Road Following for Hexapod Mobile Robot with
Adaptive Tripod Gait
Sameh Ibrahim Beaber Ahmed Y. AbdelHamid
Department of Mechanical Engineering Department of Mechanical Equipment
Military Technical College Military Technical College
Cairo, Egypt Cairo, Egypt
sibrahim@mtc.edu.eg ayahia@mtc.edu.eg
Abstract—The biological locomotion of animals shows an
incredible degree of flexibility and durability that gives them
an advantage to move across rough terrain. Even if hexapod
robots are so superior to adjust with uneven terrains, they
already have some problems to follow a smooth path exactly.
With such obstacles, regular Periodic walking gates will not be
able to respond quickly. During this study, the task to follow
an accurately predetermined route in the Cartesian region is
built through the adaptive gate. The Phantom ll model robot
case study is simulated via Matlab SimMechanicsTM toolbox to
evaluate and estimate dynamics of the hexapod and the adaptive
gate implemented. In addition, the Phantom ll case study is
evaluated in the kinematic model that consisting of two main
objectives, the direct and inverse kinematics. Inverse kinematics
is calculated geometrically and the Denavit-Hartenberg method
is applied to determine the direct kinematics. The robot margin
of stability and kinematic limitations are also taken into
consideration. Simulation results showed the suitability of the
presented adaptive gait.
Keywords—Hexapod Robot, Kinematic Model, Dynamic
Model, Mobile Robot, Adaptive Tripod Gait.
The natural locomotion of animals shows a high degree of
robustness and adaptability that allows them to move on rough
terrain and cross obstacles along a certain path. Growth in
mobile robotics has been attracting great attention in recent
years due to its relevance in real life, which can be used in
many applications and fields. There are wheeled, legged and
tracked mobile robots [1]. Walking legged robots is the main
goal of many experiments and investigations. The main feature
of legged robot is to establish a machine that simulates certain
biological plant characteristics [2].
The focus of robotics researchers goes to the legged robot,
which uses a mechanical limb for movements that make it
better on rough and unstructured terrain than tracked and
wheeled vehicles [3], [4]. There are several types of legged
robots, for example: bipeds, tripods, quadrupeds, hexapods,
and octopods, based on the number of legs for individual
robots. An exposed research problem, however, is the design
and control of legged robots motionless, and the development
of models or prototypes for these studies is still costly and
time-consuming. This reality contributes to the development
of a software model that allows such studies to be carried
out and helps to apply them in real time. Studies of control
and locomotion included these models to obtain biological
Maged M. Abou Elyazed
Department of Mechanical Equipment
Technical Research Center of Armed Forces
Cairo, Egypt
maged abou elyazed@hotmail.com
organism’s main characteristics.
Many software tools are available to help design, simulate
and test mechanical system behavior under various conditions
[5]. Such software programs make it easy to have a real-time
view of the mechanical parts. These also require the built
robot to be tested in three-dimensional space. Using these
programs does not remove the need to design the mechanical
systems ’ working prototype, it reduces the error by conducting
the required tests before the prototype is designed. Dynamic
systems simulation uses software programs that can be used
in many types of industries, with an emphasis in robotics,
due to the possibilities of modeling and simulation of many
forms of complex systems. Various software programs such as
I Webot [6], (ii) Simpack [7], (iii) Adams Multibody Dynamic
Simulation [8], and (v) Matlab / SimMechanics [9] may obtain
three-dimensional simulation of mechanical systems.
One of these software programs is the Sim-Mechanics
MATLAB toolbox. One of the benefits of such integrated
toolboxes is that they allow you to perform different tasks
using other tools available in the simulation framework. For
instance, designing control system, analyzing results of sim-
ulation, visualizing results, etc. In addition, Sim-Mechanics
makes it very easy to import CAD model into the Simulink
from a CAD environment.
There are many gaits that can be used by hexapod robots,
such as tripods, waves and rippled gait. It all depends on
rotating the robot’s legs regularly to create a cyclic motion
as a result of a fixed move of the leg. Such regular and
fixed gaits prevent the robot from taking different paths that
may be necessary to avoid an obstacle collision. Researchers
have gained tremendous potential to concentrate their work on
adaptive gaits robot hexapods.
A Phantom AX Hexapod Mark II [10] is used in this work
as a case study. Section two provides a comprehensive case
study model of kinematics and dynamics. Then, section three
discusses the methods used in this work. Results analysis and
review are presented in section four.
I. M ODELING OF H EXAPOD ROBOT
The following constraints are assumed in order to more
stable when traveling across different gradient rough terrain:
 • The body of the trunk is kept at a level of steady and TABLE I: D-H Parameters for Three Joint Legs
maintained in parallel with the locomotion terrain;
• The body of the robot moves with a constant speeds over Link αi ai di θi
the gradient;
• The hexapod is supposed to describe a tripod gate having 1 90 52 0 θ1
2 0 66 0 θ2
duty factor of 2/3 [11]; and 3 0 137 0 θ3
• The leg joints are DC motors that do not hold negative
energy.
Therefore, any negative energy came from external forces are structures as follows:
lost. To evaluate the complex relationships for both power
 
consumption and locomotion parameters, a full kinematic and c(θi ) −s(θi )c(αi ) s(θi )s(αi ) ai c(θi )
dynamic model of the hexapod robot is desired. i−1
 s(θ i) c(θi )c(αi ) −c(θi )s(αi ) ai s(θi ) 
Ti = 
 0

s(αi ) c(αi ) di 
0 0 0 1
A. Kinematic Model of the Hexapod Robot
The 3-D model of the hexapod robot in the current study
is shown in Fig. 1. The robot consists of a hexagonal trunk
shape with six legs which are distributed on two sides equally
and symmetrically around the base. Each leg consists of three
degrees of freedom and three links (coxa - femur - tibia) which
are connected by three joints, with a length of 52 mm, 66
mm, and 137 mm, respectively. Table I shows presented the
D-H parameters, link length (ai ), link angle (αi ), joint distance
(di ), and joint angel (θi ), that required to explain the legs joint
movements.

Fig. 2: Forward Schematic Model of Leg

For various leg links shown in the Fig. 2, the following

was identified to be homogeneous transformation matrices:

 
cos θ1 0 sin θ1 L1 cos θ1
0
 sin θ1 0 − cos θ1 L1 sin θ1 
T1 = 
 0
 (1)
1 0 0 
0 0 0 1

 
cos θ2 − sin θ2 0 L2 cos θ2
1
 sin θ2 cos θ2 0 L2 sin θ2 
T2 = 
 0
 (2)
0 1 0 
0 0 0 1

Fig. 1: 3-D Model of a Hexapod Robot  

cos θ3 − sin θ3 0 L3 cos θ3
2
 sin θ3 cos θ3 0 L3 sin θ3 
The transformation matrix representing the relative rotation T3 = 
 0
 (3)
0 1 0 
and translation between ith and the (i − 1) coordinate 0 0 0 1
 boundary terms of speed and angular displacement and speed
The leg tip reference frame 3 can therefore be represented are applied at the first, middle and last point of the trajectory.
in the leg reference frame 0 as shown below: Each joint acceleration and velocity in an swing leg can be
accomplished in the following equations:
3
Y
0
T3 =0 T11 T22 T3 = .i−1 T (4)
i=1
 
c(θ1 )c (θ2 + θ3 ) −c(θ1 )s (θ2 + θ3 ) s(θ1 ) (L1 + L2 c(θ2 ) + L3 c (θ2 + θ3 )) c(θ1 )
 s(θ1 )c (θ2 + θ3 ) −s(θ1 )s (θ2 + θ3 ) −c(θ1 ) (L1 + L2 c(θ2 ) + L3 c (θ2 + θ3 )) s(θ1 ) 
= 
 s (θ2 + θ3 ) c (θ2 + θ3 ) 0 L2 s(θ2 ) + L3 s (θ2 + θ3 ) 
0 0 0 1
θ̇j = aj1 + 2aj2 t + 3aj3 t2 + 4aj4 t3 + 5aj5 t4 (12)

θ̈j = 2aj2 + 6aj3 t + 12aj4 t2 + 20aj5j t3 (13)

In addition, the velocity and acceleration formulas can be

presented as shown for each leg throughout the support phase:

θ̇ = J−1 ṗ and θ̈ = J−1 (p̈ − J̇θ̇) (14)

where the position vector

T
p = [px py pz ]

and J is the Jacobian matrix.

B. Dynamic Model of the Hexapod Robot

Lagrange-Euler formula was used to establish the dynamic
equations and to determine the variations of joint torques
over the locomotion period. A standardized formulation of
Lagrange-Euler equations produces a dynamic expression
Fig. 3: Inverse schematic model of leg which can be described as shown below in the vector-matrix
form.
Geometric method is applied in order to find inverse
kinematics of each leg . In Fig. 3 the angles of the leg joints τi = [M(θ)θ̈ + H(θ, θ̇) + G(θ)]i − JiT Fi (15)
(θ1, θ2, θ3) are defined as follows from the corresponding
where M (θ) is the 3∗3 leg mass matrix, H is a 3∗1 vector of
position of the leg tip X, Y, and Z:
Coriolis and centrifugal terms, G(θ) is a 31 gravity vector
terms, i is the 31 joint torques vector and Fi is the 3∗1
θ1 = tan(Y /X) (5)
p vector of foot ’i’ ground reaction forces. There is no leg-terrain
ac = X 2 + Y 2 (6) interaction throughout the swinging phase of the leg, and Fi
equal zero. Moreover, ground interaction occurs during the
bc = ac − Lc (7) support cycle and equation (15) is undetermined, which must
p
ec = ad2 + bc2 (8) be determined with an optimization method, e.g. optimum
distribution of foot force.
θ2 = cos−1 ec2 + L2f − L2t / (2 ∗ ec ∗ Lf )



(9)
− 90 − cos−1 (ad/ec)


C. Optimum Foot Force Distribution

θ3 = 180 − cos−1 L2t + L2t − ec2 / (2 ∗ L, ∗Lt )

(10)
To ensure that the swinging legs are working smoothly,
each joint trajectory is supposed to follow the fifth degree
polynomial in time (t). In the fifth degree polynomial, the jth
joints of the swing leg represented as:
θj = aj0 +aj1 t+aj2 t2 +aj3 t3 +aj4 t4 +aj5 t5 ; j = 1, 2, 3 (11)
where aj0 , aj1 ,aj2 , aj3 , aj4 , and aj5 are coefficients.
In order to find these six parameters for each joint, the
The following assumptions were also created for the calcu-
lation of distribution foot-force:
• The supported legs are designed to hold the robot body
on its tip points with no slippage; and
• The connections between the feet tip and the ground
could be described as rough point contacts having fric-
tion.
In the current study, two approaches have been used to solve
the problem of distribution of foot force as discussed below.
 T T
1) Approach 1: Minimization of Norm of Foot Forces: [τ ] = [τp , τq , τr , τs ] and τi = [τi1 , τi2 , τi3 ] are the torque
T
Suppose that Fi = [fix , fiy , fiz ] is the force vector of vectors og leg i.
the reaction of the tip i on the ground. The wrench The equation (16) can be reformulated as follows:
T
W = [Fx , Fy , Fz , Mx , My , Mz ] that contains forces
(Fx , Fy , Fz ) and the moments (Mx , My , Mz ) that acting at [A][J][τ ] = −[B][W] (18)
the center of gravity of the robot and described the payload
[AJ ] [τ ] = −[B][W] (19)
of the robot with the consideration of the surface gradient
impact that acting on it, all external forces action on the A least squared approach has been used for obtaining the min-
robot and internal body effect. Furthermore, for simplifying imum standard solution of these above undefined equations.
the analysis, the inertial legs effects were ignored. According
to these situations, the following expression can be described II. METHODOLOGY
six steady state equations, which make a balance between This work is intended to rule the robot of the hexapod to
moments and forces when the robot legs are in supporting follow a predefined Cartesian path. The conventional gaits,
phase: however, could successfully drive the robot forward and
backward, probably struggling to keep it on an exact shaped
[A][F] = −[B][W] (16) path. Shaped paths may have special curves that distinguish
the paths of the robot legs from each other. Therefore, the
where;
common walking gaits ’ periodic fixed steps may not be

I3 I3 I3 I3
 
I3 03
 appropriate for this task and a form of adaptive gaits should
[A] = and B = be used.
Rp Rq Rr Rs Rc I3
The implemented algorithm in this work deals separately
I3 represent (3*3) identity matrix, 03 is the (3*3) null with each leg to define its unique path that could be followed
matrix and Ri describe (3*3) skew symmetric matrix of by other paths of the legs to approach the entire body path of
T
vector [xi , yi , zi ] . the robot. At the beginning of this algorithm, the predefined
    required CG frame path with respect to the global frame
0 −zi yi 1 0 0
(g) is specified. Assume that each leg joints are set to the
Ri =  zi 0 −xi  and I3 =  0 1 0 
standing position values (θ1 ,θ2 ,θ3 ) and the body path may be
−yi xi 0 0 0 1
discreetly placed (X, Y , Z) with respect to the global frame
at any time along the whole path. Hence, by substituting in
This matrix (Ri ) represent the tip foot (i) position, the kinematic equations the locus of each leg path could be
where (i=p, q, r, s using wave gait that have a duty factor presented.
equal 2/3) or which center of gravity (i=c) all with based on Now, the overall path required for each tip of the leg is
body reference frame. The axis of ith contact point between established. But the kinematic constraints of the robot and
the foot and the ground related to body reference frame, the margin of stability restrict it from taking in one shot
found at the geometric center of the body, are defined by several lengthy routes. As presented in the flow chart in Fig.
(xi , yi , zi ). With the known feet positions, the feet forces 4, the path points for the three legs to take the first step
during a whole locomotion cycle can be computed using are checked for stability point by point before the inverse
equation (16), that is indeterminate, as it is composed of six kinematic algorithm determines the corresponding joint
equations with nine unknowns. Least squared method was angles, BARYCENTRIC technique is used to test the static
used to find the solution of equation (16), that provides an stability of the robot. If the margin of stability is okay, there
minimum acceptable solution of the equilibrium equations will be an inverse kinematic and the joint variable for each
that is indeterminate. With another meaning, it is also the leg will be stored in arrays for this stage. But If the stability
solution which reduces the total value of the leg forces margin is NOT ok at any time or the joint variables for any
squares of components. leg exceed the kinematic constraints, the algorithm will stop
and begin repeating the process for the second step by the
2) Approach 2: Minimization of Norm of Joint Torques: second legs set and so on.
In that approach, equation (16) can be transformed using the The introduced algorithm makes the tripod gate more
following relations: adaptive from its cyclically and periodically fixed manner.
In another word, each leg could have its unique path by
[F] = [J] · [τ ] (17) implementing that algorithm that could be deferred from
others. Eventually, the start and end positions of each
Jp
 
03 03 03 step might identify each leg swing phase. To improve the
 03 Jq 03 03   ; Ji = JT −1 ; i =
  interaction between the tip of the leg and the ground, the
where [J] =  r i
 03 03 J 03  cubic polynomial path is used to regulate the velocity of
s joints in swing phase.
03 03 03 J
p, q, r, s and Ji is the (3 × 3) Jacobian matrix of leg i. Hence,

Fig. 4: Flow Chart of Path Planning of Phantom X Hexapod
Robot
III. S IMULATION R ESULTS
The simulation results for using the Adaptive Walking
Gait strategy are discussed in this section as well as the
stability technique that enables the hexapod robot to still stable
throughout the curved road. The tripod adaptive gait of the
hexapod is proven as each leg needs different step angles to
adjust and track the pre-defined curved path that means low
power consumption for all actuators and greater robot stability.
Fig. 5 explain the attitude of the hexapod to move on pre-
defined curved path for the first set legs (1, 3 and5) and the
second set legs (6, 4 and 2) with different steps to adjust the
length of the route . Fig. 6 and Fig. 7, respectively, illustrate
Fig. 5: Circular Path Behavior for Hexapod
the angles of the links of each leg for the first and second
set legs. It is obvious that the curved path consists of three
stages; the first and second stages are almost similar, but the
third stage is distinct as it is lower than the others and the link
angles are not similar for the same set of legs that support the
aim of the required adaptive walking gait strategy.
Fig 8 and Fig. 9 shows the friction and normal force for
Fig. 6: Angles of 1st Set of Legs Moving a Part of Circular
Fig. 7: Angles of 2nd Set of Legs Moving a Part of Circular
each leg tip. Throughout the transfer process, as the legs do
not hit the ground, the total of normal force is exactly equal
zero. While the total of the normal forces on the three legs in
the support process is nearly equal to the robot’s weight, that
is about 26 N. The normal force on leg 2 is higher than the
normal force on legs 4 and 6 when they are in the support
process.
Figs. 10 and 11 shows the variety of toques added to each
leg joints. Legs 1, 3, and 5 are start in the transfer cycle, while
the other three legs are in support phase.
IV. C ONCLUSIONS
In this study, an adaptive walking gait is represented to
overcome the task of performing an effective pre-defined
curved path while moving on the Cartesian space. Cyclic gaits
and Classical periodic do not accurately meet the requirement
to follow these paths. Forward kinematic is determined using
Denavit-Hartenberg formula, geometrically obtaining inverse
kinematic equations. In addition, consideration is given
 Fig. 8: Forces of 1st Set as reverse curved path
Fig. 9: Forces of 2nd Set as reverse curved path
to the margin of robot stability and kinematic limitations.
Results confirm the adaptive gait’s ability to follow a curved
path. Through integrating this path, any complex path
can be implemented sufficiently. On the other hand, it is
important to determine the contact forces between the leg tip
and the ground because of the legs slipping across the grounds.
V. F UTURE W ORK
Future work includes the implementation of model-based
controller in order to remove the slipping of the leg tip on
the ground. Moreover, studying moving of the hexapod robot
on an inclined plane and determine the reaction force and
the torque of each joint. Applying this study on the robot
experimentally and compare the results with the simulated
one. This research will expand the use of these robots all
over the world, in particular in dangerous environments.
ACKNOWLEDGMENT
The authors want to thank the Military Technical College
for it’s scientific and technological support for completing this
research.
Fig. 10: Torque of 1st Set as reverse curved path
Fig. 11: Torque of 2nd Set as reverse curved path
R EFERENCES
[1] S. G. Tzafestas, Introduction to mobile robot control. Elsevier, 2013.
[2] M. Silva, R. Barbosa, and T. Castro, “Multi-legged walking robot
modelling in matlab/simmechanicstm and its simulation,” in 2013 8th
EUROSIM Congress on Modelling and Simulation. IEEE, 2013, pp.
226–231.
[3] K. Walas and D. Belter, “Messor-verstatile wal king robot for serach and
rescue missions,” Journal of Automation Mobile Robotics and Intelligent
Systems, vol. 5, pp. 28–34, 2011.
[4] D. Thilderkvist and S. Svensson, “Motion control of hexapod robot using
model-based design,” 2015.
[5] S. I. Beaber, A. S. Zaghloul, M. A. Kamel, and W. M. Hussein,
“Dynamic modeling and control of the hexapod robot using matlab
simmechanics,” in ASME 2018 International Mechanical Engineering
Congress and Exposition. American Society of Mechanical Engineers,
2018, pp. V04AT06A036–V04AT06A036.
[6] Cyberbotics., “Webots: the mobile robotics simulation software,,” http:
//www.cyberbotics.com, 2011, retrieved: July 6, 2020.
[7] S. AG, “Simpack - multi-body simulation software,” http://www.
simpack.com, 2010, retrieved: July 6, 2020.
[8] M. Software, “Adams - the multibody dynamics simulation so-
lution,” http://www.mscsoftware.com/Products/CAE-Tools/Adams.aspx,
2013, retrieved: July 6, 2020.
[9] “Adams - the multibody dynamics simulation solution,” http://www.
mathworks.com/Products/products/simmechanics, 2013, retrieved: July
6, 2020.
[10] “T. robotics. (2015-06-09.). phantomx ax hexapod mark ii kit.
available:,” http://www.trossenrobotics.com/phantomx-ax-hexapod.aspx,
2013, retrieved: July 6, 2020.
[11] S.-M. Song and K. J. Waldron, Machines that walk: the adaptive
suspension vehicle. MIT press, 1989.