Robotics and Autonomous Systems 60 (2012) 7282

Contents lists available at SciVerse ScienceDirect

Robotics and Autonomous Systems

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

Effects of turning gait parameters on energy consumption and stability of a

six-legged walking robot
Shibendu Shekhar Roy a , Dilip Kumar Pratihar b,
a
Department of Mechanical Engineering, National Institute of Technology, Durgapur, India
b
Department of Mechanical Engineering, Indian Institute of Technology, Kharagpur, India

article info abstract

Article history: Minimization of energy consumption plays a key role in the locomotion of a multi-legged robot used for
Received 30 June 2010 various purposes. Turning gaits are the most general and important factors for omni-directional walking
Received in revised form of a six-legged robot. This paper presents an analysis on energy consumption of a six-legged robot during
12 August 2011
its turning motion over a flat terrain. An energy consumption model is developed for statically stable
Accepted 30 August 2011
Available online 8 September 2011
wave gaits in order to minimize dissipating energy for optimal feet forces distributions. The effects of gait
parameters, namely angular velocity, angular stroke and duty factors are studied on energy consumption,
Keywords:
as the six-legged robot walks along a circular path of constant radius with wave gait. The variations
Six-legged robot of average power consumption and energy consumption per unit weight per unit traveled length with
Turning gait the angular velocity and angular stroke are compared for the turning gaits of a robot with four different
Wave gait duty factors. Computer simulations show that wave gait with a low duty factor is more energy-efficient
Gait parameters compared to that with a high duty factor at the highest possible angular velocity. A stability analysis based
Energy consumption on normalized energy stability margin is performed for turning motion of the robot with four duty factors
Normalized energy stability margin for different angular strokes.
2011 Elsevier B.V. All rights reserved.

1. Introduction
Legged robots are preferred to wheeled robots to move through
environments which generally contain some irregularities. In such
an environment, legged robots offer better mobility than their
wheeled counterparts do. However, legged locomotion has the dis-
advantage of achieving poor energy efficiency [1]. An autonomous
walking robot cannot function satisfactorily with low energy ef-
ficiency due to the fact that it has to carry all driving and con-
trol units in addition to its trunk body and payload. Long duration
missions like exploration of planets [2], underground mining, lo-
cating and deactivating of bombs are subjected to power supply
constraints. The minimization of power consumption plays a major
role in the locomotion of an autonomous multi-legged robot used
for service applications. A reduction in energy consumption results
in robots, which can not only travel more, but also require smaller
actuators that typically yield a reduction in the robots weight and
cost.
During locomotion, a multi-legged robot might have to move
along straight paths, ascend and descend sloping terrains, take
turns to avoid various obstacles as the situation demands. Out of
Corresponding author. Tel.: +91 3222 282992; fax: +91 3222 282278.
E-mail addresses: ssroy99@yahoo.com (S.S. Roy), dkpra@mech.iitkgp.ernet.in
(D.K. Pratihar).
0921-8890/$ see front matter 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.robot.2011.08.013
all these tasks, the present study concentrates on energy efficiency
analysis for generating turning motions only. Turning gaits [3] are
the most general gaits for omni-directional walking robots [4].
These are much more complicated compared to straight-forward
wave gaits. The dynamics and energy consumption model for
generating the turning motion of a six-legged robot are also more
complex compared to those of other gaits.
Four different approaches were used by various investigators
to obtain energy efficient and statically stable multi-legged robots.
Those approaches include (i) design of energy efficient mechanical
leg structure [5,6] and leg distribution [7]; (ii) employment of
energy storage devices to recover energy [8,9]; (iii) optimal
selection of gait parameters [10,11]; and (iv) optimal solution to
foot force distribution [1216]. Lapshin [17] proposed an energy
efficient model of a walking machine, and some results were
obtained from the standpoint of gait-parameters only. Orin and
Oh [18] tried to resolve the foot force distribution for minimum
energy consumption and load balance between several legs. They
simplified the friction cone constraint to eliminate the associated
nonlinearity by inscribing a pyramid within the desired friction
cone. Because of this simplification, the solution was conservative.
Early work on minimizing power consumption considered the
actuator dynamics [19]. However, the actuators considered were
not the DC servomotors. Nahon and Angeles [20] used quadratic
programming to minimize power consumption of robotic systems
actuated by DC motors, after considering power regeneration due
to negative work.
 S.S. Roy, D.K. Pratihar / Robotics and Autonomous Systems 60 (2012) 7282
Marhefka and Orin [21] utilized quadratic programming to
solve foot force distribution in hexapod walking robots that
minimizes the power consumption in DC motors. In their
work, gains from power regeneration by the DC motors were
not considered in the optimization problem. The authors also
performed an analysis of average power consumption with respect
to the parameters of wave gaits, based on a simulation model
of a hexapod robot [10]. Kar et al. [22] performed an analysis of
energy efficiency with respect to structural parameters, interaction
forces, friction coefficient and duty factor of wave gaits, based
on a simplified model of a six-legged robot. Kar et al. [22] and
Lin and Song [23] considered instantaneous power to be the
product of instantaneous joint torques and joint velocities. Such
models ignored the fact that a considerable amount of power is
dissipated in the joints of the support legs. In order to eliminate
such drawbacks, the integral of the sum of the squares of joint
torques was considered as a criterion of dissipated energy in the
actuators. Erden and Leblebicioglu [24,25] utilized the modified
simplex method along with Lemkes Complementary Pivoting
Algorithm to compute optimum foot force and torque distributions
by considering a more practical locomotion performance objective,
that is, minimization of energy dissipation. Nishii [26] used the
integral of weighted sum of the product of instantaneous joint
torques and joint velocities, and the sum of squares of joint
torques as the energetic cost, which was analyzed with respect
to duty factor and velocity in the walking of a two joint six-
legged robot. Arikawa and Hirose [27] addressed the relationship
between power consumption and feet positions for a quadruped
robot. Silva et al. [28], Zhoga [29] and Zelinska [30] analyzed energy
expenditure and energy efficiency of multi-legged locomotion
systems by taking leg dynamics and torque into account, but
they did not consider the type of joint actuator, although its
contribution to energy consumption is significant. Moreover, they
focused their study on walking along a straight path only. Recently,
Gonzalez de Santos et al. [31] studied the energy required for a
six-legged robot using alternating tripod gaits and also derived
a method to minimize the energy consumption of the said robot
(i.e., SILO6) on irregular terrain. More recently, the locomotion of
symmetric hexapods, i.e., hexagonal six-legged robots has been
studied by Wang et al. [32]. The proposed gaits for the said robots
have been compared with those of rectangular robots from the
stability, terrain adaptability and turning ability points of view.
The studies on turning and crab gaits are relatively few
compared to those on straight-forward gait, though these are
important for omni-directional walking robots [33]. In this
connection, the works of Estremera and Gonzalez de Santos [34]
and Yang [35,36] are worth mentioning. Orin [37] analyzed the
turning motion of a hexapod robot from a kinematics point of
view. Hirose et al. [38], and Zhang and Song [39,40] adopted
analytical and graphical methods for stability analysis of spinning
gaits and circling gaits of a quadruped walking machine. Miao
and Howard [41] described an algorithm for generating a tripod
turning gait, which could maximize the rotation angle and also
optimize the stability. Recently, Estremera et al. [42] proposed a
new gait generation algorithm for the development of free crab and
turning gaits for hexapod robots on natural terrains. A non-periodic
feature was included in a conventional tripod gait to obtain a real
free gait, which worked based on some heuristic rules capable of
planning leg motions accurately and guaranteeing stability. The
problem of optimal turning gait generation of a six-legged robot
had been solved by Pratihar et al. [43] using a combined genetic
algorithm and fuzzy logic approach. Pratihar et al. [44] extended
this work to find simultaneous optimal path and gait generation of
a hexapod walking robot, but they considered a simplified model
of the robot. Moreover, they did not consider a detailed kinematics
and dynamic behavior of the leg and trunk body, although its
contribution to gait generation was significant. Due to the inherent
73
complexity of a realistic walking robot, it is not an easy task to
include inertial terms in the modeling. Most of the studies on
walking robot dynamics were conducted with simplified models
of legs and body. But, in order to have a better understanding
of its walking, dynamics and other important issues of walking,
such as dynamic stability, energy efficiency and its on-line control;
kinematics and dynamic models based on a realistic walking robot
design are necessary. To the best of the authors knowledge, no
significant study is reported on energy efficiency and stability
analysis related to turning gaits of a realistic six-legged robot.
In the present paper, attempts are made to study the effects of
gait parameters on energy consumption and stability of a real six-
legged robot.
The remaining part of this paper is organized as follows: A
few terminologies related to turning gait are defined in Section 2.
Section 3 deals with mathematical formulation of the problem. The
developed approaches are also discussed in this section. Results
are stated and discussed in Section 4. Some concluding remarks
are made in Section 5. The scopes for future study are indicated in
Section 6.
2. Terminologies related to turning gaits
Let us first review the following terms related to turning
gaits [38,40] of a multi-legged robot:
Transfer or swing phase: Transfer phase of a leg is the period
during which the foot is not on the ground.
Support phase: It is the period during which the foot is placed
on the ground.
Duty factor (): It is the time fraction of a cycle time (T ), in
which a particular leg is in the support phase, that is, =
ts /(ta + ts ), where ta and ts denote transfer and support times,
respectively.
Gravity center trajectory: It is the locus of the gravity center of
the robots body with respect to the ground.
Support trajectory of leg i: It is the locus of the foot-tip of leg
i relative to the robot body in the support phase.
Leg angular stroke (i ): It is the rotation angle of the foot-tip of
leg i about the turning center relative to the robot body during
the support phase.
Gait angular stroke (): It is the rotation angle of the gravity
center about the turning center relative to the ground in the
support phase of a leg. The maximum is the minimum of all
the largest available i s. For a periodic and regular turning gait,
all the leg angular strokes should be identical and equal to the
gait angular stroke. The gait angular stroke or angular stroke is
calculated as = ts , where is angular speed of the CG of the
robot.
Angular stride ( ): It is the rotation angle of the gravity center
about the turning center relative to the ground in one complete
turning locomotion cycle. For a periodic and regular turning
gait, = .
3. Mathematical formulation
The problem may be stated as follows: A six-legged robot has
to plan its gait parameters during turning in such a way that it can
tackle that situation by consuming a minimum amount of energy
after ensuring optimal feet forces distributions and satisfying the
conditions of static stability.
The following assumptions are made in the present study:
(a) Center of gravity (CG) of the trunk body is assumed to be
located at the geometric center of the body. The body reference
frame {B} is attached at CG of the body.
74 S.S. Roy, D.K. Pratihar / Robotics and Autonomous Systems 60 (2012) 7282

Table 1
DH parameters for three joint legs.
Link ai i di i
1 L1 90 0 1
2 L2 0 0 2
3 L3 0 0 3
+ for left side legs, for right side legs

Fig. 1. Six-legged walking robot model.

Fig. 3. A schematic showing top view of the six-legged robot walking in a circular
Fig. 2. Frame assignment as per DH notations. path.

(b) The trunk body turns in a clockwise direction with a constant where px = [L1 + L2 C 2 + L3 C (2 + 3 )]C 1; py = [L1 + L2 C 2 +
angular speed (). L3 C (2 + 3 )]S 1 ; pz = L2 S 2 + L3 S (2 + 3 ); sin and cos are
(c) During turning, the trunk body is kept at a constant height from represented by S and C , respectively.
the ground and turning radius is also kept constant. The six legs and trunk body must be integrated to solve the
(d) The joint actuators are DC geared motors, which cannot store kinematics problem of the robot. The body attached reference
negative energy. Therefore, any negative energy, that is, gain in frame {B} is located at the geometric center of the trunk body
energy supplied by external forces is lost. as shown in Fig. 3. The body reference frame {B} is represented
In this study, kinematics and dynamic models of a realistic with respect to global reference frame {G} attached at the turning
six-legged robot are considered to analyze complex relationships center, and the hip reference frame of ith leg{0i } is denoted in the
among gait parameters and its power consumption. global reference frame using the transformation matrix as given
below.
3.1. Kinematics model of six-legged robot C (t )

S (t ) 0 rG S (t )

G S (t ) C (t ) 0 rG C (t )
Fig. 1 shows a 3-D model of a real six-legged robot consid- TB = (2)
0 0 1 hG
ered in the present study. The leg mechanism of the robot is pre-
0 0 0 1
sented in Fig. 2. Each leg has three degrees of freedom and is
composed of three links connected by three rotary joints. The De- C (t ) S (t ) ri S (i + t )

0
navitHartenberg (D-H) homogenous matrix representation [45] G S (t ) C (t ) 0 ri C (i + t )
(refer to Fig. 2) is used to describe the spatial displacement be- T0,i = ;
0 0 1 hG + hib
tween neighboring link coordinate frames to obtain the kinematics 0 0 0 1
information of each leg. The DH parameters, namely link length i = leg number (1 to 6). (3)
(ai ), link twist (i ), joint distance (di ), and joint angle (i ), required
to completely describe three joint legs are given in Table 1. Here, ri is the path radius of hip of ith leg, which can be determined
The foot tip reference frame {3} can be expressed in the hip or as follows:

leg reference frame {0} as given below. 2 2
Lw Lb
3 r1 = r5 = rG + + ;
2 2

0 i1
T3 = 0 T1 1 T2 2 T3 = Ti

i =1 2 2
Lw Lb
C 1 C (2 + 3 ) C 1 S (2 + 3 ) S 1 r2 = r6 = rG + ;

px
2 2
0 S 1 cos(2 + 3 ) S 1 S (2 + 3 ) C 1 py
T3 = (1)
S (2 + 3 ) C (2 + 3 ) 0 pz Lw Lw
r3 = rG + ; r4 = rG ,
0 0 0 1 2 2
 S.S. Roy, D.K. Pratihar / Robotics and Autonomous Systems 60 (2012) 7282 75

a b

c d

Fig. 4. Gait diagrams for (a) duty factor = 1/2, (b) duty factor = 2/3, (c) duty factor = 3/4, (d) duty factor = 5/6.

Lb The joint rate and joint acceleration equations of for each leg during
1 = tan1 ; 5 = 1 ;
2rG + Lw the support phase can be expressed as follows:

Lb = J1 p, (10)
2 = tan 1
; 6 = 2 ; 3 = 4 = 0
2rG Lw
= J (p J),
1
(11)
where rG is the turning radius of the CG of the trunk body, is the
where the position vector p = [px py pz ]T , = [1 2 3 ]T and J is
peripheral speed of the CG of the robot, t is the time, Lb is the length the Jacobian matrix [45].
of the trunk body and Lw is the width of the trunk body. Standard wave gaits [1] with duty factors () equal to
The joint angles 1 , 2 and 3 can also be determined through 1/2, 2/3, 3/4 and 5/6 are considered in this study (refer to
inverse kinematics [45] calculations, as follows: Fig. 4). After solving the kinematics problem, it is necessary to
1 = a tan 2(py , px ), (4) build a dynamic model of the robot to evaluate the foot forces,
joint torques and energy consumption, which is discussed in the
2 = a tan 2 c , a2 + b2 c 2 a tan 2(a, b), (5) following sub-sections.

3.2. Dynamic model of six-legged robot

where a = 2L2 ( p2x + p2y L1 ); b = 2pz L2 ; c = [( p2x + p2y
L1 )2 + p2z + L22 L23 ],
2
p2x + p2y L1 + p2z L22 L23
3 = cos1 . (6)
2L2 L3
To ensure a smooth functioning, each joint trajectory of the
swing leg is assumed to follow a fifth-order polynomial in time (t).
The jth joint of a swing leg, that is, j can be represented in fifth-
order polynomial as follows:
j = aj0 + aj1 t + aj2 t 2 + aj3 t 3 + aj4 t 4 + aj5 t 5 ,
where aj0 , aj1 , aj2 , aj3 , aj4 , and aj5 are the coefficients, whose values
are determined using a set of boundary conditions defined over the
swing phase and j=1, 2, 3 joints. The boundary conditions of joint
angles, joint rates and joint accelerations at initial and final points
of the trajectory are applied to determine the six coefficients for
each joint. The joint rate and joint acceleration equations of each
joint of a swing leg can be obtained using the following equations:
j = aj1 + 2aj2 t + 3aj3 t 2 + 4aj4 t 3 + 5aj5 t 4
j = 2aj2 + 6aj3 t + 12aj4 t + 20aj5 t .
2 3
In order to derive dynamic equations and for finding joint
torques variations over a locomotion cycle, LagrangeEuler for-
mulation is used. The direct application of Lagrangian dynamics
formulation together with DenavitHartenbergs link coordinate
representation results in a convenient, compact, systematic al-
gorithmic description of the equations of motion. A systematic
derivation of LagrangeEuler equations yields a dynamic expres-
sion that can be written in the vector-matrix form as given below.
i = [M() + H(, ) + G()]i JTi Fi , (12)
where M() is the 3 3 mass matrix of the leg, H is a 3 1
vector of centrifugal and Coriolis terms, G() is a 3 1 vector of
(7)
gravity terms, i is the 3 1 vector of joint torques and Fi is the
3 1 vector of ground reaction forces of ith foot. During the legs
swing phase, there is no footterrain interaction, and Fi becomes
equal to zero. However, during the support phase, ground contact
exists and Eq. (12) becomes undetermined, which has to be solved
using an optimization criterion, e.g., optimal foot force distribution.
The derivation of dynamic equations for the swing leg is presented
in [46].
For computing foot-force distributions, the following assump-
(8) tions are made:
(i) The ground legs are assumed to be supporting the trunk body
(9) without any slippage on their tip points.
76 S.S. Roy, D.K. Pratihar / Robotics and Autonomous Systems 60 (2012) 7282

(ii) The contacts of the tip of the feet with ground can be This matrix defines the position of tip of a ground foot i (i = p, q, r
modeled as hard point contacts with friction, which indicates that for = 1/2; i = p, q, r , s for = 2/3 and = 3/4; i =
the interaction between the tip of the leg and ground is limited to p, q, r , s, t for = 3/4 and = 5/6) or that of the center of gravity
three components of force: one normal and two tangential to the (i = c) with respect to the body reference frame. The coordinates
surface. of ith footground contact point with respect to the body refer-
(iii) The effect of inertia of swing legs on trunk body is assumed ence frame, located at the bodys geometric center, are denoted by
to be negligible for simplicity. Therefore, the robots center of (xi , yi , zi ).
gravity does not change with leg movements; transfer legs exert The values of Fx , Fy , Fz , Mx , My and Mz are to be found for
no forces on the trunk body. In the present study, this problem of turning motion as given below.
foot force distribution is solved using two approaches as explained
below. Fx = FIx ; Fy = mrG 2 ; Fz = mgz ; and
Approach 1: minimization of norm of feet forces d
Mx = B Ixz + B Iyz 2 ;
Let us assume that Fi = [fix , fiy , fiz ]T is the ground-reaction force dt
vector on foot i, where i = 1, 2, . . . , n; n is the number of ground d
legs at a particular instant [for example, n = 3 for tripod gait and My = B Iyz B Ixz 2 ;
dt
n = 4 for tetrapod gait]. The wrench W = [Fx , Fy , Fz , Mx , My , Mz ]T d
contains the forces (Fx , Fy , Fz ) and moments (Mx , My , Mz ) acting on Mz = B Izz
the robots center of gravity, and represents the robots payload, dt
including the effect of surface gradient, any externally applied where m(=3.5 kg) is the total mass of the robot including payload,
forces and inertial effects of the robots body. However, the inertial FIx is inertia force component.
effects of the legs are neglected to simplify the study. Under With the known feet positions, the feet forces during a whole
these conditions, six equilibrium equations that balance forces and locomotion cycle can be computed using Eq. (19), which is
moments can be obtained as follows (for n number of ground legs): indeterminate, because it consists of six equations but there are
n
more than six unknowns. The solution of Eq. (19) is obtained using
the least squared method, which gives the minimum norm solution

fix + Fx = 0 (13)
i =1
of the indeterminate equilibrium equations.
n
Approach 2: Minimization of norm of joint torques
fiy + Fy = 0 (14) In this approach, the Eq. (19) can be re-formulated by using the
i =1
following relations.
n
[F] = [D].[] (20)
fiz + Fz = 0 (15) p
J 03 03

for = 1/2;
i =1 q
where [D] = 03 J 03
n n r
03 03 J 99
yi fiz zi fiy + yc Fz zc Fy + Mx = 0 (16) p
i =1 i =1
J 03 03 03
q
03 J 03 03
n n [D ] = r for = 2/3;
03 03 J 03
zi fix xi fiz + zc Fx xc Fz + My = 0 (17) s
03 03 03 J 1212
i =1 i =1
p
n n J 03 03 03 03
q
xi fiy yi fix + xc Fy yc Fx + Mz = 0. (18) 03 J 03 03 03
for = 5/6.
r
[D] = 03 03 J 03 03

i =1 i=1
03 s
03 03 J 03
These Eqs. (13)(18) can be written in a matrix form as follows:
03 03 03 03 t J 1515
[A].[F] = [B].[W] (19)
For = 3/4, during the periods: (0 to 3T/12) and (6T/12 to 9T/12),
where
[ ] [D] will have the dimensions of 15 15 (the same as that with
[A] =
I3 I3 I3
for = 1/2; = 5/6), whereas during the remaining period, [D] will have the
Rp Rq Rr
69
dimensions of 12 12 (the same as that with = 2/3).
T 1
[
I3 I3 I3 I3
] i
J =Ji ; Ji is the (3 3) Jacobian matrix of leg i. Here,
[A] = for = 2/3 [] = [p , q , r ]T for = 1/2; [] = [p , q , r , s ]T for = 2/3
Rp Rq Rr Rs
612
[ ] and = 3/4; [] = [p , q , r , s , t ]T for = 3/4 and = 5/6;
I3 I3 I3 I3 I3
[A] = for = 5/6. and i = [i1 , i2 , i3 ]T is the torque vector containing three joint
Rp Rq Rr Rs Rt
615 torques at leg i.
For = 3/4, during the periods: (0 to 3T/12) and (6T/12 to 9T/12), The Eq. (19) can be re-written as follows:
[A] will have the dimensions of 6 15 (the same as that with [A].[D].[] = [B].[W] (21)
= 5/6), whereas during the remaining period, [A] will have the
dimensions of 6 12 (the same as that with = 2/3). [AJ ].[] = [B].[W]. (22)
[
I3 03
] The minimum norm solution of the above indeterminate equations
[B] = . is obtained using a least squared method.
Rc I3
66

I3 is the (3 3) identity matrix, 03 is the (3 3) null matrix and 3.3. Power consumption of a six-legged robot
Ri is the (3 3) skew-symmetric matrix of vector [xi , yi , zi ]T .

0 zi yi

1 0 0
The energy consumption of a legged robot is mainly due to the
Ri = zi 0 xi and I3 = 0 1 0 . energy consumed by an actuator at each joint of the legs. As a joint
y i xi 0 0 0 1 is driven by a DC motor, the consumed energy in the motor during
 S.S. Roy, D.K. Pratihar / Robotics and Autonomous Systems 60 (2012) 7282 77

Thus, average power consumption can be calculated as follows:

T 6
3
E 1
P = = [(ij ij ) + ij ]dt .
2
(33)
T T 0 i =1 j =1

The first term denotes positive mechanical power and the second
term indicates the energy dissipation per unit time, which is
proportional to the sum of the squares of joint torques.
For a legged robot, one of the most challenging problems is to
perform a task with minimum energy consumption. The total en-
ergy consumption for a given task is more important than the aver-
Fig. 5. Energy model of a DC motor. age power consumption. The main objective of the present study is
to minimize the energy expenditure for a fixed distance travel. The
a time (T ) is given by: energy consumption depends on the torque and velocity of each
T joint, which, in turn, depends on various walking parameters re-
E= ua idt , (23) lated to the gait, vehicle and terrain conditions. Gait parameters
0 include duty factor, angular stroke, phase difference, cycle time
where ua is the applied voltage and i is the armature current. and foothold location. Vehicle parameters are related to weight,
The behavior of DC motors (refer to Fig. 5) can be explained by height and angular velocity of the trunk body. Terrain parameters
its torque and voltage equations as mentioned below [45]. take care of the slope and friction. In the present work, weight
m = Kt i (24) and height of the trunk body are kept constant for all duty factors.
Therefore, energy consumption depends on angular stroke, angu-
ue = Kv m (25)
lar velocity and duty factor. Since energy is the integral of power
di consumption over time, minimization of energy results if the aver-
ua = ue + Ri + L , (26)
dt age power consumption is minimized. It may be possible that a task
where m is the torque (N m) at the rotor, Kt is the torque constant, is accomplished in a shorter time with higher average power con-
Kv is the voltage constant, m is angular velocity of the rotor, R is sumption, so that the total dissipated energy becomes less. There-
the armature resistance, L is armature inductance, ue is the induced fore, the energy consumption per unit length of travel is more
voltage in the armature windings opposing the applied voltage. relevant for the walking of multi-legged robots. In the past, spe-
The output angular velocity ( ) and torque ( ) of the geared cific resistance [47], [23], that is, energy consumed per unit weight
motor are computed as follows: and per unit traveled length was used to compare the energy effi-
ciencies of various types of locomotion of legged robot. The same

m = , (27) is considered as the index of energy efficiency in our study.
Gs
m = Gs , (28) 3.4. Normalized energy stability margin
where Gs is the speed ratio of the geared motor.
The inductance of the winding is usually neglected. Thus, One of the most popular stability criteria for multi-legged robot
Eq. (23) can be re-written as follows: is Static Stability Margin (SSM), which was proposed by McGhee
and Frank [49]. SSM is defined for a given support polygon as the
T T T
RG2s

distances from the projection of CG to the edges of the support
E= (ue + Ri)idt = 2 dt . dt +
(29)
0 Kt2 0 0 polygon. This stability criterion is based on geometric concepts
The first term is mechanical energy and the second term represents and is independent of CG height. A better method of stability
energy loss by heat emissions. Although a negative value for measurement through Energy Stability Margin (ESM) was given by
the first term, that is, mechanical energy indicates a gain in Messuri and Klein [50], which is defined as the minimum potential
energy supplied by external forces, the DC motor cannot store this energy required for tumbling the robot around the edges of support
energy [47,48]. Therefore, the energy consumed by the DC motor polygon. It can be expressed as follows:
during time T is given by the following expression: ns
T T SESM = min(mghi ), (34)
RG2s

i
E= [( )]dt + 2 dt , (30)
0 0 Kt2 where i denotes the segment of support polygon considered as the

if > 0 rotation axis, ns is the number of support legs, and hi is the variation
where ( ) = 0 if 0
. of CG height during the tumble, which is given as hi = |Ri |(1
The energy consumed by the motor of jth joint of ith leg during cos ) cos , where Ri is the distance from the CG of the robot
a time T is given by the following expression: to the rotation axis, is the angle that Ri forms with the vertical
T T plane, and is the inclination angle of the rotation axis relative
RG2s

Eij = [(ij ij )]dt + (31) ij2 dt . to the horizontal plane. Hirose et al. [51] normalized the ESM with
0 Kt2 0 respect to the robots weight and proposed the Normalized Energy
Now, total energy consumed by all motors in a six-legged robot can Stability
be determined as follows:
SESM ns
T 6
3 T 6
3
RG2
Margin (NESM) as SNESM = = min(hi ). (35)
E= [(ij ij )]dt + s
ij2 dt mg i
0 i=1 j=1 0 i =1 j =1
Kt2
T 6
3 4. Results and discussion
E= [(ij ij ) + ij2 ]dt , (32)
0 i=1 j=1 In this section, simulation results related to overall energy
RG2s
consumption and stability of a six-legged robot during its walking
where = . along a circular path are discussed in detail. Table 2 shows the
Kt2
78 S.S. Roy, D.K. Pratihar / Robotics and Autonomous Systems 60 (2012) 7282

Table 2
Physical parameters of the six-legged robot.
Body parameters

Dimensions (103 m) Length (Lb ) 440

Width (Lw ) 160
Height (Lh ) 80
Moment of inertia (104 kg m2 ) B
Ixx 50
B
Iyy 270
B
Izz 260
Mass of the body including payload (kg) mb 1.712
Leg parameters
Link parameters Link 1 Link 2 Link 3
Mass (kg) mi 0.152 0.04 0.106
Length (103 m) Li 85 115 100
Position of center of mass (103 m) x 71.22 71.40 97.33
y 14.04 2.47 0.98
z 0.00 8.21 3.43
Moment of inertia (104 kg m2 ) Ixx 1.00 0.23 0.22
Iyy 8.28 3.07 10.00
Izz 9.09 2.91 10.01
Product of inertia (104 kg/m2 ) Ixy 1.57 0.141 0.103
Ixz 0.113 0.364 0.376
Iyz 0.037 0.018 0.0036

Table 3 be proportional to average dissipated power (average heat loss)

Average values of the squares of joint torques during turning motion with different
duty factors.
Duty factor ( ) Average of the squares of joint torques (N m)2
Approach 1
1/2 7.2513
2/3 5.5217
3/4 4.9860
5/6 4.5577
Angular stroke = 8 deg., Angular velocity = 2 deg./s
Height of trunk body = 0.13 m, Turning radius = 1 m
physical parameters of a real six-legged robot used for simulation.
The values of moment of inertia and positions of center of gravity
of this real robot have been computed using CATIA CAD/CAE
software [52]. Fig. 6 illustrates a flowchart describing overall
procedure adopted in computer simulations for the said robot. In
this simulation, turning radius and body height is assumed to be
equal to 1.0 m and 0.13 m, respectively.
Table 3 shows the average values of the squares of joint torques
of the robot for generating wave-turning gait patterns with various
duty factors, as obtained by approaches 1 and 2. Results indicate
that the average value of the squares of joint torques during one
complete locomotion cycle decreases with the increase of duty
factor for both the approaches. The average value of the squares
of joint torques of the robot as obtained by Approach 1 is seen
to be higher than that yielded by Approach 2 for all duty factors.
Since the average of the squares of joint torques is considered to
of the joint motor, it can be concluded that Approach 2 is more
energy efficient than Approach 1. Moreover, the average values of
the squares of joint torques for the wave-turn gait pattern with
Approach 2
duty factor of 5/6 are seen to be less than those obtained for other
4.0773 duty factors. In the wave-turning gait pattern with duty factor
2.9939
of 1/2, the front and rear legs of one side and middle leg of the
2.6133
2.2946 other side of the longitudinally symmetric six-legged robot are
always in support. Moment balance about the longitudinal axis, as
required to prevent the robot from rolling sideways, reveals that
the middle leg must exert more force compared to the front and
rear legs, and thus requires more torque and hence more power.
The forces required to support the body are distributed more
evenly among the legs, when duty factor increases and thereby,
the contribution (in terms of torque and power) of each support
leg is reduced. Since the average of the squares of joint torques is
considered to be proportional to average dissipated power of the
joint motor, it may be concluded that power dissipation and total
energy requirement become less in wave-turning gait with higher
duty factor in comparison with that involving lower duty factors.
4.1. Effect of angular velocity and duty factor on energy consumption
The effects of angular velocity on average power consumption
over a locomotion cycle of the robot for four different duty factors
are displayed in Table 4. For a particular value of duty factor,

Table 4
Variations of average power consumption with angular velocity during turning motion.
Angular velocity (deg./s) Average power consumption (in W)
= 1/2 = 2/3 = 3/4 = 5/6
Approach 1 Approach 2 Approach 1 Approach 2 Approach 1 Approach 2 Approach 1 Approach 2

0.5 0.3550 0.2228 0.2726 0.1731 0.2444 0.1539 0.2224 0.1362

1.0 0.3531 0.2407 0.2722 0.1938 0.2449 0.1764 0.2239 0.1579
1.5 0.3527 0.2600 0.2728 0.2156 0.2466 0.2001 0.2275 0.1817
2.0 0.3529 0.2800 0.2741 0.2380 0.2496 0.2250
2.5 0.3535 0.3004 0.2760 0.2610
3.0 0.3544 0.3210 0.2848 0.2787
3.5 0.3555 0.3419
4.0 0.3631 0.3569
Angular stroke = 6 deg., Height of trunk body = 0.13 m, Turning radius = 1 m
 S.S. Roy, D.K. Pratihar / Robotics and Autonomous Systems 60 (2012) 7282 79

Table 5
Variations of consumed energy per unit weight per unit traveled length with angular velocity during turning motion.
Angular velocity (deg./s) Energy per unit weight per unit traveled length
= 1/2 = 2/3 = 3/4 = 5/6
Approach 1 Approach 2 Approach 1 Approach 2 Approach 1 Approach 2 Approach 1 Approach 2

0.5 1.1849 0.7437 0.9099 0.5777 0.8157 0.5137 0.7421 0.4545

1.0 0.5892 0.4016 0.4542 0.3234 0.4087 0.2944 0.3737 0.2636
1.5 0.3923 0.2892 0.3035 0.2398 0.2744 0.2226 0.2531 0.2021
2.0 0.2945 0.2336 0.2287 0.1986 0.2082 0.1877
2.5 0.2360 0.2005 0.1842 0.1742
3.0 0.1971 0.1786 0.1584 0.1550
3.5 0.1695 0.1630
4.0 0.1515 0.1489
Angular stroke = 6 deg., Height of trunk body = 0.13 m, Turning radius = 1 m

Table 6
Variations of average power consumption with angular stroke during turning motion.
Angular stroke (deg.) Average power consumption (in W)
= 1/2 = 2/3 = 3/4 = 5/6
Approach 1 Approach 2 Approach 1 Approach 2 Approach 1 Approach 2 Approach 1 Approach 2

8 0.3724 0.3107 0.2836 0.2569 0.2596 0.2428 0.2429 0.2244

7.5 0.3666 0.3028 0.2807 0.252 0.2564 0.2382 0.2399 0.22
7 0.3614 0.2951 0.2781 0.2472 0.2537 0.2336 0.2374 0.2159
6.5 0.3569 0.2875 0.2759 0.2425 0.2515 0.2292
6 0.3529 0.28 0.2741 0.238 0.2496 0.225
5.5 0.3495 0.2727 0.2726 0.2336
5 0.3465 0.2656 0.2714 0.2294
4.5 0.3439 0.2586 0.2705 0.2253
4 0.3418 0.2519 0.2699 0.2215
3.5 0.34 0.2453
3 0.3384 0.2391
Angular velocity = 2 deg./s, Height of trunk body = 0.13 m, Turning radius =1 m

average power consumption is found to increase with the increase
in angular velocity, as expected. It is interesting to observe that
Approach 2 is able to provide more energy efficient gait (lying
in the range of 2% to 38%) compared to Approach 1 for all duty
factors. It is also important to note that for a high value of duty
factor, the robot may not be able to move with high angular
velocity due to dynamic constraints of the joint motors because
of the requirement of the faster leg transfers. Thus, the velocity
should be as low as possible to minimize power consumption for
a particular duty factor. However, traveling with a low velocity
takes more time to cover a fixed distance, and consequently,
total energy consumption may be increased. The energy required
to travel a fixed distance can be quantified using a parameter
called specific resistance, that is, the energy consumed per unit
weight and per unit traveled length. Table 5 displays the effects
of variation of velocity on energy per unit weight per unit traveled
length (that is, specific resistance) during turning over a flat terrain
with four different duty factors. The data of Tables 4 and 5 related
to Approach 2 are plotted in Fig. 7. Specific resistance is found
to decrease with the increase of angular velocity for a particular
value of duty factor. However, average power consumption is seen
to increase with the increase in angular velocity. Moreover, for a
high value of duty factor, angular velocity cannot be increased to a
high value due to dynamic constraints of joint actuators. The blank
cells () of Tables 4 and 5 correspond to angular velocities and
duty factors at which the robot is unable to walk because of the
violation of dynamic constraints of the motors. Approach 2 is seen
to yield more efficient gaits compared to Approach 1 for all duty
factors. The average power consumption is found to be minimum
(that is, 0.1362 W) corresponding to a duty factor of 5/6 and at
an angular velocity of 0.5 deg./s. On the other hand, the specific
resistance is seen to be the minimum (that is, 0.1489) for a duty
factor of 1/2 and at an angular velocity of 4.0 deg./s. Moreover, the
value of specific resistance corresponding to the minimum average
power consumption is found to be equal to 0.4545 units, whereas
it is seen to be the minimum for an average power consumption of
0.3569 W.
4.2. Effect of angular stroke and duty factor on energy consumption
Results related to the effects of angular stroke on average
power consumption and specific resistance during turning of the
robot with wave gaits of different duty factors are presented in
Tables 6 and 7, respectively. Approach 2 is seen to provide an
energy saving of 7%29% in comparison with Approach 1. The
data shown in Tables 6 and 7 related to Approach 2 are plot-
ted in Fig. 8. For a given angular velocity, both average power
consumption and specific resistance are found to increase with
angular stroke for all duty factors. Moreover, for a particular an-
gular stroke, average power consumption and specific resistance
are seen to decrease with the increase of duty factor for both ap-
proaches 1 and 2. It is interesting to observe that Approach 2 has
provided more energy efficient solutions compared to Approach 1
for all angular strokes. It is also to be noted that a few boxes of
Tables 6 and 7 are kept blank (), as no feasible solutions are ob-
tained for some combinations of angular stroke and duty factor due
to dynamic constraints of the robot.
4.3. Effect of angular stroke and duty factor on NESM
Table 8 and Fig. 9 show the minimum values of NESM (that is,
hmin ) obtained by the robot during turning on flat terrain with four
80 S.S. Roy, D.K. Pratihar / Robotics and Autonomous Systems 60 (2012) 7282

Table 7
Variations of consumed energy per unit weight per unit traveled length with angular stroke during turning motion.
Angular stroke (deg.) Energy per unit weight per unit traveled length
= 1/2 = 2/3 = 3/4 = 5/6
Approach 1 Approach 2 Approach 1 Approach 2 Approach 1 Approach 2 Approach 1 Approach 2

8 0.3107 0.2593 0.2366 0.2144 0.2166 0.2026 0.2027 0.1873

7.5 0.3059 0.2527 0.2342 0.2103 0.214 0.1987 0.2002 0.1836
7 0.3016 0.2462 0.2321 0.2063 0.2117 0.1949 0.1981 0.1801
6.5 0.2978 0.2399 0.2302 0.2024 0.2098 0.1913
6 0.2945 0.2336 0.2287 0.1986 0.2082 0.1877
5.5 0.2916 0.2275 0.2274 0.1949
5 0.2891 0.2216 0.2264 0.1914
4.5 0.287 0.2158 0.2257 0.188
4 0.2852 0.2101 0.2252 0.1848
3.5 0.2837 0.2047
3 0.2824 0.1995
Angular velocity = 2 deg./s, Height of trunk body = 0.13 m, Turning radius = 1 m

Fig. 6. Flowchart for computer simulations.

Table 8
Variations of minimum values of NESM with angular stroke during turning motion
with different duty factors over one locomotion cycle. Fig. 7. Variations of (a) average power consumption, (b) specific resistance, with
Angular stroke (deg.) Minimum value of normalized energy angular velocity using Approach 2.
stability margin (hmin ) (mm)
= 1/2 = 2/3 = 3/4 = 5/6
different values of duty factors. For a particular angular stroke, hmin
8 1.35 19.41 29.53 38.79
7.5 1.99 19.76 29.25 37.87 is seen to increase with the increase in duty factor. It happens due
7 2.75 20.10 28.96 36.95 to the fact that the number of ground legs increases as duty factor
6.5 3.64 20.45 28.68 36.05
6 4.64 20.79 28.39 35.15
increases and thereby the stability margin improves. It is also to
5.5 5.75 21.14 28.11 34.25 be observed that hmin decreases with angular stroke particularly
5 6.98 21.48 27.82 33.36 for low values of duty factor (that is, = 1/2 and = 2/3). For
4.5 8.32 21.83 27.54 32.49
4 9.77 22.17 27.25 31.62 higher values of duty factor (say 2/3 and 3/4), hmin does not vary
3.5 11.33 22.52 26.96 30.75 much with the angular stroke, due to a large number of ground legs.
3 12.99 22.86 26.68 29.89 However, for a duty factor of 5/6, an increasing trend is observed
Height of trunk body = 0.13 m, Turning radius = 1 m
for hmin with angular stroke.
 S.S. Roy, D.K. Pratihar / Robotics and Autonomous Systems 60 (2012) 7282
a considering the dynamic constraints of the actuators). Moreover,
b NESM decreases with the increase of angular stroke, whereas a
Fig. 8. Variations of (a) average power consumption, (b) specific resistance, with
angular stroke using Approach 2.
Fig. 9. Variations of NESM with angular stroke for different duty factor.
5. Conclusions
In this study, an attempt is made to minimize energy
consumption of a six-legged robot during its turning on flat
terrain. A power consumption model is derived for statically
stable wave gaits by minimizing dissipating power for optimal
foot force distribution and minimizing total energy expenditure
for optimal selection of gait parameters, namely angular velocity,
angular stroke and duty factor. Two approaches, namely Approach
1 (that is, minimization of norm of feet forces) and Approach 2
(that is, minimization of norm of joint torques) are developed
and their performances are tested through computer simulations
for generating the turning gait of a six-legged robot with four
different duty factors. It is important to mention that Approach
2 is seen to be more energy efficient compared to Approach 1 for
all duty factors. The variations of average power consumption and
energy consumption per unit weight per unit traveled length with
angular velocity and angular stroke are studied for the turning
of the six-legged robot with four different duty factors. The most
energy-efficient gait is obtained for the minimum value of duty
factor and maximum value of angular velocity (which is decided by
81
for a particular value of duty factor, specific resistance is seen to
decrease with the lower values of angular stroke, and it is found to
decrease further with the higher values of duty factor. Moreover,
in order to minimize total energy consumption, the velocity should
be as high as possible for a particular duty factor. However, as
angular velocity increases, the maximum reachable duty factor
is reduced due to the dynamic constraints of joint actuators. The
method adopted in this study may be helpful to decide on suitable
motors for all joints of the robot. The effects of angular stroke and
duty factors on the minimum value of NESM over a locomotion
cycle are also analyzed. It is observed that the minimum value of
NESM increases with the duty factor for a particular angular stroke.
Moreover, for a low value of duty factor, the minimum value of
reverse trend is observed for the same for higher values of duty
factor, due to the presence of more of ground legs.
6. Scope for future work
The present work will be extended in future to tackle the
problems related to gait generations of the walking robots over
rough terrains. This study can also be further extended to address
the scalability issues. The present paper deals with static stability
of the robot. However, the robot has to be dynamically stable also,
particularly when it moves with a high speed. It has also been kept
in the scope for future study.
Acknowledgments
The authors are profoundly grateful to anonymous reviewers,
whose suggestions have helped to improve the quality of this paper
significantly. The first author acknowledges all help from Mr. Ajay
Kr. Singh, Department of Mechanical Engineering, NIT, Durgapur,
India.
References
[1] S.M. Song, K.J. Waldron, Machines That Walk: The Adaptive Suspension
Vehicle, The MIT Press, Cambridge, Massachusetts, 1989.
[2] F. Cordes, I. Ahrns, S. Bartsch, T. Birnschein, A. Dettmann, S. Estable, S. Haase,
J. Hilljegerdes, D. Koebel, S. Planthaber, T.M. Roehr, M. Scheper, F. Kirchner,
LUNARES: lunar crater exploration with heterogeneous multi robot systems,
Journal of Intelligent Service Robotics 4 (1) (2011) 6189. Special issue on
Space Robotics.
[3] C.D. Zhang, S.M. Song, Stability analysis of wave-crab gaits of a quadruped,
Journal of Robotic Systems 7 (2) (1990) 243276.
[4] V. Kumar, K.J. Waldron, Gait analysis for walking machines for omnidirectional
locomotion on uneven terrain, in: A. Morecki, G. Bianchi, K. Kectzior
(Eds.), Proc. of 7th CISM-IFToMM Symp. Theory and Practice of Robots and
Manipulators, Udine, Italy, 1988, pp. 3762.
[5] S.M. Song, V.J. Vohnout, K.J. Waldron, G.L. Kinzel, Computer-aided design of a
leg for an energy efficient walking machine, Mechanism and Machine Theory
19 (1994) 1724.
[6] S. Hirose, Y. Umetani, Some consideration on a feasible walking mechanism
as a terrain vehicle, in: Proc. of 3rd International CISM-IFToMM Symposium,
Udine, Italy, 1978, pp. 357375.
[7] P. Gonzalez de Santos, E. Garcia, J. Estremera, Improving walking robot
performances by optimizing leg distribution, Autonomous Robots 23 (4)
(2007) 247258.
[8] E. Shin, D.A. Streit, An energy of spring efficient quadruped with two stage
equilibrator, Journal of Mechanical Design, ASME 115 (1993) 156163.
[9] R.M. Alexander, Three uses of springs in legged locomotion, The International
Journal of Robotics Research 9 (1990) 5361.
[10] D.W. Marhefka, D.E. Orin, Gait planning for energy efficiency in walking
machines, in: Proc. of IEEE International Conference on Robotics and
Automation, Albuquerque, NM, April, 1997, pp. 474480.
[11] D.C. Kar, K.K. Issac, K. Jayarajan, Gaits and energetics in terrestrial legged
locomotion, Mechanisms and Machine Theory 38 (2) (2003) 355366.
[12] J.S. Chen, F.T. Cheng, K.T. Yang, F.C. Kung, Y.Y. Sun, Optimal force distribution
in multilegged vehicles, Robotica 17 (1999) 159172.
[13] C.A. Klein, S. Kittivatcharapong, Optimal force distribution for the legs of a
walking machine with friction cone constraints, IEEE Transactions on Robotics
and Automation 6 (1) (1990) 7385.
82 S.S. Roy, D.K. Pratihar / Robotics and Autonomous Systems 60 (2012) 7282
[14] D.W. Hong, R.J. Cipra, Optimal contact force distribution for multi-limbed
robots, Journal of Mechanical Design, ASME 128 (3) (2006) 566573.
[15] M. Mahfoudi, K. Djouani, S. Rechak, Optimal force distribution for the legs of an
hexapod robot, in: Proc. of International Conference on Control Applications,
CCA, 2003, pp. 657663.
[16] P. Gonzalez de Santos, J.A. Cobano, E. Garcia, A six-legged robot based system
for humanitarian demining missions, Mechatronics 17 (2007) 417430.
[17] V.V. Lapshin, Energy consumption of a walking machine: model estimations
and optimization, in: Proc. 7th Conference on Advanced Robotics, San Feliu de
Guixols, 1995, pp. 420425.
[18] D.E. Orin, Y. Oh, A mathematical approach to the problem of force distribution
in locomotion and manipulation system containing closed kinematic chains,
in: Proc. of 3rd International CISM-IFToMM Symposium, Udine, Italy, 1978,
pp. 123.
[19] D.E. Orin, Y. Oh, Control of force distribution in robotic mechanism containing
closed kinematic chains, Journal of Dynamic Systems, Measurement, and
Control, ASME 102 (1981) 134141.
[20] M.A. Nahon, J. Angeles, Minimization of power losses in cooperating
manipulators, Journal of Dynamic Systems, Measurement, and Control, ASME
114 (1992) 213219.
[21] D.W. Marhefka, D.E. Orin, Quadratic optimization of force distribution in
walking machines, in: Proc. of IEEE International Conference on Robotics and
Automation, Leuven, Belgium, May 1998, pp. 477483.
[22] D.C. Kar, K.K. Issac, K. Jayarajan, Minimum energy force distribution for a
walking robot, Journal of Robotic Systems 18 (2) (2001) 4754.
[23] B.S. Lin, S.M. Song, Dynamic modeling, stability and energy efficiency of a
quadrupedal walking machine, Journal of Robotic Systems 18 (11) (2001)
657670.
[24] M.S. Erden, K. Leblebicioglu, Torque distribution in a six-legged robot, IEEE
Transactions on Robotics 23 (1) (2007) 179186.
[25] M.S. Erden, K. Leblebicioglu, Analysis of wave gaits for energy efficiency,
Autonomous Robots 23 (2007) 213230.
[26] J. Nishii, Gait pattern and energetic cost in hexapods, in: Proc. of 20th Annual
International Conference of the IEEE Engineering in Medicine and Biology
Society, vol. 20, 1998, pp. 24302433.
[27] K. Arikawa, S. Hirose, Study of walking robot for 3 dimensional terrain, in: Proc.
of IEEE ICRA-95, Nagoya, Japan, vol. 1, 1995, pp. 703708.
[28] M.F. Silva, J.A. Tenreiro Machado, A.M. Endes Lopes, Energy analysis of multi-
legged locomotion systems, in: Proc. of 4th Conf. on Climbing and Walking
Robots, Karlsruhe, Germany, 2001, pp. 143150.
[29] V.V. Zhoga, Computation of walking robots movement energy expenditure, in:
Proc. of IEEE Int. Conf. on Robotics and Automation, Leuven, 1998, pp. 163168.
[30] T. Zelinska, Efficiency analysis in the design of walking machine, Journal of
Theoretical and Applied Mechanics 38 (2000) 693708.
[31] P. Gonzalez de Santos, E. Garcia, R. Ponticelli, M. Armada, Minimizing energy
consumption in hexapod robots, Advanced Robotics 23 (2009) 681704.
[32] Z.Y. Wang, X.L. Ding, A. Rovetta, Analysis of typical locomotion of a symmetric
hexapod robot, Robotica 28 (2010) 893907.
[33] A.A. Durge, K.K. Issac, An algorithm for optimal gait generation for
level ground walking of an omnidirectional hexapod, in: Proc. of IEEE
International Conference on Robotics and Automation, New Orleans, LA, 2004,
pp. 30993104.
[34] J. Estremera, P. Gonzalez de Santos, Generating continuous free crab gaits for
quadruped robots on irregular terrain, IEEE Transactions on Robotics 21 (6)
(2005) 10671076.
[35] J.M. Yang, Fault-tolerant crab gaits and turning gaits for a hexapod robot,
Robotica 24 (2006) 269270.
[36] J.M. Yang, Omnidirectional walking of legged robots with a failed leg,
Mathematical and Computer Modelling 47 (2008) 13721388.
[37] D.E. Orin, Supervisory control of a multi-legged robot, The International
Journal of Robotics Research 1 (1982) 7991.
[38] S. Hirose, H. Kikuchi, Y. Umetani, The standard circular gait of a quadruped
walking vehicle, Advanced Robotics 1 (2) (1986) 143164.
[39] C.D. Zhang, S.M. Song, Gaits and geometry of a walking chair for the disabled,
Journal of Terramechanics 26 (34) (1989) 211233.
[40] C.D. Zhang, S.M. Song, Turning gait of a quadrupedal walking machine, in: Proc.
of IEEE Int. Conf. on Robotics and Automation, Sacramento, California, April,
1991, pp. 21062112.
[41] S. Miao, D. Howard, Optimal tripod turning gait generation for hexapod
walking machines, Robotica 18 (2000) 639649.
[42] J. Estremera, J.A. Cobano, P. Gonzalez de Santos, Continuous free-crab gaits
for hexapod robots on a natural terrain with forbidden zones: an application
to humanitarian demining, Robotics and Autonomous Systems 58 (2010)
700711.
[43] D.K. Pratihar, K. Deb, A. Ghosh, Optimal turning gait of a six-legged robot using
GA-fuzzy approach, Artificial Intelligence for Engineering Design, Analysis and
Manufacturing 14 (2000) 207219.
[44] D.K. Pratihar, K. Deb, A. Ghosh, Optimal path and gait generations simulta-
neously of a six-legged robot using a GA-fuzzy approach, Robotics and Au-
tonomous Systems 41 (2002) 120.
[45] K.S. Fu, R.C. Gonzalez, C.S.G. Lee, Robotics: Control, Sensing, Vision, and
Intelligence, McGraw Hill, Singapore, 1987.
[46] S.S. Roy, A.K. Singh, D.K. Pratihar, Estimation of optimal feet forces and
joint torques for on-line control of six-legged robot, Robotics and Computer
Integrated Manufacturing 27 (5) (2011) 910917.
[47] J. Nishii, Legged insects select the optimal locomotor pattern based on the
energetic cost, Biological Cybernetics 83 (2000) 435442.
[48] J. Nishii, An analytical estimation of the energy cost for legged locomotion,
Journal of Theoretical Biology 238 (2006) 636645.
[49] R.B. McGhee, A.A. Frank, On the stability properties of quadruped creeping
gaits, Mathematical Biosciences 3 (1968) 331351.
[50] D. Messuri, C. Klein, Automatic body regulation for maintaining stability of a
legged vehicle during rough-terrain locomotion, IEEE Journal of Robotics and
Automation RA-1 (3) (1985) 132141.
[51] S. Hirose, H. Tsukagoshi, K. Yoneda, Normalized energy stability margin:
generalized stability criterion for walking vehicles, in: Proc. of the Int. Conf.
on Climbing and Walking Robots, Brussels, Belgium, 1998, pp. 7176.
[52] http://www.3ds.com.
Shibendu Shekhar Roy received the B.E and M.Tech
degrees in Mechanical Engineering from the R.E. College,
Durgapur (Presently, NIT, Durgapur) in 1999 and 2001,
respectively. He obtained his Ph.D from IIT, Kharagpur,
India, in 2011. Currently. He is an Assistant Professor
in the Department of Mechanical Engineering, National
Institute of Technology, Durgapur, India. From March
2001 to December 2006 he was a Scientist at the Central
Mechanical Engineering Research Institute, Durgapur,
CSIR, India. He has a number of research papers in journals
and conferences and has filed number of patents in
product development. His research interests are in the modeling and simulation
of multi- legged robots and Soft Computing.
Dilip Kumar Pratihar (B.E. (Hons.), M.Tech., Ph.D.) re-
ceived his Ph.D. from IIT Kanpur, India, in the year 2000.
Besides several scholarships and best paper awards, he
received the University Gold Medal for securing the high-
est marks in the University in 1988, the A.M. Das Memo-
rial Medal in 1987, the Institution of Engineers Medal in
2002, and others. He completed his post-doctoral stud-
ies in Japan (6 months) and Germany (1 year) under the
Alexander von Humboldt Fellowship Programme. He is
working at present as a Professor in the Department of Me-
chanical Engineering, IIT Kharagpur, India. His research ar-
eas include robotics, soft computing and manufacturing science. He has published
more than 130 papers in different journals and conference proceedings. He has writ-
ten a textbook on Soft Computing, which has been published by Narosa Publish-
ing House, New Delhi and Alpha Science International Publisher, UK. Recently, this
book has been translated into Chinese. He has edited a book on Intelligent and Au-
tonomous Systems, which was in 2010 published by Springer-Verlag, Germany. He
has been included as a member of the program committee for several international
conferences. He has been selected as the Editorial Board Member of eight interna-
tional journals. He has been elected as a Fellow of the Institution of Engineers (I)
and Member of IEEE.