracic one (hind leg). As the simulation result
Next, we turn to the tetrapod coordination pattern. In this
case, usually four out of six legs are simultaneously on the
ground at any instant of time. Looking only at one side, i.e., at
the three ipsilateral legs, a wave-like process of leg movements
can be observed (Fig. 4). That is, the legs carry out basically
the same kind of movement (lift-off, touch-down, protraction
and retraction, extension and flexion) but with a constant phase
delay within a period. The wave starts at the hind leg and
propagates in the anterior direction (Graham 1972; Grabowska
et al. 2012).
For finding the rules and constraints for this coordination
pattern, we used the same general principles outlined in
Basic Principles of the Interleg Coordination. First of all, a
simple reasoning shows that if the hind leg is lifted, then
“both” the middle and the front leg must stay on the ground.
Hence, the levator CPG neuron of both LD local control
networks must be inhibited; otherwise, the central drive to
the CPG would still prevail. However, this condition, although necessary, has proved to be insufficient for performing the tetrapod coordination pattern. It turned out that
signals carrying information on the actual values of the
angular velocities of the levation angle ␤, i.e., the quantities
d␤/dt at each segment, are also required. They determine
whether a leg is being lifted (d␤/dt ⬎ 0) or is touching down
(d␤/dt ⬍ 0). For each leg, two states are defined: “ready-tostep,” and “stepping” state. The coordination process consists
of two stages: in the first one, the actual states just mentioned
are determined by using the angular velocities; in the second
one, it is decided whether the levator CPG neuron of the LD
local control network will be inhibited. These stages are
specified and formalized in the signal flow charts of Fig. 10.
Thus at any instant of time, it is decided whether the hind leg
is in the “ready-to-step” state, i.e., is allowed to carry out the
next step. If so, the levator CPG neuron of the LD local control
network of the middle leg will be inhibited to keep the middle
leg on the ground. Next, it is determined whether the hind leg
is being lifted (condition: d␤hleg/dt ⬎ 0). If it is, the hind leg’s
state changes to the “stepping” state. It will only be reset to the
“ready-to-step” state when the middle leg finishes its step
(touches down). The hind leg thus remains in the “stepping”
state until the reset happens. Exactly the same procedure is
repeated with regard to the middle and front leg to determine
the state of the middle leg. In the second stage, it is determined
whether the hind leg is lifted. If so, the levator CPG neurons of
the LD local control network of the middle and front leg will
J Neurophysiol • doi:10.1152/jn.00693.2015 • www.jn.org
Downloaded from http://jn.physiology.org/ by 10.220.32.247 on June 17, 2017
middle leg lifted?
inhibit lev. CPG n.
in front leg
shows, this connection has destroyed the tripod coordination
pattern; hence, it cannot be part of the intersegmental connections that are used to convey signals coordinating the legs’
movement during tripod. It is obvious that a too low level of
connectivity precludes the occurrence of the tripod coordination pattern. This example shows that a too high level of
connectivity can also lead to the disruption of the tripod
coordination pattern.
In summary, our simulation results show that the two intersegmental connections used in the model (Fig. 8), together
with the simple rules (Fig. 7) associated with these connections, are sufficient to bring about the coordination of the leg
movements during tripod.
Interleg Coordination During Tetrapod
βmleg
Yes
893
894
INTERLEG COORDINATION
PR control network
LD control network
EF control network
g
g
g
g
d1
d2
g app1 g app2
Pro−thorax (FL)
g
MN
IN2
C38
C1
C2
g
MN1P
C19
MN2R
C20
MN
g
MN
IN6
C42
C3
C4
g
MN3D
C21
g
MN
IN10
C46
C5
C6
g
MN5F
C23
MN6E
C24
IN7
C43
Ext. m.
γ
IN8
C44
g d8
IN11
C47
IN12
C48
EF control network
g
g
g
d14
IN14
C50
C7
C8
MN8R
C26
MN
g
MN
d18
g app9 g app10
g inh9
IN13
C49
g
d17
IN17
C53
IN18
C54
C9
C10
MN10L
C28
MN
g
MN
IN21
C57
IN22
C58
C11
C12
g
MN11F
C29
CPG
CPG
Pro. m.
IN16
C52
Flex. m.
Lev. m.
β
IN15
C51
MN12E
C30
MN
CPG
Dep. m.
Ret. m.
d22
g app11 g app12
g
MN9D
C27
g
d21
Ext. m.
γ
IN20
C56
IN19
C55
IN23
C59
IN24
C60
g d20
PR control network
g
g
d25
d26
g
MN
IN25
C61
IN26
C62
C13
C14
g
MN13P
C31
M14R
C32
MN
g
MN
g
g
d29
d30
IN29
C65
IN30
C66
C15
C16
MN16L
C34
MN
g
MN
IN27
C63
β
IN33
C69
IN34
C70
C17
C18
MN18E
C36
MN
CPG
Lev. m.
IN31
C67
d34
g
MN17F
C35
CPG
Dep. m.
Ret. m.
g
d33
g app17 g app18
g
MN15D
C33
CPG
Pro. m.
IN28
C64
EF control network
g
g app15 g app16
g app13 g app14
Meta−thorax (HL)
LD control network
IN32
C68
Flex. m.
Ext. m.
IN35
C71
IN36
C72
γ
g d32
Fig. 8. The 3-leg model with the intersegmental connections that underlie the tripod coordination pattern. The intersegmental thick lines are inhibitory synaptic
pathways from the posterior segment to the next anterior one. The synaptic strengths, i.e., the actual values of the conductances ginh3 (prothoracic LD CPG) and
ginh9 (mesothoracic LD CPG) are completely determined by the actual values of ␤ in the next posterior segment. Note that there is no such inhibitory synaptic
connection on the LD CPG of the metathoracic (HL) segment.
be inhibited, preventing both legs from lifting off. If the
condition is not fulfilled then it is determined whether the hind
leg is still in the “ready-to-step” state. If not, then the levator
CPG neuron of the LD local control network of the middle leg
receives the excitatory signal required for starting the oscillation of the CPG. In addition, it is also determined in this case
whether the middle leg is lifted (using the angle ␤mleg, as
shown in Fig. 10). If the middle leg is lifted, the levator CPG
neuron of the LD local control network of the front leg will be
inhibited to keep the front leg on the ground; otherwise, it will
be decided whether the middle leg is in the “ready-to-step”
state. If not, the levator CPG neuron of the LD local control
network of the front leg will receive its excitatory drive
necessary to perform the next lift-off. In addition, one should
bear in mind that in tetrapod the central drive to the CPGs is
different from that in tripod, i.e., different gapp values apply.
They bring about the change in the length of the oscillatory
period and in the duty cycle.
J Neurophysiol • doi:10.1152/jn.00693.2015 • www.jn.org
Downloaded from http://jn.physiology.org/ by 10.220.32.247 on June 17, 2017
LD control network
g
d13
MN
CPG
Flex. m.
PR control network
g
MN7P
C25
MN
Lev. m.
β
IN3
C39
d10
IN9
C45
CPG
g app7 g app8
MN
MN4L
C22
Dep. m.
Ret. m.
IN4
C40
g
g
d9
g app5 g app6
IN5
C41
CPG
Pro. m.
Meso−thorax (ML)
d6
g app3 g app4
g inh3
IN1
C37
g
d5
INTERLEG COORDINATION
895
Failed tripod
o
α( )
130
105
80
55
30
60
50
50
o
β( )
60
40
40
30
115
115
90
90
o
γ ( )
30
65
40
2000
3000
4000
5000
65
40
2000
t (ms)
3000
4000
5000
t (ms)
V (mV)
10
-10
PR
-30
-50
-70
hind leg
middle leg
V (mV)
10
front leg
-10
-30
LD
-50
-70
2000
3000
4000
5000
t (ms)
Fig. 9. A normal and a “failed” tripod as indicated are displayed. For the normal tripod: 1st row left: shows the protractor-retractor (horizontal) movement (angle
␣); 2nd row left: the time evolution of the angle ␤ (vertical movement) of the femur; 3rd row left: the extension-flexion movement of the tibia (angle ␥). In
addition, the electrical activity of the protractor CPG neurons (4th row left) and the levator CPG neurons (5th row left) of all 3 legs are displayed. In the 5th
row left, the strong inhibitory effects on the levator CPG neurons of the LD local control networks of the middle and front leg (green and red line, respectively)
are clearly discernible, while there is no such effect on the hind leg (blue line). For the “failed” tripod (right), only the angular movements are shown. Note that,
in this case, the front leg is permanently in the air and in protracted position (minimal value of ␣) with maximal extension of the leg (minimal value of ␥).
Figure 11 displays the three-leg model with the intersegmental connections required for performing the tetrapod coordination pattern. Here, there is a direct inhibition from the metathoracic segment to both the meso- and prothoracic segment, in
contrast to the intersegmental connections in tripod (cf. Fig. 8).
The other important difference is that, beside the position
signals ␤, their time derivatives, the angular velocities d␤/dt
also play an important gating role in generating tetrapod. This
is symbolically indicated in Fig. 11, the gating sites are also
shown as enlarged insets. (Note that the PR local networks of
the three-leg model were omitted from Fig. 11 for the sake of
simplicity.)
Using all the aforementioned rules and conditions in the
simulations, as outlined in the signal flow charts of Fig. 10, we
obtained results that are satisfactorily close to the tetrapod
coordination pattern. Figure 12 shows an example of a simulated tetrapod produced by the three-leg model. The tetrapod
coordination pattern is clearly visible in all rows, i.e., with
regard to all angle movements, and the electrical activity of the
corresponding protractor and levator CPG neurons. The simu-
lated tetrapod had a period of ⬇1,171 ms, which falls into the
range of values observed in the experiments. Also, the relative
length of ⬃0.27 of the swing phase is in good agreement with
the experimental data (e.g., Graham 1972).
The example of a “failed” tetrapod in the simulations (Fig.
13) emphasizes the importance of using the angular velocity
signals when determining the rules and conditions for the
tetrapod coordination pattern. In this example, the angular
velocity signals were omitted from the conditions. They are
thus indispensable sensory signals for shaping the tetrapod
coordination pattern.
Finally, we show that interleg coordination affects intraleg
coordination. For this, we return to the simulations in which
there was no interleg coordination, and the strengths of the
intraleg connections were kept permanently weak or strong (cf.
Fig. 6). There, we found that when the strength of the EF¡LD
coupling was weak, or the LD¡EF coupling was strong, all
three legs remained lifted (in the air) in the steady state, which
is physiologically unrealistic. We repeated the simulations but
now with interleg connections characteristic to the tetrapod
J Neurophysiol • doi:10.1152/jn.00693.2015 • www.jn.org
Downloaded from http://jn.physiology.org/ by 10.220.32.247 on June 17, 2017
γ (o)
β (o)
α (o)
Normal tripod
130
105
80
55
30
896
INTERLEG COORDINATION
At any instant of time t:
βhleg
d βhleg /dt
hind leg ready
to step?
d βhleg /dt >d φcr /dt
Yes
βhleg
d βmleg /dt
hind leg ready
to step?
d βmleg /dt <−d φcr /dt
Yes
inhibit lev. CPG n.
in middle leg
hind leg during
stepping
Yes
No
middle leg is
finishing step
hind leg ready
to step
Repeat the same with hind leg
middle leg, and middle leg
front leg.
βhleg
Yes
hind leg lifted?
inhibit lev. CPG n.
in middle and front leg
No
βmleg
hind leg ready to step?
No
normal drive to lev. CPG n.
in middle leg
middle leg lifted?
No
Yes
inhibit lev. CPG n.
in front leg
middle leg ready to step?
No
normal drive to lev. CPG n.
in front leg
Fig. 10. Signal flow chart of the conditions governing tetrapod. ␤hleg Is vertical position of the femur of the hind leg; d␤hleg/dt is its angular velocity. For the
middle leg, the variables ␤mleg and d␤mleg/dt have analogous meaning. The quantity d␾cr/dt is a threshold value of the angular velocity used for reasons of
numerical computation. The expression “hind leg¡middle leg” means that the variables and parameters relating to the middle leg are substituted in the above
signal flow chart for those of the hind leg. The expression “middle leg¡front leg” has the same meaning (but using different legs). The other abbreviations are
the same as in Fig. 7.
coordination pattern. The results are displayed in Fig. 14. In
these simulations, the same connection strengths were used as
in the earlier ones with no interleg connections. While the
cases “weak LD¡EF” and “strong EF¡LD” connection show
essentially the same steady state as before, the two other cases
do not. In the new simulations, the front and the middle leg end
up touching the ground in the steady state, contrary to what
was seen in the earlier simulations with no interleg connec-
tions. This indicates that interleg coordination plays a part in
shaping the steady state of the legs in certain conditions. By
this, interleg coordination improves the physiological plausibility of the model. It does not escape, however, one’s attention
that in both of the latter cases (weak EF¡LD and strong
LD¡EF connection), the hind leg remains lifted. To establish
ground contact of this leg, i.e., to stop it properly, a different
mechanism becomes active, which has been introduced and
J Neurophysiol • doi:10.1152/jn.00693.2015 • www.jn.org
Downloaded from http://jn.physiology.org/ by 10.220.32.247 on June 17, 2017
AND
INTERLEG COORDINATION
LD control network
EF control network
g
g
g
MN
g
d5
d6
IN5
C41
IN6
C42
C3
C4
MN4L
C22
MN
g
MN
IN9
C45
IN10
C46
C5
C6
g
MN5F
C23
CPG
dβ
Flex. m.
Lev. m.
β
MN6E
C24
MN
CPG
Dep. m.
Dep. m.
d10
g app5 g app6
g
MN3D
C21
g
d9
g app3 g app4
g inh3
Pro−thorax (FL)
897
IN7
C43
Ext. m.
γ
IN8
C44
g d8
IN11
C47
IN12
C48
LD control network
EF control network
g
g
g
MN
d18
g app9 g app10
g inh9
Meso−thorax (ML)
g
d17
IN17
C53
IN18
C54
C9
C10
MN10L
C28
MN
g
MN
IN21
C57
IN22
C58
C11
C12
g
MN11F
C29
CPG
dβ
Flex. m.
Lev. m.
β
MN12E
C30
MN
CPG
Dep. m.
Dep. m.
d22
g app11 g app12
g
MN9D
C27
g
d21
Ext. m.
γ
IN20
C56
IN19
C55
IN23
C59
IN24
C60
g d20
dβ
LD control network
EF control network
g
g
g
d29
d30
g app15 g app16
Meta−thorax (HL)
g
MN
dβ
dβ
IN29
C65
IN30
C66
C15
C16
MN16L
C34
MN
g
MN
Lev. m.
IN31
C67
IN34
C70
C17
C18
MN18E
C36
MN
CPG
Dep. m.
β
IN33
C69
g
MN17F
C35
CPG
Dep. m.
d34
g app17 g app18
g
MN15D
C33
g
d33
Flex. m.
IN32
C68
Ext. m.
IN35
C71
IN36
C72
γ
g d32
Fig. 11. Intersegmental connections in the 3-leg model required for the tetrapod coordination pattern. The conductances ginh3 and ginh9 have the same meaning
as in Fig. 8 but here, the inhibitory effect on the levator CPG neuron of the LD local control network of the prothoracic segment can directly be evoked by the
␤-signal of the metathoracic segment (direct thick line from the metathoracic segment to the prothoracic one). The pentagons with d␤ in them symbolize the
sources of the sensory signals that convey information on the angular velocities d␤/dt. They are enlarged as insets for better visibility next to their location in
the network. The circles on the connecting pathways express the gating function the angular velocities exert on the (inhibitory) signal flow from the meta- and
mesothoracic segments, respectively. Downward directed arrows from the d␤ pentagons pointing to these circles symbolize their influence on the next caudal
leg (segment), upward pointing arrows to the next anterior one. The protractor-retractor local control networks of all segments are omitted from this figure for
the sake of simplicity.
J Neurophysiol • doi:10.1152/jn.00693.2015 • www.jn.org
Downloaded from http://jn.physiology.org/ by 10.220.32.247 on June 17, 2017
dβ
898
INTERLEG COORDINATION
o
α( )
Normal tetrapod
130
105
80
55
30
o
β( )
60
50
40
30
o
γ ( )
65
40
2000
3000
4000
5000
t (ms)
10
V (mV)
Fig. 12. Simulated tetrapod produced by the 3-leg
model using the conditions in Fig. 10. The rows
display the same variables, and the color codes for
the legs are the same as in Fig. 9. Note again the
strong inhibition of the levator CPG neurons of the
LD local control networks of the middle and front
leg in the bottom row.
90
-10
PR
-30
-50
-70
V (mV)
10
-10
-30
LD
-50
-70
2000
described in detail in Tóth et al. (2013b). This mechanism
involves slow motoneurons and muscle fibers, which are omitted from the present model for the sake of simplicity and
clarity. We could do so because slow muscles, although being
important in maintaining steady state position of the animals,
have only a small contribution to shaping the stepping movements (Bässler and Stein 1996; Gabriel et al. 2003; Guschlbauer et al. 2007; Tóth et al. 2013a).
Transition Between Tripod and Tetrapod
Having successfully reproduced the tripod and tetrapod
coordination patterns in the simulations, the questions naturally
arise: how do transitions between them take place? What
mechanisms and what sensory and (possibly) central neuronal
signals play a role in the transition processes?
First of all, we can make some simple observations and thus
formulate basic necessary conditions to be fulfilled during
transitions from one type of coordination pattern to the other.
3000
4000
5000
t (ms)
Such a simple observation is that no more than two legs must
be lifted at the same time. Obviously, three lifted ipsilateral
legs would cause the animal to lose balance and to fall on its
side on which all three legs are in the air. Figure 15 shows such
a “forbidden” transition. As can easily be seen, all three
␤-angles, which reliably reflect the legs’ vertical position, have
values greater than 50° at one point of time in the interval
[3,000,4,000] ms. That is, at that point of time, all three
ipsilateral legs are lifted.
We should also bear in mind that we, in the present study,
only deal with the stepping of ipsilateral legs in the two
coordination patterns. The inclusion of all six legs into the
considerations in some future model can, and most likely will,
lead to a stricter constraint, i.e., that, occasionally, only a single
leg is allowed to be lifted on one side during transition from
one coordination pattern to the other. In view of this, we
endeavored to adhere to this stricter constraint, with partial
success. Figure 16 shows a successful transition from tetrapod
J Neurophysiol • doi:10.1152/jn.00693.2015 • www.jn.org
Downloaded from http://jn.physiology.org/ by 10.220.32.247 on June 17, 2017
115
INTERLEG COORDINATION
899
α (o)
Failed tetrapod
130
105
80
55
30
Fig. 13. Example of a “failed” tetrapod. It was obtained by
omitting the information carried by the angular velocities
(cf. Fig. 10). The error is that the front and the hind leg lift
off simultaneously, and that the middle and front leg are in
the air at the same time. These anomalies are best seen in
the middle row showing the time course of the ␤-angles but
they are also discernible in the time courses of the ␣ (top)and ␥ (bottom)-angles. The color codes are the same as in
Fig. 9.
β (o)
60
50
40
30
γ (o)
90
65
40
2000
3000
4000
5000
to tripod where this stricter constraint is fulfilled. The transition
period is here virtually absent because the finishing part of the
last tetrapod (front leg stepping) coincides with the starting of
the tripod (concurrent lifting of the front and the hind leg).
In the simulations, we systematically investigated whether
all points of time within a step cycle can equally well serve as
a starting point of a transition. This turned out not to be the
case. The simulations showed that transition from one coordination pattern to the other one is successful only if the angle
variable ␤ and the corresponding angular velocity d␤/dt of the
different legs obeyed specific conditions. Moreover, these
conditions proved to be different depending on the direction of
the transition.
Transition from tetrapod to tripod. First, we treat the transition from tetrapod to tripod. Here, we found by means of
simulations that the front leg or the hind leg had to be lifting off
in order that the transition be successful. More precisely, the
angle ␤ had to exceed a threshold value and the corresponding
angular velocity had to be positive. (Practically, it had also to
be greater than a threshold value.) If these conditions were
satisfied then the conductances (gapp) of the central driving
currents of all nine CPGs were set to their values suitable for
the tripod coordination pattern. This means that these conductance values ensured that the oscillations with the correct
oscillatory period (Tp ⬇605 ms) were produced by the CPGs.
It should be borne in mind that with the transition, the “tripod
rules” of intersegmental coordination (Fig. 7) as well as the
corresponding intersegmental connections (Fig. 8) henceforth
automatically applied. Figure 17 summarizes the transition
conditions. Interestingly, the middle leg plays no part in these
conditions. Indeed, the simulations demonstrated that the transition from tetrapod to tripod failed, if the same kind of
conditions was applied to the middle leg, i.e., to its ␤-angle and
the corresponding angular velocity d␤/dt.
Figure 18 shows two cases when either the hind leg or the
front leg satisfies the transition condition. The presumably
central commands initiating the transitions are marked by
arrows. These commands can be thought of as neuronal signals
that activate the processes that underlie the transition, e.g.,
t (ms)
setting the driving excitatory signals to the CPGs, once the
transition conditions are fulfilled. At present, there is no direct
experimental evidence for their existence. It would indeed be
difficult to devise an experiment that could prove or disprove
the existence of such a command. It thus remains, for the time
being, hypothetical. In Fig. 18, top row, the unperturbed
tetrapod coordination pattern is displayed with these arrows to
show the exact instants of time when the commands appear.
Figure 18, middle row, illustrates the case when the transition
command arrives at the start of the lift-off of the hind leg. A
transitory period of about one tetrapod oscillatory period follows during which, at some point of time, two legs, the middle
and the hind leg, are simultaneously in the air. The next time
the hind leg is being lifted off, it is doing so together with the
front leg, i.e., they are already in tripod. In Fig. 18, bottom row,
we illustrated the case when the front leg satisfies the transition
condition at the arrival of the central command for it. Here, the
transitory phase is virtually absent, and the tripod coordination
pattern immediately commences.
Transition from tripod to tetrapod. Now, we deal with the
transition from tripod to tetrapod. Here, too, systematic simulations varying the command signals within one oscillatory
period of the tripod coordination pattern revealed that successful transitions from tripod to tetrapod would occur if both the
hind leg and the middle leg were moving slowly, or were
approaching the ground. More precisely, the angular velocities
in the vertical direction (Fig. 18, top row) of the legs involved
were below a small positive threshold. This also includes
negative velocities when the leg is approaching the ground. If
this condition was fulfilled the central drives to all nine CPGs
were adjusted to produce periodic oscillation suitable to tetrapod, i.e., one with a period of Tp ⬇1,171 ms. The adjustment
was brought about by changing the values of all conductances
(gapp) of the central drives to all CPGs. In addition, the rules of
the tetrapod coordination pattern (Fig. 10) as well as the
corresponding intersegmental connections (Fig. 11) automatically took effect. Figure 19 summarizes the transition conditions just described. Note that the “AND” condition is almost
automatically satisfied, since if one of the legs involved is in
J Neurophysiol • doi:10.1152/jn.00693.2015 • www.jn.org
Downloaded from http://jn.physiology.org/ by 10.220.32.247 on June 17, 2017
115
900
INTERLEG COORDINATION
3000
4000
5000
α β γ (o)
2000
α β γ (o)
125
100
75
50
25
1000
125
100
75
50
25
125
100
75
50
25
125
100
75
50
25
1000
t (ms)
3000
4000
5000
t (ms)
α β γ (o)
2000
α β γ ( o)
125
100
75
50
25
1000
125
100
75
50
25
125
100
75
50
25
α β γ ( o)
α β γ (o)
HL
α β γ ( o)
ML
125
100
75
50
25
α β γ ( o)
FL
3000
4000
5000
t (ms)
5000
t (ms)
Strong EF −−> LD connection
Strong LD −−> EF connection
125
100
75
50
25
2000
125
100
75
50
25
1000
2000
3000
4000
Fig. 14. Rerunning the simulations with weak and strong intraleg coupling in the presence of interleg connection (in tetrapod). The constant phase difference
between the FL, ML, and HL traces is a clear sign of its presence. Note that when the EF¡LD connection is weak (top right) or the LD¡EF connection strong
(bottom left), the FL and the ML are on the ground in steady state (cf. Fig. 6). The color code for the traces and the role of the arrows in the individual panels
are the same as in Fig. 6.
the air and approaching the ground the other must be on the
ground. However, it was easier to formulate the transition
condition by using the “AND” operation. Note also that the
front leg is not explicitly involved in the condition but, since it
moves concurrently with the hind leg during tripod, its ␤ angle
and the corresponding angular velocity are (nearly) identical to
those of the hind leg.
Figure 20 illustrates a successful transition from tripod to
tetrapod. In this example, the central command arrives when
both the hind leg and the middle leg are moving very slowly
near, or on, the ground. However, the region of admissible
timing of the central command for transition is much larger.
The time intervals in which the central commands are
effective are of importance for the transitions in both directions. Figure 21 displays such intervals for both types of
transition as indicated in the figure. They are identified by a
pair of thin vertical bars. They arise from the conditions
formulated for the two transition types (cf. Figs. 17 and 19). In
Fig. 21, left and right, the top traces are the time courses of the
angles ␤ and the bottom traces are those of the angular
velocities d␤/dt. The horizontal lines, parallel to the time axes,
are the threshold values: 31° for ␤ in Fig. 21, left, and 5.0 1/s
for d␤/dt in Fig. 21, left and right.
During tetrapod¡tripod transition (Fig. 21, left), the central
command is effective in the time interval in which the angular
velocity d␤/dt and ␤ exceed their own threshold. During tripod¡tetrapod transition (Fig. 21, right), the command is effective
while d␤/dt is below the threshold. Besides the intervals shown in
Fig. 21, there is one such additional interval each in both cases,
although they are not displayed in Fig. 21. During transition from
tetrapod to tripod, this additional interval is about the same length
commencing when the front leg starts lifting off the ground.
During transition into the opposite direction, there is also an
analogous condition that gives rise to an additional interval in
which the command to change coordination pattern is immediately effective: the middle leg is stepping and the hind leg remains
on the ground. We have thus two intervals each for the two types
of transition. The intervals in the case of transition from tripod to
tetrapod are much longer: 256 ms than in the opposite case: 45 ms.
If the command signal arrives outside the intervals just identified,
the transition will be delayed and will not start until the transition
conditions are fulfilled.
J Neurophysiol • doi:10.1152/jn.00693.2015 • www.jn.org
Downloaded from http://jn.physiology.org/ by 10.220.32.247 on June 17, 2017
125
100
75
50
25
Weak EF −−> LD connection
α β γ (o)
HL
α β γ (o)
ML
α β γ (o)
FL
125
100
75
50
25
α β γ (o)
Weak LD −−> EF connection
α (o)
INTERLEG COORDINATION
901
130
105
80
55
30
β (o)
60
Fig. 15. Example of a “forbidden” transition from tetrapod
to tripod. The rows display the same angles, and the color
codes are the same, as in Fig. 9. In the traces of the ␤-angles
(middle), all 3 ␤ are greater than 50°, meaning that all 3 legs
are high in the air at the same time, at one point after the
3,000-ms mark.
50
40
30
90
65
40
2000
3000
4000
5000
DISCUSSION
Building on previous models (Tóth et al. 2012; Knops et al.
2013), we constructed a three-leg model to simulate the normal
locomotion (walking) of the stick insect. Our approach differed
from that of other models mentioned in the Introduction (e.g.,
Cruse 1980, 1990; Cruse et al. 1998; Dürr et al. 2004; Ekeberg
et al. 2004; von Twickel et al. 2011, 2012; Schilling et al.
2013). Those models used artificial neural networks to build
the control systems that governed the muscles that move the
individual leg segments. In contrast to them, we built our
models of modules, called local neuronal control networks, that
were constructed using anatomical data and experimental results if they were available (Bässler 1983; Büschges 1995,
1998, 2005; Westmark et al. 2009) and reasonable physiological assumptions (Daun-Gruhn 2011; Daun-Gruhn and Tóth
2011; Daun-Gruhn et al. 2011; Tóth et al. 2012; Knops et al.
2013), where no relevant data existed. On the same basis, we
defined specific connections, sensory synaptic pathways, between the local control networks both intra- and intersegmentally. No such approach has, to the best of our knowledge, been
applied by other models in this field. The model created thus
and the simulation results produced by it can therefore be
considered as novel. In the present model, for each leg, the
most important antagonistic muscle pairs and their neuronal
control networks have been implemented: the protractor-retractor, the levator-depressor, and the extensor-flexor neuromuscular networks. They are interconnected in a suitable way to
bring about synchronized periodic oscillation of all three neuromuscular networks. The synchronization is such as to produce stepping movements of the individual legs. We kept all
legs mechanically identical using data of the middle leg. This
had the advantage that differing mechanical properties, which
could otherwise have obscured the main functions of the local
control networks of the individual legs and leg joints in the
interleg coordination, were not present. At the same time, this
simplification did not amount to a serious shortcoming of the
model, since it is a kinematic rather than a dynamic one, as far
as its mechanical properties are concerned.
During walking, interleg (intersegmental) coordination of
the ipsilateral legs is also required. As stated in the Introduc-
t (ms)
tion, the main objective of the present study has been to
establish this required coordination between their movements
by means of a neuronal model. Thus we looked for suitable
intersegmental connections between the control networks of
the individual legs. We reasoned that connections between the
segmental levator-depressor control networks would play a
primary role. The sensory signal expressing the vertical position ␤ and the corresponding angular velocity d␤/dt completely
characterize the state the leg is in (stance or swing state, and
the vertical direction of movement in the latter). Ground
contact of the leg can uniquely be described by a critical value
of the levation angle ␤ at which the stance phase commences.
The levation angle is controlled by the neuromuscular LD
system at each leg. Strong experimental evidence for its primary role in shaping locomotion (walking) was found by Ayres
and Davis (1977), although not in the stick insect, but in the
lobster.
Using the aforementioned sensory signals, in the form of
(actual) values of the levation angle ␤ and its time derivative
d␤/dt, we could derive conditions that determined the coordinated movements of the individual legs during the tripod and
tetrapod coordination patterns. We found important differences
between the conditions for tripod and tetrapod. The main
qualitative difference is that the conditions for tripod only
require the actual values of the angles ␤, whereas tetrapod also
requires the sensory signals encoding the angular velocities
d␤/dt. Figures 9 and 13 demonstrate this finding. In physiological terms, this means that the tetrapod coordination pattern
requires more sensory input and hence is more heavily dependent on it, to take place than the tripod coordination pattern.
The intersegmental connections between the levator-depressor
control networks are more complex in the former than in the
latter case (cf. Figs. 8 and 11).
The posterior-anterior propagation of stepping during the
two coordination patterns is well known to happen in the stick
insect (e.g., Graham 1972, 1985; Grabowska et al. 2012).
Moreover, Grabowska et al. (2012) have also shown that the
pair of middle legs plays a crucial part in the interleg coordination. If these legs are amputated, the coordination between
the front and hind legs ceases. This is in good agreement
J Neurophysiol • doi:10.1152/jn.00693.2015 • www.jn.org
Downloaded from http://jn.physiology.org/ by 10.220.32.247 on June 17, 2017
γ (o)
115
INTERLEG COORDINATION
o
α( )
902
130
105
80
55
30
o
β( )
60
50
40
30
o
Fig. 16. Transition from tetrapod to tripod. The
rows display the same variables, and the color codes
are the same, as in Fig. 9. The timing of the
transition command is labeled by an arrow in the
row of the ␤-traces. Note that the transition period
is virtually absent. The last part of the tetrapod (step
of the front leg) coincides with the start of the tripod
(simultaneous lifting of the front and hind leg).
90
65
40
2000
3000
4000
5000
t (ms)
V (mV)
10
-10
-30
PR
-50
-70
V (mV)
10
-10
-30
LD
-50
-70
2000
with the model since all coordination patterns, whether
tripod or tetrapod, require sensory signals from the middle
leg. Hence the interleg coordination in the model, too, is
abolished if these signals are missing.
As far as the transitions between tetrapod and tripod are
concerned, we could also derive simple conditions for them.
They are, however, dependent on the direction of the transition
(cf. Figs. 17 and 19). Moreover, the direction of the transition
also determines the time intervals in which a central command
for transition can become effective (cf. Fig. 21). The constraints that arise show that during transition from tetrapod to
tripod, these time intervals within an oscillatory period are
rather small (in total ⬃90 ms), in the opposite direction,
however, much larger (in total ⬃512 ms). This asymmetry can
be interpreted with respect to the locomotion patterns of the
stick insect that tetrapod seems to be preferred to tripod in the
animal.
There remains an important question to be answered: are the
coordination structures used in the model for tripod and tetra-
3000
4000
5000
t (ms)
pod separate? The answer simply is: no, they are not. A look at
Figs. 8 and 11 suffices to see that the coordinating system that
is active during tripod is a subset of that which is active during
tetrapod. Hence, it is sufficient to assume the existence of a
single intersegmental coordinating system in the model. This
system is fully activated during tetrapod, including the sensory
pathways that convey information on the angular velocities.
During tripod, it is only partially activated as shown in Fig. 8.
The function of the parts that are active during both tripod and
tetrapod is the same, namely temporary inhibition of the
corresponding CPG. The (hypothetical) command signal that
initiates the switch between the coordination patterns could
also activate or inactivate parts of the interleg connections
according to whether they are required in the actual coordination pattern.
Even though the conditions for the tetrapod and tripod
coordination patterns have been implemented in the model in
an abstract way illustrated by signal flow charts (cf. Figs. 7, 10,
17, and 19), they could, in principle, be converted into neuronal
J Neurophysiol • doi:10.1152/jn.00693.2015 • www.jn.org
Downloaded from http://jn.physiology.org/ by 10.220.32.247 on June 17, 2017
γ ( )
115
INTERLEG COORDINATION
903
At any instant of time t:
At any instant of time t:
βfleg
d βfleg /dt
front leg is lifting off?
βhleg
d βhleg /dt
d βhleg /dt
hind leg is slow or
approaching ground?
hind leg is lifting off?
Yes
d βmleg /dt
middle leg is slow or
approaching ground?
Yes
Yes
Yes
AND
set central drive of all
CPGs to 3pod values
Fig. 17. Conditions for the transition from tetrapod to tripod. These conditions
are the same for both the front and hind leg. It suffices if one of these legs
satisfies them, hence, the “OR” logical operation. Note that the middle leg
plays no part in the transition conditions.
networks that would produce the behavior described by those
charts. However, a practical implementation would face difficult problems. The main problem is that currently, there are no
data, not even hypotheses, of what such networks would look
like. Would they comprise just a few dozens of neurons or
hundreds of them? What kind of neurons might they be, etc.?
Another problem would likely be finding suitable values for the
parameters of a large number of additional neurons without
knowing their anatomical and biophysical properties. Furthermore, technical difficulties would also arise from including
additional model neurons into the model. The inclusion would
substantially increase the size of the differential equation
system to be solved in the simulations and hence reduce the
numerical stability of the solutions. Moreover, it would sub-
set central drive of all
CPGs to 4pod values
Fig. 19. Conditions for transition from tripod to tetrapod. The “AND” condition is almost trivially satisfied, since if one of the legs is lifted and approaching the ground the other must be on the ground.
stantially prolong the computational time. Finally, the additional neuronal networks might distract attention from the main
properties of the model, namely the connections between the
LD systems of the individual legs. Hence, an extension of
the kind would, at present, implant more shortcomings into the
model than advantages.
There exist other types of models, too, which can simulate
even full insect walking, i.e., using all six legs. One example is
the model by von Twickel et al. (2011, 2012). It is based by
constructing appropriate artificial neural networks, which are
then trained to carry out the task of controlling the movement
of the leg. The single-leg controller (von Twickel et al. 2011)
β (o)
60
50
40
30
Fig. 18. Two cases of successful transition from tetrapod to
tripod. Top row: unperturbed tetrapod with the ␤-angles and
with the instants of time when the (central) command for
transition appears. Middle row: the case when the command
occurs at the beginning of the lift-off of the hind leg. A
transitory phase of about one tetrapod oscillatory period can
be discerned. Note also that during this period, there is a short
time interval when both the middle and the hind leg are lifted
(at the crossing of the green and the blue line at ⬃40°).
Bottom row: the case when the front leg satisfies the transition condition, the command for it arriving at the 2nd (later)
arrow as indicated at both top and bottom rows. Here,
virtually no transitory phase is discernible; the tripod coordination pattern commences immediately.
β (o)
60
50
40
30
β (o)
60
50
40
30
3000
4000
5000
t (ms)
J Neurophysiol • doi:10.1152/jn.00693.2015 • www.jn.org
Downloaded from http://jn.physiology.org/ by 10.220.32.247 on June 17, 2017
OR
INTERLEG COORDINATION
o
α( )
904
130
105
80
55
30
o
β( )
60
50
40
30
o
90
65
40
6000
Fig. 20. Transition from tripod to tetrapod. The rows
display the same variables, and the color codes are the
same, as in Fig. 9. The arrow indicates the timing of the
central transition command. The transitory phase is
about one tripod oscillatory period (⬇605 ms).
7000
8000
9000
10000
11000 t (ms)
V (mV)
10
-10
-30
PR
-50
-70
V (mV)
10
-10
-30
LD
-50
-70
6000
uses an earlier leg model by Ekeberg et al. (2004). With the use
of it, a six-legged control system (robot) (von Twickel et al.
2012) was developed. Thus the model by von Twickel et al.
(2012) can emulate six-leg walking. Another example is Walknet (Cruse 1980; Cruse et al. 1998; Dürr et al. 2004; Schilling
et al. 2007, 2013), which is a complex control system. Parts of
it are artificial neural networks, other parts are built of the usual
functional units of control systems, such as integrators, nonlinear threshold elements. Walknet uses all angle variables (␣,
␤, and ␥) to be controlled, and several versions of it also the
angular velocities, as additional control variables. The system
works by implementing a number of rules, the so-called “Cruse
rules” (e.g., Dürr et al. 2004; Schilling et al. 2013). These are
rules that determine the interleg coordination in Walknet. As a
result, Walknet is capable of mimicking six-leg walking in a
number of (simulated) environmental conditions. By contrast,
our model, having only three (simulated) legs, cannot replicate
the full walking patterns with six legs. It can only produce the
ipsilateral projection of the coordination (walking) patterns
7000
8000
9000
10000
11000 t (ms)
tetrapod and tripod. However, it was built on the basis of a
quite different concept: a close and direct relation to the
morphological and physiological properties of its biological
counterpart, the stick insect. Consequently, it can highlight
new aspects of the mechanisms that generate intersegmental
coordination during walking. Despite the substantial differences between these models concerning their concept and the
modeling techniques applied, they implement basically the
same coordination rules in their own way. This important
common property therefore binds them.
A model of intersegmental coordination not having mechanical variables and parameters was presented by Daun-Gruhn
(2011), Daun-Gruhn and Tóth (2011), and Tóth et al. (2015).
This model was obtained by using the experimental data by
Borgmann et al. (2007 2009). In the majority of these experiments, all but a single front leg were amputated, in others, two
ipsilateral legs (e.g., front and middle leg) were left intact. In
both cases, functional connections, in particular, possible coordination between the PR systems of these legs were studied
J Neurophysiol • doi:10.1152/jn.00693.2015 • www.jn.org
Downloaded from http://jn.physiology.org/ by 10.220.32.247 on June 17, 2017
γ ( )
115
INTERLEG COORDINATION
905
tripod −−−> tetrapod
70
60
60
o
β( )
70
50
50
40
40
30
30
dβ/dt (1/s)
dβ/dt (1/s)
β (o)
tetrapod −−−> tripod
20
10
-10
2000
10
0
2100
2200
2300
2400
t (ms)
-10
2000
2100
2200
2300
2400
t (ms)
Fig. 21. Intervals in which the commands for transition are effective. Top curves: time course of the angles ␤ of the legs; bottom curves: time course of the
corresponding angular velocities d␤/dt. The horizontal lines, parallel to the time axes represent the threshold values for ␤ and d␤/dt, respectively. In the
simulations, the threshold value 5.0 1/s was used for d␤/dt and 31° for ␤.
(Borgmann et al. 2009). The results suggested that there were
descending excitatory and inhibitory connections between the
CPGs of the PR systems of the ipsilateral legs, modulated,
presumably, by peripheral (load) sensory signals. Thus a descending chain of connections was identified in the experiments. In the model, we completed this descending chain by an
additional connection from the meta- to the prothoracic PR
system. With the use of this model, tetrapod- and tripod-like
simulated electrical activity could be produced in the protractor-retractor CPGs. Also, the transition between the coordination patterns was successfully simulated. These earlier results
do not contradict our present ones, since, as pointed out, in the
experiments by Borgmann et al. (2007, 2009), at least one of
the three ipsilateral legs were amputated; hence, the LD and EF
systems of the absent legs, deprived from their vital peripheral
connections, were rendered ineffective. It is therefore no surprise that no connection, let alone coordination, could be
observed between the LD systems of the legs. Conversely, in
the intact animal, the connections between the PR systems of
the legs may well be masked by the much stronger ones
between the LD systems. We conjecture that the connections
between the PR systems of the different legs might play a part
in the mutual adjustment of the anterior and posterior end
positions of the different legs during general locomotion,
including coordination patterns other than just those of normal
walking.
The properties of our present model and the simulation
results that were achieved by using it can, in general, be
directly interpreted in physiological terms (e.g., observed coordination patterns and transitions between them, neuronal
mechanisms in the model and animal). However, the model
makes some predictions of a basic nature. First and foremost,
it predicts that the main components of the intersegmental
connections are between the levator-depressor neuronal control
networks of the individual segments. They are crucial in
producing locomotion, more precisely walking. A further important prediction is that the tripod coordination pattern requires sensory signals represented by the angle ␤ of the hind
and middle leg, while the tetrapod one also requires the angular
velocities d␤/dt of all three legs. Yet another prediction is how
the transition between the two coordination patterns (tetrapod
and tripod) takes place, including the timing of those transitions (cf. Fig. 21). These predictions have not been or could not
yet be checked, i.e., confirmed or rejected, in experiments.
Similarly, the details of the intersegmental connections in the
model await scrutiny in suitable experiments. We thus see the
main merit of our work in the predictions just mentioned. They
provide further goals for the experimental investigations in a
logically consistent theoretical framework. These predictions
are, or will become, practicable to be tested in experiments.
Finally, it is also worth briefly examining what implications
our results might have for the locomotion of other legged
animals. As already mentioned earlier, we adopted the idea
about the primary role of the LD systems in locomotion from
the lobster. It is, however, a reasonable assumption that our
results would not only apply to the lobster but also to the much
bigger family of crustaceans (cf. Orlovsky et al. 1999) or even
arthropods. To make a further and larger jump, as far as
four-legged vertebrates, and mammals in particular, are concerned, the details of the mechanisms underlying their walking
are much more complex than in insects and crustaceans (Orlovsky et al. 1999). A complete description and analysis of the
organization of locomotion at the level of local neuronal
networks have so far not been possible. There are, however,
functional relations between our results and walking in fourlegged mammals. In the latter, too, local neuronal networks
control the workings of groups of antagonistic muscle groups
providing for them the rhythmic excitation and inhibition
needed in walking. Also, proprioceptive sensory signals can
and do modify the basic walking pattern depending on the
environmental conditions. Interestingly, the middle and the
hind leg in our model perform “normal” walking pattern,
similar to those in four-legged animals, if the front leg is
isolated from the rest (Tóth and Daun-Gruhn, unpublished
results). This is in very good agreement with the aforementioned experimental results by Grabowska et al. (2012). Hence,
the same basic features of locomotion (walking) can be detected in a very broad spectrum of species. We would therefore
J Neurophysiol • doi:10.1152/jn.00693.2015 • www.jn.org
Downloaded from http://jn.physiology.org/ by 10.220.32.247 on June 17, 2017
0
20
906
INTERLEG COORDINATION
venture to suggest that the primary role of the neuromuscular
systems that bring about lifting and downward movement of
the leg may be a general property in many kinds of animals,
including legged mammals.
GRANTS
This work was supported by the Deutsche Forschungsgemeinschaft Grants
DA1182/1-1 and GR3690/4-1 (to S. Daun-Gruhn).
DISCLOSURES
No conflicts of interest, financial or otherwise, are declared by the author(s).
AUTHOR CONTRIBUTIONS
REFERENCES
Akay T, Ludwar BC, Göritz ML, Schmitz J, Büschges A. Segment
specificity of load signal processing depends on walking direction in the
stick insect leg muscle control system. J Neurosci 27: 3285–3294, 2007.
Ayres JL, Davis W. Neuronal control of locomotion in the lobster, Homarus
americanus. J Comp Physiol A 115: 1–27, 1977.
Bässler U. Sensory control of leg movement in the stick insect Carausius
morosus. Biol Cybern 25: 61–72, 1977.
Bässler U. Neural Basis of Elementary Behavior in Stick Insects. Berlin,
Germany: Springer Verlag, 1983.
Bässler U, Stein W. Contributions of structure and innervation pattern of the
stick insect extensor tibiae muscle to the filter characteristics of the musclejoint system. J Exp Biol 199: 2185–2198, 1996.
Borgmann A, Scharstein H, Büschges A. Intersegmental coordination: influence of a single walking leg on the neighboring segments in the stick
insect walking system. J Neurophysiol 98: 1685–1696, 2007.
Borgmann A, Hooper SL, Büschges A. Sensory feedback induced by
front-leg stepping entrains the activity of central pattern generators in caudal
segments of the stick insect walking system. J Neurosci 29: 2972–2983,
2009.
Borgmann A, Tóth TI, Gruhn M, Daun-Gruhn S, Büschges A. Dominance
of local sensory signals over inter-segmental effects in a motor system:
experiments. Biol Cybern 105: 399 – 411, 2011.
Büschges A. Role of local nonspiking interneurons in the generation of
rhythmic motor activity in the stick insect. J Neurobiol 27: 488 –512, 1995.
Büschges A. Inhibitory synaptic drive patterns motoneuronal activity in
rhythmic preparations of isolated thoracic ganglia in the stick insect. Brain
Res 783: 262–271, 1998.
Büschges A. Sensory control and organization of neural networks mediating
coordination of multisegmental organs for locomotion. J Neurophysiol 93:
1127–1135, 2005.
Büschges A, Gruhn M. Mechanosensory feedback in walking: from joint
control to locomotor patterns. Adv Insect Physiol 34: 193–230, 2008.
Cruse H. A quantitative model of walking incorporating central and peripheral
influences. Biol Cybern 37: 137–144, 1980.
Cruse H. What mechanisms coordinate leg movement in walking arthropods?
Trends Neurosci 13: 15–21, 1990.
Cruse H, Kindermann T, Schumm M, Dean J, Schmitz J. Walknet–a
biologically inspired network to control six-legged walking. Neural Netw
11: 1435–1447, 1998.
Daun-Gruhn S. A mathematical modeling study of inter-segmental coordination during stick insect walking. J Comput Neurosci 30: 255–278, 2011.
Daun-Gruhn S, Tóth TI. An inter-segmental network model and its use in
elucidating gait-switches in the stick insect. J Comput Neurosci 31: 43– 60,
2011.
Daun-Gruhn S, Tóth TI, Büschges A, Borgmann A. Dominance of local
sensory signals over inter-segmental effects in a motor system: modeling
studies. Biol Cybern 105: 413– 426, 2011.
J Neurophysiol • doi:10.1152/jn.00693.2015 • www.jn.org
Downloaded from http://jn.physiology.org/ by 10.220.32.247 on June 17, 2017
Author contributions: T.I.T. and S.D.-G. conception and design of research;
T.I.T. performed experiments; T.I.T. analyzed data; T.I.T. and S.D.-G. interpreted results of experiments; T.I.T. prepared figures; T.I.T. drafted manuscript; T.I.T. and S.D.-G. edited and revised manuscript; S.D.-G. approved
final version of manuscript.
Delcomyn F. Insect locomotion on land. In: Locomotion and Energetics in
Arthropods, edited by Herreid CF, Fourtner CR. New York: Plenum, 1981,
p. 103–125.
Dürr V, Schmitz J, Cruse H. Behaviour-based modelling of hexapod locomotion: linking biology and technical application. Arthropod Struct Dev 33:
237–250, 2004.
Ekeberg O, Blümel M, Büschges A. Dynamic simulation of insect walking.
Arthropod Struct Dev 33: 287–300, 2004.
Gabriel JP, Scharstein H, Schmidt J, Büschges A. Control of flexor
motoneuron activity during single leg walking of the stick insect on an
electronically controlled treadwheel. J Neurobiol 56: 237–251, 2003.
Grabowska M, Godlewska E, Schmidt J, Daun-Gruhn S. Quadrupedal gaits
in hexapod animals–inter-leg coordination in free-walking adult stick insects. J Exp Biol 215: 4255– 4266, 2012.
Graham D. A behavioural analysis of the temporal organisation of walking
movements in the 1st instar and adult stick insect (Carausius morosus). J
Comp Physiol 81: 23–52, 1972.
Graham D. Pattern and control of walking in insects. Adv Insect Physiol 18:
31–140, 1985.
Guschlbauer CH, Scharstein H, Büschges A. The extensor tibiae muscle of
the stick insect: biomechanical properties of an insect walking leg muscle.
J Exp Biol 210: 1092–1108, 2007.
Hughes GM. The co-ordination of insect movements. I. The walking movements of insects. J Exp Biol 29: 167–285, 1952.
Knops S, Tóth TI, Guschlbauer C, Gruhn M, Daun-Gruhn S. A neuromechanical model for the neuronal basis of curve walking in the stick insect. J
Neurophysiol 109: 679 – 691, 2013.
Laurent G, Burrows M. Distribution of intersegmental inputs to nonspiking
local interneurons and motor neurons in the locust. J Neurosci 9: 3019 –
3029, 1989a.
Laurent G, Burrows M. Intersegmental interneurons can control the gain of
reexes in adjacent segments of the locust by their action on nonspiking local
interneurons. J Neurosci 9: 3030 –3039, 1989b.
Orlovsky GN, Deliagina TG, Grillner S. Neuronal Control of Locomotion.
From Mollusc to Man. Oxford, UK: Oxford Univ. Press, 1999.
Schilling M, Cruse H, Arena P. Hexapod walking: an expansion to Walknet
dealing with leg amputations and force oscillations. Biol Cybern 96: 323–
340, 2007.
Schilling M, Hoinville T, Schmitz J, Cruse H. Walknet, a bio-inspired
controller for hexapod walking. Biol Cybern 107: 397– 419, 2013.
Schmitz J. The depressor trochanteris motoneurones and their role in the
coxa-trochanteral feedback loop in the stick insect Carausius morosus. Biol
Cybern 55: 25–34, 1986a.
Schmitz J. Properties of the feedback system controlling the coxa-trochanter
joint in the stick insect Carausius morosus. Biol Cybern 55: 35– 42, 1986b.
Schmitz J, Schneider A, Schilling M, Cruse H. No need for a body model:
positive velocity feedback for the control of an 18-DOF robot walker. Appl
Bionics Biomech 5: 135–147, 2008.
Tóth TI, Knops S, Gruhn S. A neuro-mechanical model explaining forward
and backward stepping in the stick insect. J Neurophysiol 107: 3267–3280,
2012.
Tóth TI, Schmidt J, Büschges A, Gruhn S. A neuro-mechanical model of a
single leg joint highlighting the basic physiological role of fast and slow
muscle fibres of an insect muscle system. PLoS One 8: e78247, 2013a.
Tóth TI, Grabowska M, Schmidt J, Büschges A, Gruhn S. A neuromechanical model explaining the physiological role of fast and slow muscle
fibres at stop and start of stepping of an insect leg. PLoS One 8: e78246,
2013b.
Tóth TI, Grabowska M, Rosjat N, Hellekes K, Borgmann A, Daun-Gruhn
S. Investigating inter-segmental connections between thoracic ganglia in the
stick insect by means of experimental and simulated phase response curves.
Biol Cybern 109: 349 –362, 2015.
von Twickel A, Büschges A, Pasemann E. Deriving neural network controllers from neuro-biological data: implementation of a single-leg stick insect
controller. Biol Cybern 104: 95–119, 2011.
von Twickel A, Hild M, Siedel T, Patel V, Pasemann F. Neural control of
a modular multi-legged walking machine: simulation and hardware. Robot
Auton Syst 60: 227–241, 2012.
Westmark S, Oliveira EE, Schmidt J. Pharmacological analysis of tonic activity in
motoneurons during insect walking. J Neurophysiol 102: 1049–1061, 2009.
Wilson DM. Insect walking. Ann Rev Entomol 11: 103–122, 1966.