robotics

Article
A Dynamic Approach to Low-Cost Design, Development, and
Computational Simulation of a 12DoF Quadruped Robot
Md. Hasibur Rahman 1,2 , Saadia Binte Alam 1,2, *, Trisha Das Mou 1,2 , Mohammad Faisal Uddin 1,2
and Mahady Hasan 1,2

1 Department of Computer Science Engineering, IUB, Dhaka 1229, Bangladesh

2 RIoT Research Center, IUB, Dhaka 1229, Bangladesh
* Correspondence: [email protected]

Abstract: Robots equipped with legs have significant potential for real-world applications. Many
industries, including those concerned with instruction, aid, security, and surveillance, have shown
interest in legged robots. However, these robots are typically incredibly complicated and expensive
to purchase. Iron Dog Mini is a low-cost, easily replicated, and modular quadruped robot built for
training, security, and surveillance. To keep the price low and its upkeep simple, we designed our
quadruped robot in a modular manner. We provide a comparative study of robotic manufacturing
cost between our proposed robot and previously established robots. We were able to create a
compact femur and tibia structure with sufficient load-bearing capacity. To improve stability and
motion efficiency, we considered the novel Watt six-bar linkage mechanism. Using the SolidWorks
modeling software, we analyzed the structural integrity of the robot’s components, considering their
respective material properties. Furthermore, our research involved developing URDF data for our
quadruped robot based on its CAD model. Its gait trajectory is planned using a 14-point Bezier curve.
We demonstrate the operation of the simulation model and briefly discuss the robot’s kinematics.
Computational methods are emphasized in this research, coupled with the simulation of kinematic
and dynamic performances and analytical/numerical modeling.

Citation: Rahman, M.H.; Alam, S.B.; Keywords: kinematics; simulations; quadruped robot; watt six-bar linkage; PyBullet; URDF; 12DoF
Mou, T.D.; Uddin, M.F.; Hasan, M. A
Dynamic Approach to Low-Cost
Design, Development, and
Computational Simulation of a
1. Introduction
12DoF Quadruped Robot. Robotics
2023, 12, 28. https://doi.org/ Numerous quadrupedal robots have been developed in recent decades. The potential
10.3390/robotics12010028 of mobile robots to replace people in high-risk situations has made this a fascinating area
of robotics study [1], including first response, flammable and toxic substance disposal, and
Academic Editor: Raffaele Di
lidar field navigation. Most remotely operated robots fall into three categories: Those with
Gregorio
wheels, those with a crawler-type under-carriage, and those with legs [2]. Although mobile
Received: 16 December 2022 robots that move on wheels or crawlers may be effective on the ground level, their mecha-
Revised: 29 January 2023 nisms are severely limited by obstacles such as hills and mountains. Legged robots have a
Accepted: 30 January 2023 wider variety of potential applications, as they can traverse difficult terrain [3]. The number
Published: 17 February 2023 of feet may be used to categorize robots into three groups: Bipeds, quadrupeds, and multi-
legged. More people are interested in quadruped robots than in the past, as they are more
stable, can carry more weight than biped robots, have better mobility performance than
multi-legged robots, and are more efficient at moving than the aforementioned robots [4].
Copyright: © 2023 by the authors.
Rapid and accurate evaluation of dynamic characteristics is essential for precise modeling,
Licensee MDPI, Basel, Switzerland.
estimation, and control of robots. Scientists in the field of robotics can greatly benefit from
This article is an open access article
having access to techniques that can reduce error and speed up the development process.
distributed under the terms and
In this study, we provide a computationally based, organized simulation model run in
conditions of the Creative Commons
the Pybullet physics engine, in order to address a wide variety of issues that arise when
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
constructing quadruped robots. Our research here describes the connection between a
4.0/).
kinematics equation and the characteristics of the joints of a four-legged robot in depth. The

Robotics 2023, 12, 28. https://doi.org/10.3390/robotics12010028 https://www.mdpi.com/journal/robotics

Robotics 2023, 12, 28 2 of 23

robot’s design features legs with three degrees of freedom. in this study, we emphasize the
use of computational methods, such as modeling and simulation of kinematic and practical
implementations.
Bipedal and quadrupedal robots are incredibly well-known for their complex, sophis-
ticated mechanics and mathematical techniques, although robotics researchers deal with a
wide variety of robots at present. Parts of the quadruped robot are relevant to creatures
with four legs. Simulating a high-quality, mechanically constructed, and hydraulically op-
erated quadruped robot requires the employment of at least twelve degrees of freedom [5].
Scientists have previously developed robot engine controls for a quadruped robot [6].
Implementing a four-legged robot relies heavily on simulations run in software. This
research suggested a simulation model with outlined components and parameters, such
as kinematics, designs, and a standard robot description format. To simulate the motion
of a robot’s 2DoF leg, kinematics solutions have been implemented in a purpose-built
3D program [7]. One research highlight was demonstrating a novel robot design using
electrically operated motors [8]. Additionally, most quadruped robots, such as wildcat [9],
use a two-joint construction for their primary leg component. This framework is uncom-
plicated, understandable, and straightforward to manage. However, there are significant
biological advantages to the three-joint limb structure of toed animals, such as cats, dogs,
and lions, regarding walking velocity. MIT Cheetah utilized a three-segment construction
that allowed for a running velocity of 6 m/s and the efficient passage of obstacles [10].
Additionally, the Cheetah-cub was created with a pantograph leg arrangement, in order to
simplify controlling three links with only two joints and allowing it to achieve a trot [11].
Additionally, Pneupard—a genuine “cat-sized” robot—has utilized the same pantograph
technique [12]. It has been shown that placing motors at the shoulder works well for high-
velocity locomotion. Ming Lu has highlighted a 2DoF-based parallel leg formation [13]. A
hybrid-legged wheeled robot, which uses a wheel and leg for walking, has been described
in [14]. Kinematic analysis can be used as the determining element for quadruped robots,
and recent studies [15,16] have detailed exquisite mathematical kinematics techniques. A
motion observer study for a four-legged robot capable of traversing rugged terrain has
been carried out [17]. Gait pattern creation is critical for quadruped robots, as they require
precise trajectory adjustment. Sooyeong Yi has described the two-phase discontinuous gaits
of quadruped walking robots [18]. Thanhtam Ho has created a biomimetic self-contained
quadruped bounding robot [19].
The researcher gave an overview of a quadruped robot model developed in a dynamic
simulator with an alternate gait creation mechanism [20]. Through structural simulation,
one of the essential load test analyses of a robot leg has been detailed [21]. Knowing
how much weight it can withstand and how long it can withstand pressure is beneficial.
The researcher demonstrated a variable-based design for a quadruped robot using the
equivalent motion, a validation approach, and a non-programmable method [22–24]. Previ-
ously, researchers have employed computer modeling and analysis before constructing a
massive robot, thus reducing instrument losses and making the model construction process
faster. MIT researchers have used 3D design control to demonstrate improved design and
advancement [25].
In SolidWorks, we created a working prototype of a four-legged robot with precise
joint measurements, specifications, and dimensions. Initially, we intended to construct our
prototype using standard methods, where the servomotors are usually in joint areas like the
coxa–femur joint and femur–tibia joint; however, the size and weight of our servomotors
became assembly limiting considerations. Due to the poor weight distribution in the body and
joints of our quadruped robot, the design was precarious. To address this problem, we built a
Watt six-bar linkage mechanism with all servomotors in the coxa joint region and linkages
connecting the femur and tibia to their respective servomotors. This structure allowed us to
improve the weight distribution. Figure 1 shows the quadruped robot iron dog mini. Table 1
shows some feature comparisons between the existing and proposed robots.
 Robotics 2023, 12, x FOR PEER REVIEW 3 of

Robotics 2023, 12, 28 robot iron dog mini. Table 1 shows some feature comparisons between 3the
of 23
existing an
proposed robots.

Figure 1. Iron Dog mini,1.aIron

Figure low-cost quadruped
Dog mini, robot
a low-cost with Watt
quadruped six-bar
robot withlinkage mechanism.
Watt six-bar linkage mechanism.

Table 1. Comparison of existing

The models with
contributions and the proposedof
motivation model. The table
this work includes the
are described technology
below:
adoption, features,
• and
We introduce the novel Watt six-bar linkage mechanism for betterrobots.
mechanisms of previous and current research based on quadruped walking motion
• The motivation of this research was to develop a low-cost and modular quadrupe
assistiveState-of-the-Art
robot platformTechnology Adoption
for use in security and surveillance operations.
Robot Year • We innovatively designed robot Power
parts to make it modular, such that users can quick
Light Heavy Carry Low
assemble and Modular
disassemble the robot. Effi- Mechanism Agile
Weight Weight Load Cost
• cient
We explain the kinematic equations, demonstrate the URDF process, and test sever
Sony Aibo 1999 X - commands
- in the
- PyBullet -physics engine.
X Electric -
BigDog [26] 2005 - X • We discuss
X the material
- characteristics
- - and structural analysis
Hydraulic of the
Xrobot’s parts.
Scalf1 [6] 2011 - XTable 1. Comparison
X of-existing models
- with the- proposed model.
Hydraulic - the technolog
The table includes
Frog [27] 2013 - Xadoption, features,
X and-mechanisms
- of previous- and current research
- based on quadruped
- robot

Alpha Dog [28] 2012 - X X - - -

State-of-the-Art Adoption X
Hydraulic
Technology
Robot Year Light Heavy Carry Low Power
Hydraulic and Ef-
HyQ [29] 2011 - X X - - - Modular Mechanism
X Agile
Weight Weight Load Cost Electric
ficient
Sony Aibo 1999 ✓ - - - ✓
- Serial–parallel Electric -
Baby Elephant [30] 2013 - X X - - - X
BigDog [26] 2005 - ✓ ✓ - - Hybrid- Hydraulic ✓
AnyMal [31] 2016 Scalf1 [6]
- X 2011 X - - ✓ -✓ -X - Electric- Hydraulic
X -
Spot [32] 2017 Frog [27]
- X 2013 X - - ✓ X ✓ - X - Electric- -
X -
Alpha Dog [28] 2012 - ✓ ✓ - - - Hydraulic ✓
MIT Cheetah 3 [25] 2018 - X X - - X Electric X
Hydraulic
Unitree Laikago 2017 HyQ [29]
- X 2011 X - - ✓ X ✓ - X - Electric- X
and Electric
✓
Electric-five bar Serial–paral-
Stoch 2 [33] 2019 - [30]
Baby Elephant X 2013 - - - ✓ -✓ -- - X ✓
linkage- lel Hybrid
AnyMal [31] 2016 - ✓ ✓ - ✓bar
- Electric-Six Electric ✓
Proposed 2022 X - X X X X X
Spot [32] 2017 - ✓ ✓ - ✓ Linkage ✓ Electric ✓
MIT Cheetah 3 [25] 2018 - ✓ ✓ - - ✓ Electric ✓
UnitreeThe contributions and motivation of this work are described below:
2017 - ✓ ✓ - ✓ ✓ Electric ✓
Laikago
• We introduce the novel Watt six-bar linkage mechanism for better walking motion.
• The motivation of this research was to develop a low-cost and modular quadruped
assistive robot platform for use in security and surveillance operations.
• We innovatively designed robot parts to make it modular, such that users can quickly
assemble and disassemble the robot.
• We explain the kinematic equations, demonstrate the URDF process, and test several
commands in the PyBullet physics engine.
• We discuss the material characteristics and structural analysis of the robot’s parts.
 Electric-five
Stoch 2 [33] 2019 - ✓ - - - - ✓
bar linkage
Electric-Six
Proposed
Robotics 2023, 12, 28 2022 ✓ - ✓ ✓ ✓ ✓ ✓
4 of 23
bar Linkage

2. Design Principle
2. Design Principle
Most quadruped robots have 8–16 degrees of freedom (DoF). While these are the
most Most quadruped
common, robots
there are have other
various 8–16 degrees of freedom (DoF).
viable architectural While
options. Thethese
axis are theeight
of all most
common, there are various other viable architectural options. The axis of
joints (four hips, four knees) is parallel to itself, making up all eight degrees of freedom. all eight joints
(four hips,an
Although four knees)
8-DoF is parallel
robot to itself,
lacks the making
hip joint’s up all eight
transverse degrees
swing of freedom.
flexibility, Although
they are none-
an 8-DoF robot lacks the hip joint’s transverse swing flexibility, they are nonetheless
theless easy to control and capable of fast forward and backward motion. However, the
easy to control and capable of fast forward and backward motion. However, the motion
motion performance of 8-DoF quadruped robots is hindered by their poor steering capa-
performance of 8-DoF quadruped robots is hindered by their poor steering capabilities and
bilities and inability to carry out transverse motion. A quadruped robot with 16 DoF has
inability to carry out transverse motion. A quadruped robot with 16 DoF has more joints
more joints for agile maneuvering, but is more complicated and, hence, more difficult to
for agile maneuvering, but is more complicated and, hence, more difficult to control. To
control. To allow for three rotatable joints in each leg, we developed a 12 DoF design. The
allow for three rotatable joints in each leg, we developed a 12 DoF design. The notion of a
notion of a four-legged robot was mostly inspired by domesticated feline and canine ani-
four-legged robot was mostly inspired by domesticated feline and canine animals, which
mals, which are distinguished from other domestic species by the presence of an endo-
are distinguished from other domestic species by the presence of an endoskeleton, allowing
skeleton, allowing for greater movement. The development of a quadruped robot with 12
for greater movement. The development of a quadruped robot with 12 degrees of freedom
degrees of freedom is a key part of our study. Our intricate quadruped robot is made up
is a key part of our study. Our intricate quadruped robot is made up of many different
of many different elements with rotating joints. The coxa, femur, and tibia are the most
elements with rotating joints. The coxa, femur, and tibia are the most vital components of
vital components
the proposed of the proposed
quadruped robot. quadruped robot.
A
A rod
rod end,
end, servo
servo arm,
arm, rod
rod end
endlinker,
linker, and
and servo
servo horn
hornare
arealso
alsoincluded
includedin inthe
thedesign.
design.
Three
Three typical bones make up each leg: The coxa (hip bone), femur (thigh bone), and
typical bones make up each leg: The coxa (hip bone), femur (thigh bone), and tibia
tibia
(shin
(shin bone).
bone). Twelve
Twelve servo
servo actuator
actuator motors
motors are
are employed
employed to to translate
translate the
the links
links and
and modify
modify
the
the joints.
joints. The
The different
different parts
parts of
of robot
robot are
are shown
shown in in Figure
Figure 2.
2.

Figure 2.
Figure Differentessential
2.Different essentialparts
partsof
ofthe
thequadruped
quadrupedrobot
robotequipped
equippedwith
withsix-bar
six-barlinkage
linkagemechanism.
mechanism.

The locomotion
The locomotion ofof the
the quadruped
quadruped robot
robot depends
depends onon the
the translations
translations and
and rotations
rotations of
of
the coxa joint, femur, and tibia. The kinematic analysis defines the motion of the quadruped
the coxa joint, femur, and tibia. The kinematic analysis defines the motion of the quadru-
robot’s
ped legs,legs,
robot’s where we can
where correlate
we can the the
correlate forward andand
forward inverse kinematics.
inverse The
kinematics. objective
The of
objective
this study is to create a quadruped robot with a body made entirely of 3D-printed parts
of this study is to create a quadruped robot with a body made entirely of 3D-printed parts
and reasonably priced servo motors. Given its size and capabilities, the robot has been
and reasonably priced servo motors. Given its size and capabilities, the robot has been
built to be able to carry a reasonable payload. As such, we designed the robot’s torso in
built to be able to carry a reasonable payload. As such, we designed the robot’s torso in
such a way that we additional components can be mounted on our robot. Our design
such a way that we additional components can be mounted on our robot. Our design
principle allows the quadruped robot to be modular, such that the user can easily replace
principle allows the quadruped robot to be modular, such that the user can easily replace
any broken or defective components without replacing the whole robot, massively reducing
any broken or defective components without replacing the whole robot, massively reduc-
the affordability and maintenance costs of our quadruped robot. Our modular-based design
ing the affordability and maintenance costs of our quadruped robot. Our modular-based
is shown in Figure 3. Table 2 shows essential parameters of Iron dog mini.
design is shown in Figure 3. Table 2 shows essential parameters of Iron dog mini.
Robotics
Robotics 2023, 12, 28
2023, 12, x FOR PEER REVIEW 55 of
of 24
23

Figure 3.
Figure Different modules
3. Different modules of
of Iron
Iron dog
dog mini.
mini.

Table 2.
2. Essential Parameters of Iron Dog mini, including the height, weight, and mass values
values of
different links.
Value Value
Value Total Value Total
Parameter Single Part Total Parameter
Single Part Total
Parameter Height
Height Width
Width Number
Number Single Part Mass
Mass (g) Parameter Height
Height Width
Width Number Single Part
Number Mass
Mass (g)
(cm) (cm) of Uses Mass
Mass (g) (cm) (cm) ofof
Uses Mass
(cm) (cm) of Uses (g) (cm) (cm) Uses Mass (g) (g)
(g) (g)
Base 30 6 2 92 184 Servo Arm 2.4 0.63 4 5 20
Base 30 6 2 92 184 Servo Arm 2.4 0.63 4 5 20
Femur 12 2.5 4 55 220 Long rod end 4.5 0.25 4 7 28
Femur 12 2.5 4 55 220 Long rod end 4.5 0.25 4 7 28
Tibia 17 2.5 4 48 192 Short Rod end 1.2 0.2 4 1.8 7.2
Tibia 17 2.5 4 48 192 Short Rod end 1.2 0.2 4 1.8 7.2
FrontHead
Front Head
Actuator
Actuator 5.65.6 22 12
12 8080 960
960 77 1212 11 5555 55
55
Module
Module
RollingServo
Rolling Back
BackShell
Shell
Servo 4 4 22 44 1010 4040 77 1212 11 4848 48
48
mount Fortorso
For torso lock
lock
mount
SideServo Side
Sidepitch
pitch
Side
Servo 4 4 22 44 88 3232 mount
mount(double
(double 1212 5.6
5.6 22 120
120 240
240
Mount
Mount ServoModule)
Module)
Servo
ServoHorn
Servo 0.14 0.025
0.14 0.025 44 66 2424 Screws
Screws 0.30.3 0.05
0.05 74
74 11 74
74
Horn
Free linker 6.32 4.60 4 12 48 Battery - - 1 250 250
Free Total Weight of Robot
6.32 4.60 4 12 48 Battery - After Fully
- 1 250 250
Other
linkerParts - - - 150 150 2572.2g
Mounted
Other
- - - 150 150 Total Weight of Robot After Fully Mounted 2572.2 g
Parts
3. Working Mechanism
Iron dog
3. Working mini consists of 12 servo motors that control the joint angles at the coxa,
Mechanism
femur, and tibia. Figure 2 shows detailed information on the locations and orientations of
Iron dog mini consists of 12 servo motors that control the joint angles at the coxa,
the servo motors. Our design provides better weight balance, as we concentrated the
femur, and tibia. Figure 2 shows detailed information on the locations and orientations
weight close to the four corners of the robot’s main body by locating the three servo mo-
of the servo motors. Our design provides better weight balance, as we concentrated the
tors of a single leg at the hip joint area. Actuator 1 controls the joint angle of the coxa. The
weight close to the four corners of the robot’s main body by locating the three servo motors
hip
of a of the leg
single robot is directly
at the hip jointcoupled with the
area. Actuator servo motor,
1 controls where
the joint theofhip
angle the joint
coxa.helps to
The hip
stabilize the is
of the robot robot during
directly motion.
coupled with the servo motor, where the hip joint helps to stabilize
Actuator 2 controls
the robot during motion. the Femur, and the femur link is directly coupled with the servo
motor. Actuator
Actuator 3 controls
2 controls thethe Tibia,and
Femur, butthe
thefemur
Tibia link
jointisisdirectly
not directly
coupledcoupled with
with the the
servo
servo motor. To control the Tibia link, we introduced the Watt six-bar linkage
motor. Actuator 3 controls the Tibia, but the Tibia joint is not directly coupled with the mechanism
into
servoour design,
motor. which isthe
To control a unique approach.
Tibia link, Conventional
we introduced the Wattmechanisms such mechanism
six-bar linkage as four-bar
linkage
into ourmechanisms
design, which have
is asome
uniquemotion constraints
approach. due to their
Conventional limited design
mechanisms such asvariables.
four-bar
We have pointed out some advantages of the six-bar linkage mechanism
linkage mechanisms have some motion constraints due to their limited design variables. over the four-bar
linkage
We havemechanism.
pointed out some advantages of the six-bar linkage mechanism over the four-bar
•linkage
Themechanism.
watt six-bar linkage mechanism provides a greater range of motion for leg actu-
• ation thansix-bar
The watt the four-bar
linkagelinkage mechanism.
mechanism provides a greater range of motion for leg actua-
tion than the four-bar linkage mechanism.
Robotics 2023, 12, 28 6 of 23

Robotics 2023, 12, x FOR PEER REVIEW 6 of 24

• The watt six-bar linkage mechanism produces leg motion during gait generation,
which is very close to the leg motion of a four-legged animal compared to the four-bar
• The watt mechanism.
linkage six-bar linkage mechanism produces leg motion during gait generation,
• which
The four-barclose
is very to the
linkage leg motionhas
mechanism of amany
four-legged
motionanimal compared
constraints. to the four-
Therefore, the four-bar
bar linkage mechanism.
linkage mechanism robot has a limited range of motion for its leg. The four-bar linkage
• The four-bar linkage mechanism has many motion constraints. Therefore, the four-
has a total of eight design variables.
bar linkage mechanism robot has a limited range of motion for its leg. The four-bar
• The watt six-bar linkage mechanism has fourteen design variables.
linkage has a total of eight design variables.
•• Thewatt
The six-bar linkage
six-bar has
linkage more motion
mechanism parameters
has fourteen than
design the four-bar linkage, increasing
variables.
• the range of motion.
The six-bar linkage has more motion parameters than the four-bar linkage, increasing
the range of motion.
Additionally, the gait generation produced by the watt six-bar linkage mechanism
Additionally,
is found the gaitcomparable
to be highly generation produced
to that byof the watt six-bar linkage
a four-legged animal, mechanism
making it is an ideal
found to befor
candidate highly
use incomparable
quadrupedalto that of a four-legged
robots. Therefore,animal, making
the six-bar it an is
linkage ideal candi-
a more convenient
date for use in
mechanism forquadrupedal robots. of
the leg actuation Therefore,
quadruped the six-bar
robots.linkage is a more convenient
mechanism for the leg actuation of quadruped robots.
We designed and optimized the six-bar linkage mechanism to manipulate the tibia link
We designed and optimized the six-bar linkage mechanism to manipulate the tibia
effectively, and the bar linkages are shown in Figure 4. The Watt six-bar linkage mechanism
link effectively, and the bar linkages are shown in Figure 4. The Watt six-bar linkage mech-
has more stability, adequate movement, and better motion efficiency than conventional
anism has more stability, adequate movement, and better motion efficiency than conven-
four- and
tional four-five-bar
and five-barlinkage mechanisms.
linkage mechanisms. The
Theactuator
actuatorthat
that is
is used has aasignificant
used has significant impact
on the performance of a quadruped robot’s movement. Considering
impact on the performance of a quadruped robot’s movement. Considering the robot’s the robot’s weight,
weight, the actuator must provide significant torque while maintaining a rapid response time and
the actuator must provide significant torque while maintaining a rapid response
a compact
time footprint.
and a compact Brushless
footprint. motors
Brushless are widely
motors used,
are widely due
used, dueto to
their
theirexcellent
excellent dynamic
qualities;qualities;
dynamic however,however,
they arethey
typically more costly
are typically more and more
costly andextensive in size.inThus,
more extensive size. we used
Thus, we used a servo
a high-voltage high-voltage
motor servo
whichmotor
has awhich hasgear
built-in a built-in gear Moreover,
reducer. reducer. Moreover,
we used a metal
we
gearused a metal
servo motor,gearproviding
servo motor,35 providing
kg torque35per kg centimeter.
torque per centimeter. The single-leg
The single-leg configuration is
configuration is
shown in Figure 5. shown in Figure 5.

Robotics 2023, 12, x FOR PEER REVIEW 7 of 24

WattSix-Bar
Figure4.4.Watt
Figure Six-Bar linkage
linkage mechanism.
mechanism.

Single-leg
Figure5.5.Single-leg
Figure configuration.
configuration.

3.1. Mathematical Analysis

Quadruped robots use two types of kinematics: Forward and inverse kinematics. Ro-
bot end-effector joint positions are determined using forward kinematics, whereas manip-
Robotics 2023, 12, 28 7 of 23

3.1. Mathematical Analysis

Quadruped robots use two types of kinematics: Forward and inverse kinematics.
Robot end-effector joint positions are determined using forward kinematics, whereas
manipulator joint values can be derived from inverse kinematics. In order to calculate
the forward kinematics, we multiply the number of matrices. To better understand how
forward kinematics operate in quadruped robots, we constructed a block-based dummy
quadruped robot and figured out all the frames. In our previous work, we have demon-
strated a comprehensive derivation of the true nature of forward kinematics [15]. Figure 6
depicts all the various frames and coordinate systems in detail. Additionally, Figure 6
was used to reach the foot frame, which serves as the illustration’s final effector. The
zeroth frame and hip joints represent the coxa coordinate X0 , Y0 , Z0 . To find the forward
kinematics equation, we need to multiply each frame from zeroth to the fourth. If we
want to reach the zeroth frame from the first frame, there will be a translation, denoted
by L1 . When reaching the first frame from the second frame, there is no translation but,
instead, a rotation, denoted by θ1 . Passing from frame two to frame three involves L2 and
θ2 for translation and rotation, respectively. Finally, we have L3 translation and θ3 rotation
between the third and fourth frames. After multiple translations and rotations, we obtain
the final transformation matrix, according to Denavit Hartenberg’s convention. We can
determine the matrix representing all possible transformations between frames 0 and 4 [15]
as T04 = T01 ∗ T12 ∗ T23 ∗ T34 :
 
φ11 φ12 φ13 φ14
φ21 φ22 φ23 φ24 
Robotics 2023, 12, x FOR PEER REVIEW T04 = 
φ31 φ32 φ33 φ34 .
 8 of (1)
24

φ41 φ42 φ43 φ44

Figure6.
Figure Frame-by-framecoordinate
6. Frame-by-frame coordinate system
system of
of aa quadruped
quadruped robot
robot using
using 3D
3D design.
design.

From this transformation matrix, our end effector value can be determined from the
Based on Figure 7,
fourth column (i.e., φ14 , φ24 , and φ34 ). After the forward kinematics are established, inverse
kinematics are required to regulate the jointθ1settings.
= α3 − α We (2)
1 . sketched through the actual model
for better understanding.
In △ ADC,
Based on Figure 7,
θ1 = α3√x−42α+1 y. 42 − L21 (2)
α3 = arctan ( ). (3)
L1

In ◻ ABCE,
α1 + α2 = 90∘ (4)

⇒ α1 = 90∘ − α2 . (5)
Furthermore, in △ ABC,
α2 + α4 + ∠B = 180∘ , (6)

α2 + α4 = 180∘ − 90∘ [∠B = 90∘ ], (7)

 Now, substituting the value of AC2 into Equation (32),
L22 + L23 − X42 − Y42 − L21 − z42 AD = L2
α = arccos ( )[ ], (36)
2L2 L3 CD = L3

L22 + L23 − x42 − y42 − L21 − z4

Robotics 2023, 12, 28 θ3 = 180∘ − arccos ( ). 8(37)
of 23
2L2 L3

Determiningθθ11utilizing
Figure7.7.Determining
Figure utilizingthe
thefront
frontview
viewof
ofthe
therobot
robotleg.
leg.

In 4ADC,
q 
x24 + y24 − L21
α3 = arctan . (3)
L1

In ABCE,
α1 + α2 = 90◦ (4)
⇒ α1 = 90◦ − α2 . (5)
Furthermore, in 4ABC,
α2 + α4 + ∠B = 180◦ , (6)
◦ ◦ ◦
α2 + α4 = 180 − 90 [∠B = 90 ], (7)
α2 = 90◦ − α4 . (8)
FigureSubstituting this
8. Determining θ2 value into
utilizing theEquation
side view.(5) :

α1 = 90◦ − (90◦ − α4 ) (9)

⇒ α1 = α4 . (10)
In 4ABC,
− y4
 
α4 = arctan . (11)
x4
Now, substituting the values of α1 and α3 into Equation (2),

θ1 = α3 − α1 (12)
q 
x42 + y24 − L21
 − arctan −y4 .
 
⇒ θ1 = arctan (13)
11 x4
Figure 9. Determining θ3.
Based on Figure 8,
θ2 = −90◦ + α1 . (14)
In summary, in the derivation process of θ1, θ2 and θ3 in the inverse kinematic anal-
ysis, we took the left front leg of our quadruped robot into consideration. We positioned
Robotics 2023, 12, 28 9 of 23
Figure 7. Determining θ1 utilizing the front view of the robot leg.

Figure 8.
Figure Determining θθ22 utilizing
8. Determining utilizing the
the side
side view.
view.

From ABFE,
α1 + α2 + α3 = 90◦ (15)
∴ α1 = 90◦ − α2 − α3 . (16)
From 4ABC,  
− Z4
α3 = arctan q . (17)
x42 + y24 − L12

From 4 ACF,  
CF
α2 = arctan . (18)
AF
Furthermore, from 4CDF,
CF
sin θ3 = , (19)
Figure 9. Determining θ3. CD
CF = CD sin θ3 , (20)
In summary, in the derivation process of θ1, θ2 and θ3 in the inverse kinematic anal-
our=quadruped
ysis, we took the left front leg ofCF L3 sin θ3 [CD = Linto
robot 3 ], consideration. We positioned
(21)
DF
cos θ3 = , (22)
CD
DF = L3 cos θ3 [CD = L3 ], (23)
AF = AD + DF, (24)
AF = L2 + L3 cos θ3 [ AD = L2 ]. (25)
Similarly,  
L3 sin θ3
∴ α2 = arctan . (26)
L2 + L3 cos θ3
Now, substituting the values of α2 and α3 into Equation (16),
 
−
 
L 3 sin θ 3 Z4
α1 = 90◦ − arctan − arctan q , (27)
L2 + L3 cos θ3 x +y −L 2 2
4 4 1

θ2 = −90◦ + α1 (28)
 
− z4
 
L3 sin θ3
⇒ θ2 = −arctan −arctan q . (29)
L2 + L3 cos θ3 x + y2 − L
2
4 4 1
Robotics 2023, 12, 28 10 of 23

Based on Figure 9,
θ = 180◦ − α.
3 (30)
Figure 8. Determining θ2 utilizing the side view.

Figure 9.
Figure Determining θθ33..
9. Determining

4ACD, in the derivation process of θ1, θ2 and θ3 in the inverse kinematic anal-
In summary,
In
2 2
AC
ysis, we took the left front leg =
ofAD + CD2 − 2 ·robot
our quadruped AD · CD
into· consideration.
cos α (31)
We positioned
!
AD2 + CD2 − AC2
⇒ α = arccos . (32)
2 · AD · CD

In 4ABC,
AC2 = AB2 + BC2 , (33)
q 2
AC2 = X24 + y24 + L21 + (−z4 )2 , (34)

AC2 = x24 + y24 + L21 + z24 . (35)

2
Now, substituting the value of AC into Equation (32),
!
L22 + L23 − X24 − Y24 − L21 − z24

AD = L2
α = arccos , (36)
2L2 L3 CD = L3
!
2 2 2 − y2 − L2 − z
L + L − x 4
θ3 = 180◦ − arccos 2 3 4 4 1
. (37)
2L2 L3

In summary, in the derivation process of θ1 , θ2 and θ3 in the inverse kinematic analysis,

we took the left front leg of our quadruped robot into consideration. We positioned the
robot’s leg to mimic its normal standing posture (side view). Using the front view of our
leg, we positioned the coxa angle, which mirrors the roll position of the robot, allowing us
to identify θ1 . From the two right-angle triangles ABC and ADC in Figure 7, we were able
to obtain the equation for θ1 that we required. Using the profile view of our leg in Figure 8,
we determined θ2 . Two right-angle triangles, ABC and CDF, and one obtuse triangle, and
ADC, are shown in Figure 8, and from these triangles we derived the desired equation for
θ2 . Using the profile of our leg, we were able to compute θ3 from Figure 9. Figure 9 depicts
two triangles, one with a right-angle (ABC) and another with an acute angle (ADC), from
which our desired equation for θ3 was derived.
q  
θ1 = arctan2 2 2 2
x4 + y4 − L1 , L1 − arctan2(−y4 , x4 ), (38)

 q 
θ2 = −arctan2( L3 sin(θ3 ), ( L2 + L3 cos(θ3 )))− arctan2 −z4 , x42 + y24 − L21 , (39)
 𝐿22 + 𝐿23 − 𝑥42 − 𝑦42 − 𝐿21 − 𝑍42 (40)
𝜃3 = 𝜋 − arccos ( ).
2𝐿2 𝐿3

3.2. Simulation Model Workflow

Robotics 2023, 12, 28 Workflow Procedure 11 of 23

Next, we describe the computational method used to simulate our quadruped robots.
We used the Python-based Pybullet physics engine in the Ubuntu!operating system for
L22 + Lstages,
this scientific simulation following multiple 2 − x 2 − y2 − L2 − Z 2
3 4 as presented
4 1 in
4 Figure 10. Parts were
θ3 = π − arccos . (40)
created in the SolidWorks software using precise 2L measurements.
L
2 3
A great deal of thought and care must go into the design of each component before
it can
3.2. be assembled
Simulation Modelinto a functioning robot. There are several essential parts for robots.
Workflow
About
Workflow 64 unique parts were conceived for this study. Each of the robot’s legs consists of
Procedure
threeNext,
separate parts (thethe
we describe coxa, femur, and method
computational tibia), and thetodesign
used simulateis based on the principle
our quadruped robots.
of
We used the Python-based Pybullet physics engine in the Ubuntu operatingthere
three degrees of freedom. Throughout the entirety of the design phase, wasfor
system a
continuous
this scientificprocess of considering
simulation the design’s
following multiple framework
stages, in addition
as presented to10.
in Figure theParts
materials
were
that were
created into beSolidWorks
the utilized. software using precise measurements.

Figure 10.
Figure Workflow diagram
10. Workflow diagram for
for the
the proposed
proposed simulation
simulation model.
model.

A great tests,
Certain deal of thought
such andelement
as finite care must go into the
modeling design
(FEM), of each
finite component
element analysis before
(FoS),
it can be assembled into a functioning robot. There are several essential parts for robots.
and strain, must be performed on the robot’s core components to ascertain their durability
About
and the64 unique parts
maximum werethe
payload conceived for withstand.
design can this study. In
Each
theof the robot’s
“Material legs consists
Analysis” of
section,
three separate parts (the coxa, femur, and tibia), and the design is based on the principle
we summarize relevant information in this aspect.
of three degrees of freedom. Throughout the entirety of the design phase, there was a
continuous process of considering the design’s framework in addition to the materials that
were to be utilized.
Certain tests, such as finite element modeling (FEM), finite element analysis (FoS), and
strain, must be performed on the robot’s core components to ascertain their durability and
the maximum payload the design can withstand. In Section 4.1, we summarize relevant
information in this aspect.
After making all the pieces, they must be put together in a way that allows the robot to
operate well. As there are three movable and one fixed joint, each joint must be mated with
extreme precision. The six-bar linkages were created to generate exterior revolute joints in
addition to the leg joints. A rod-end and actuator were utilized to rotate the desired joint in
the six-bar linkages.
For the simulation process, creating a URDF script was very much necessary. URDF
contains all of the robot data, such as link length, link mass, material, inertia, geometry,
and the path of the robot STL file. STL files are exported for every part of the robot, which
contain the 3-dimensional surface geometry of the parts. This format is suitable for use
with the PyBullet physics engine.
After that, we had to link our robot’s gait pattern to the kinematics algorithm. The
various link parameters and joint angles were contained in the Kinematics Algorithm,
which provides 3D foot positions for us. We tested several gait patterns. To develop
a walking motion for a quadruped robot, kinematics is insufficient. Gait generation is
required to operate the robot in both real and virtual simulations. There are various gait
patterns available, including the creep, trot, pace, cantor, and gallop gaits. We created a
creep gait, trot gait, and pace gait for our simulation investigation, through investigating
the walking patterns of genuine dogs and cats. Manually adjusting the arms and joints
Robotics 2023, 12, 28 12 of 23

of the robot to the appropriate angle is challenging, although it is plausible to adjust the
motor inclination by issuing repeated instructions, this is considered inappropriate
A kinematics-based system for controlling both virtual and actual robots was created
as part of this research project. Using the simulation model, complete control can be exerted
over the angle of each joint, as well as the foot locations and the end effectors. This method
incorporates both forward and inverse kinematics. The values of the end effectors are
provided as output after all the robot’s connection parameters, and the manipulator’s
required angle is considered.
We can also set the end effector value to achieve the appropriate location. One of the
most difficult challenges is taking control of a quadruped robot from its center of gravity. If
the four-legged equations are not linked to the center of gravity point, the robot will never
be able to stand on its own. For example, if the robot’s rear side is heavier than its front side,
then the device will always be unable to walk, as it will always fall on its back. To solve this
problem, we first located the center points of the robot and then directly connected them to
each leg. After completing all the stages (i.e., Part design, Assembly, URDF creation, Gait
pattern, and kinematics setting), we executed all the scripts for simulation using the PyBullet
physics engine. We connected our implemented robot with the simulation model during the
simulation process. The hardware sensors transmitted sensory data every 20 milliseconds.
We can control the simulation remotely, either using a controller or giving a command.

3.3. Step Trajectory and Gait Generation

There are various different gait patterns used by quadruped robots, including creep,
trot, and gallop, among others. Creep and trot are often employed and swapped, based on
the surroundings and the desired pace of mobility. Trot serves as the most frequent gait.
The two diagonal legs move in tandem, with the left front leg and right back leg moving
first and the right front leg and left back leg moving second. According to the trot diagram
in Figure 11, the trot is a type of gait in which two legs are activated in tandem. At one
moment, the front right leg and the back left leg swing, while the right back leg and left
Robotics 2023, 12, x FOR PEER REVIEW 13 of 24
front leg form a stance. Two motions are present: Swing and stance. The stance phase
means that the toe has contact with ground; during the swinging phase, the toe is in the air.

Figure11.
Figure 11.Trot
TrotGait
GaitPatterns.
Patterns.

We
Weused usedthe theBernstein
Bernsteinpolynomial polynomialand andBezierBeziercurve curvefor forour ourrobot’s
robot’strajectory
trajectorygenera- gener-
tion
ation [34].
[34].We Wecreated
createda a14-point 14-pointBezierBeziercurve curvetotocontrolcontrolthe thefootstep
footsteptrajectory.
trajectory.The Thecontrol
control
points
pointsthat thatmake makeup upthe theBezier
Beziercurvecurvewere wereused usedto tobuild
buildthe theswingswingphase. phase. EquationEquation (41) (41)
shows
shows the the method
method for calculation calculation of of the
thebinomial
binomialfactor factor(d, (d,l),l),and
andEquation
Equation(42) (42)is is used
used to
to simplify
simplify thethe Bezier
Bezier curvecurve calculation,
calculation, wherewhere K isKthe is the currentcurrent point pointindex; index;t is the t is time;
the time; and
and c denotes
c denotes the coordinate.
the coordinate. Here,Here,we have we have P0(a0,Pd00(a ), 0P, 1d
(a01),, dP11),(aP12,(ad21, ),d2P),2 (a
P32(a, 3d, 2d),3),PP3 (a
4(a3 ,4,dd34),
),
PP45(a
(a45, d54),), PP65(a
(a6,5 ,dd6),5 ),P7P(a6 (a 6 ,7),
7, d d6P),8(a
P78,(ad78,),dP79),(aP9,8d
(a9),
8, Pd108 ),
(aP 10,9 (a , dP911),(aP1110, (a
d109 ), d1110),, P
d12
10(a),12P,11d(a 11P, 13
12), d(a ),,
1113
Pd12
13(a 12 , the
) as d12 ),14Pcontrol
13 (a13 , d 13 ) as the
points. 14 control
Equations (43)points.
and (44) Equations
are used (43) for curveand (44) are used for
construction. In
curve construction. In Figure 12a, we show the
Figure 12a, we show the trajectory steps of swing and stance, where t = 0.5 seconds; trajectory steps of swing and stance, where in
Figure 12b, the parametric Bezier curve is shown (the curve was plotted using MATLAB).
We show all the control point values in Table 3.
Robotics 2023, 12, 28 13 of 23

t = 0.5 seconds; in Figure 12b, the parametric Bezier curve is shown (the curve was plotted
using 11.
Figure MATLAB).
Trot Gait We show all the control point values in Table 3.
Patterns.
d!
We used the Bernstein polynomial f(d, l)and= Bezier curve , for our robot’s trajectory gener- (41)
K!(d − l)!
ation [34]. We created a 14-point Bezier curve to control the footstep trajectory. The control
points that make up the Bezier curve were used to build theK swing phase. Equation (41)
b(t, K, c) = f(N, K) · (1 − t)(N−K)·t ·c , (42)
shows the method for calculation of the binomial factor (d, l), and Equation (42) is used to
a =the
simplify (b(Bezier
t, 0, a0 )curve (t, 1, a1 ) + bwhere
+ bcalculation, (t, 2, a2K ) +isbthe
(t, 3, a3 ) + bpoint
current 4 ) + bt(is
(t, 4, aindex; t, 5,
thea5time;
) and
c denotes the coordinate.+Here, b(t, 6,we + b(P
a6 )have t,07,
(aa0,7d) 0+
), b t, 8,
P(1(a a81),
1, d )+P2b(a(2t,, d9,2),
a9P) 3(a3, d3), P4(a4, d (43)
4),

P5(a5, d5), P6(a6, d6), P7(a7, +

d7),b(Pt,8(a
10,8, a +9(a
8),) P
d10 b9(,t,d11,
9), Pa10 ) 10
11(a P11a(a1211), +
+, bd(10t,),12, d11b),(t,P12
13,
(a12 , d))12,), P13(a13,
a13
d13) as the 14 control points. Equations (43) and (44) are used for curve construction. In
d = (b(t, 0, d0 ) + b(t, 1, d1 ) + b(t, 2, d2 ) + b(t, 3, d3 ) + b(t, 4, d4 ) + b(t, 5, d5 )
Figure 12a, we show the trajectory steps of swing and stance, where t = 0.5 seconds; in
+ b(t, 6, d6 ) + b(t, 7, d7 ) + b(t, 8, d8 ) + b(t, 9, d9 ) (44)
Figure 12b, the parametric Bezier curve is shown (the curve was plotted using MATLAB).
+ b(t, 10, d10 ) + b(t, 11, d11 ) + b(t, 12, d12 ) + b(t, 13, d13 )).
We show all the control point values in Table 3.

(a) (b)
Figure
Figure12.
12.(a)(a)
Step trajectory
Step stance
trajectory andand
stance swing phase;
swing (b) trajectory
phase; generation
(b) trajectory using 14-point
generation Bezier
using 14-point
curve.
Bezier curve.

Table 3. Control Points for Bezier curve. d!

f(d, l) = , (41)
K! (d − l)!
Control Points (mm) (mm)
K
p0 K) ⋅ (1 − t)(N−K)⋅t ⋅cd,0
a0 b(t, K, c) = f(N,100 −252.4 (42)
pa1 = (b(t, 0, a0 ) + b(t,
a1 1, a1 ) + b(t, 2, a100
2 ) + b(t, 3, a3 ) + b(t,
d1 4, a4 ) + b(t, 5, − )
a5250
p2 +a2b(t, 6, a6 ) + b(t, 7,
140a7 ) + b(t, 8, a8 ) +db(t,
2
9, a9 ) −250 (43)
p3
+a b(t, 10, a10 ) + b(t,
150
11, a11 ) + b(t, 12, a
d 12
) + b(t, 13, a13 ) ),
−180
3 3

pd
4 =
(b(t, 0, d0 ) + ab(t, )
4 1, d1 + b(t, 2,150 d2 ) + b(t, 3, d3 ) + b(t,
d4 4, d4 ) + b(t, 5,−d180
5)

p5 +a5 b(t, 6, d 6 ) + b(t, 7, d

150 7
) + b(t, 8, d )
8 d+ b(t, 9, d 9 ) −180
(44)
5
+ b(t, 10, d10 ) + b(t, 11, d11 ) + b(t, 12, d12 ) + b(t, 13, d13 ) ).
p6 a6 −100 d6 −70
p7 a7 0 d7 −180
p8 a8 0 d8 −160.5
p9 a9 −151.5 d9 −160.5
p10 a10 −151.5 d10 −160.5
p11 a11 −141 d11 −250
p12 a12 −100 d12 −250
p13 a13 −90 d13 −260
Robotics 2023, 12, 28 14 of 23

4. Results and Discussion

4.1. Material Analysis
As our quadruped robot was designed with a low cost in mind, we had to choose a
material for 3D printing that is both cheap and strong for fabrication. Therefore, we chose
ABS plastic (Acrylonitrile butadiene styrene) as our printing material.While designing
and developing our low-cost quadruped robot, we used SolidWorks to create around
64 distinct components. We conducted load simulation in SolidWorks for various portions
(i.e., femur and tibia) of our quadruped robot, in order to better understand the strength
and stiffness of our design and the material used. We obtained analytical data, such as the
strain, displacement, and Factor of Safety (FoS) for the femur and tibia of our four-legged
robot throughout the simulation process. By looking at these data, we could make changes
to how the components were made, thus improving their performance in the long run.
Table 4 shows the mechanical properties of abs plastic.

Table 4. Mechanical Properties of ABS Plastic.

Property Value Unit

Tensile strength 30 MPa
Mass density 1020 kg/m3
Elastic modulus 2000 MPa
Shear modulus 318.9 MPa
Poisson’s ratio 0.394 N/A

In SolidWorks, we have utilized our femur and tibia designs to perform load analysis
simulation. During load simulation, we considered approximately 7.5 kg of payload on
our quadruped robot. The weight of our quadruped robot is around 2.752 kg. So, with
the added payload, the total weight would be close to 10.25 kg, equivalent to a 100 N load.
During the simulation process, we have obtained diagrams showing the performance of our
designed parts under load. In Figure 13a,b, the displacement diagrams for the femur and
tibia parts are shown, with the scale on the right indicating the degree of deformation of
the components (in mm). We can see that, under a 100 N load on the femur, the maximum
Robotics 2023, 12, x FOR PEER REVIEW 15 of
deformation was 6.596 × 10−3 mm. When the 100 N load was applied on the Tibia, the
maximum deformation was 5.254 × 10−1 mm. For visual representation of the deformation,
the components are highlighted with multiple colors, indicating the most deformed area
(red color) and deformed area (red color)
the least deformed and the
area (blue least deformed
color), accordingarea (blue
to the color),
scale. Noteaccording
that theto the sca
Note that the deformation scale was adjusted
deformation scale was adjusted for clearer visual analysis. for clearer visual analysis.

Figure 13. (a)analysis

Figure 13. (a) Displacement Displacement analysis
diagram diagram
for femur; for (b)
and femur; and (b) Displacement
Displacement analysis diagram
analysis diagram
for tibia. tibia.

In the simulation process, we also obtained Factor of Safety diagrams, as shown

Figure 14a,b for the femur and tibia, respectively. The Factor of Safety indicates the safe
of any mechanical component during operations; therefore, we can determine how sa
the component is under certain conditions, according to the factor of safety scale.
Robotics 2023, 12, 28 15 of 23
Figure 13. (a) Displacement analysis diagram for femur; and (b) Displacement analysis diagram for
tibia.

In the simulation
In the process,
simulationwe also obtained
process, FactorFactor
we also obtained of Safety diagrams,
of Safety as as
diagrams, shown
shownin in
Figure 14a,b for the14a,b
Figure femur forand tibia,and
the femur respectively. The Factor
tibia, respectively. of Safety
The Factor indicates
of Safety the
indicates thesafety
safety
of any mechanical
of anycomponent during operations;
mechanical component therefore,
during operations; we can we
therefore, determine how safe
can determine how the
safe
component isthe component
under certainis conditions,
under certainaccording
conditions,toaccording to the
the factor of factor
safetyofscale.
safety scale.

Figure 14. (a) Factor

Figure of
14.Safety analysis
(a) Factor diagram
of Safety analysis for femur;
diagram for and (b)and
femur; Factor of Safety
(b) Factor analysis
of Safety diagram
analysis diagram
for tibia. for tibia.

A base number indicates the minimum factor of safety where the component is safe.
A base number indicates the minimum factor of safety where the component is safe.
A value lower than the base number will indicate the failing point of the component. The
A value lowerfactorthanofthe base number will indicate the failing point of the component. The
safety can vary for different materials, and the process of finding the factor of
factor of safety can vary
safety varies forforductile
different materials,
and brittle andFor
materials. the our process of finding
ABS Plastic, which isthe factor
a brittle of
mate-
safety varies for ductile and brittle materials. For our ABS Plastic, which is a brittle
rial, the factor of safety is determined as the ratio of ultimate tensile stress to working material,
the factor of safety
stress. is determined as the ratio of ultimate tensile stress to working stress.
The diagrams The diagrams
show show the
the Factor of Factor
Safetyofscale
Safetyon scale
theon the right,
right, whichwhich indicates
indicates thethesafety
safety
factor of components. Here, the factor of safety for the femur ranged from 3.6 up3.5to×
factor of components. Here, the factor of safety for the femur ranged from 3.6 up to
104, depending
104 , REVIEW
× PEER on theonworking
the working
stressstress in different parts of the femur. Forthe
thetibia,
tibia,the
the
Robotics 2023, 12,3.5
x FOR depending in different parts of the femur. For 16 of 24
factor of safety ranged from 2.510 to 3.111 7 × 107. The components are highlighted with
factor of safety ranged from 2.510 to 3.111 × 10 . The components are highlighted with
multiple colors, indicates the areas where the components are safe (blue area) and where
the components are at colors,
multiple the risk of failing
indicates (red where
the areas area). the components are safe (blue area) and where
From the the components
simulation are at the
process, we risk of failing
also obtained(red area).
strain analysis diagrams, as shown
in Figure 15a,b forFrom the simulation
the Femur process,
and Tibia, we also obtained
respectively. Strain strain analysis diagrams,
is associated with theas shown
ratio in
of deformation under load to the original state. The strain is directly proportional toratio
Figure 15a,b for the Femur and Tibia, respectively. Strain is associated with the the of
deformation under load to the original state. The strain is directly proportional to the ap-
applied stress. We can analyze the strain condition of our femur and tibia from the load
plied stress. We can analyze the strain condition of our femur and tibia from the load
analysis. For the femur,
analysis. Forwe
thecanfemur,seewe
that,
canwith a deformation
see that, with a deformation scale scale
of 1838.92, thethe
of 1838.92, strain
strain
ranged from 4.004 × 10 −9 to 4.358−9× 10−5 . For−5the tibia, with a deformation scale of 27.71,
ranged from 4.004 × 10 to 4.358 × 10 . For the tibia, with a deformation scale of 27.71, the
the strain ranged from 3.457 × 10 −10 × 10×−10 3 .−3The components are highlighted
strain ranged from 3.457 ×up toup
10−10 3.257
to 3.257 . The components are highlighted with
with multiple multiple
colors, indicates the areas
colors, indicates where
the areas wherestrain
strainisismore
more significant. Note
significant. Note thatthat the
the defor-
deformation scale mationwasscale was adjusted
adjusted for clear
for clear visualvisual analysis.
analysis.

Figure
Figure 15. (a) Strain 15. (a) Strain
analysis analysis
diagram fordiagram
femur; for
andfemur; and (b)
(b) strain strain analysis
analysis diagram diagram for tibia.
for tibia.

4.2. Unified Robot Description Format Verification

The URDF refers to a robot as a tree of links connected by joints. The joints define
how one link moves concerning another, thereby defining the location of the links in
space. The links represent the moving components of the robot. URDF contains every
Robotics 2023, 12, 28 16 of 23

4.2. Unified Robot Description Format Verification

The URDF refers to a robot as a tree of links connected by joints. The joints define how
one link moves concerning another, thereby defining the location of the links in space. The
links represent the moving components of the robot. URDF contains every link’s inertia,
mass, and rotation axis information. To create a scientific simulation of kinematics after
finishing the assembly of the robot, we created a URDF file for the robot, which contains
the joint parameter information such as translation and rotation along the x, y, and z axes,
as well as roll, pitch, and yaw movements. Table 5 shows the urdf verification parameters.
Robotics 2023, 12, x FOR PEER REVIEW 17 of 24
Table 5. Verification at various angles with URDF.

Robot Hips Joint Angle Femur Joint Angle Tibia Joint

Positions
For this issue, after creating the URDF
(Degree)
file with meshes, we verified
(Degree)
the design using a
Angle (Degree)
Web-based URDF visualization tool. Scenario 1 shows the ideal state of our quadruped
Coxa FR 0 Femur FR 0 Tibia FR 0
robot. We created the URDF file in SolidWorks, where we set the positions of all joint
Ideal State Coxa FL angles and links,
0 as shown Femur FL 16a. All of0the links were
in Figure Tibia FL zero position,0 relative
set to
Figure 16a Coxa BR to the base link
0 (black body). ForBR
Femur each leg, the coxa
0 was the parent link, the femur0 was the
Tibia BR
Coxa BL child link relative
0 to the coxa, and
Femur BL the tibia was 0the child linkTibia BL to the Femur.
relative 0 Figure
16b shows the standing state of our quadruped robot, where all the links and joint angles
Coxa FR 0 Femur FR −45 Tibia FR 90
were manipulated within their respective parameter ranges; in particular, the coxa re-
Standing State Coxa FL mains at zero 0 position, while
FemurtheFLfemur link is45set to 45° relative
Tibia to
FL the coxa and90the tibia
Figure 16b Coxa BR link is set to 090° relative to the Femur.
Femur BR −45 idea of testing
The main Tibiathe
BR URDF is to verify
90 the
Coxa BL angles of the0 coxa associated with
Femur BL the other links
45 (i.e., the femur
Tibia BL and tibia). Figure
90 17
shows the URDF tree.

Figure 16. (a) Ideal state position; and (b) standing state position.

In the URDF file, we can utilize visualization and material information to create the
simulation effectively. If the URDF file has a problem, the simulation results will be poor. For
this issue, after creating the URDF file with meshes, we verified the design using a Web-based
URDF visualization tool. Scenario 1 shows the ideal state of our quadruped robot. We created
the URDF file in SolidWorks, where we set the positions of all joint angles and links, as shown
in Figure 16a. All of the links were set to zero position, relative to the base link (black body).
For each leg, the coxa was the parent link, the femur was the child link relative to the coxa,
and the tibia was the child link relative to the Femur. Figure 16b shows the standing state
of our quadruped robot, where all the links and joint angles were manipulated within their
respective parameter ranges; in particular, the coxa remains at zero position, while the femur
link is set to 45◦ relative to the coxa and the tibia link is set to 90◦ relative to the Femur. The
main idea of testing the URDF is to verify the angles of the coxa associated with the other
links (i.e., the femur and tibia). Figure 17 shows the URDF tree.

Figure 17. Unified robot description format tree. This diagram shows every joint detail with x, y, z
and r, p, y values. Each leg has four joints and four links; three joints are revolute, and one is fixed.
Robotics 2023, 12, 28 17 of 23
Figure 16. (a) Ideal state position; and (b) standing state position.

Robotics 2023, 12, x FOR PEER REVIEW 18 of 24

Figure 17.
Figure Unified robot
17. Unified robot description
description format
format tree.
tree. This
This diagram
diagramshows
showsevery
everyjoint
jointdetail
detailwith
withx,x,y,
y, zz
and r, p, y values. Each leg has four joints and four links; three joints are revolute, and one is fixed.
and r, p, y values. Each leg has four joints and four links; three joints are revolute, and one is fixed.
4.3. Dynamic Simulation Results
4.3. Dynamic Simulation Results
We created a quadruped robot using a novel mobility strategy for the leg called the
We created a quadruped robot using a novel mobility strategy for the leg called the
Watt six-bar linkage mechanism. The Pybullet physics engine platform was used for test-
Watt six-bar linkage mechanism. The Pybullet physics engine platform was used for testing
ing of the developed robot, and the robot’s dynamic model was established through sim-
of the developed robot, and the robot’s dynamic model was established through simulation.
ulation. The robot was divided into multiple parts, according to its components, and each
The robot was divided into multiple parts, according to its components, and each piece was
piece was handled as a separate entity. The maximum torque and speed of each joint, as
handled as a separate entity. The maximum torque and speed of each joint, as well as the
well as the mass, inertia, and collision geometry of each component, were precisely en-
mass, inertia, and collision geometry of each component, were precisely entered into the
tered into the simulation environment. In Figure 18, we show our quadruped robot from
simulation environment. In Figure 18, we show our quadruped robot from various angles.
various angles. We also provide detailed explanations of the mechanism and the kine-
We also provide detailed explanations of the mechanism and the kinematic equations with
matic equations with pertinent figures below.
pertinent figures below.

Figure
Figure 18.
18. Computational
Computational simulation
simulation results.
results.

Before the
Before the simulation
simulation process,
process, we
we connected
connected gait
gait patterns
patterns (e.g.,
(e.g.,trot
trot and
and creep
creep gait)
gait)
in our simulation, in order to observe the walking motion of the quadruped
in our simulation, in order to observe the walking motion of the quadruped robot. In Fig- robot. In
Figure
ure 18, 18, Scenario
Scenario 1 shows
1 shows thethe natural
natural standing
standing positionofofthe
position thequadruped
quadrupedrobot.robot. Scenarios
Scenarios
2–6 demonstrate translation of the body of our quadruped robot along its x, y, and zzaxes.
2–6 demonstrate translation of the body of our quadruped robot along its x, y, and axes.
We can see the pitch of the body along its y-axis in Scenarios 7 and 8, while
We can see the pitch of the body along its y-axis in Scenarios 7 and 8, while Scenarios Scenarios 9 and
9
10 show the roll of the body along its x-axis. Finally, we can see the yaw
and 10 show the roll of the body along its x-axis. Finally, we can see the yaw movement movement of our
robot
of ourin Scenarios
robot 11 and 11
in Scenarios 12.and
Throughout the simulation,
12. Throughout we captured
the simulation, several alternative
we captured several al-
orientations of the robot. Specifically, we demonstrated our quadruped
ternative orientations of the robot. Specifically, we demonstrated our quadruped robot’s ideal state,
robot’s
roll, pitch, and yaw positions. Following that, we built a real robot that was coupled with a
ideal state, roll, pitch, and yaw positions. Following that, we built a real robot that was
simulation engine. We put our quadruped robot through specific tests, in order to evaluate
coupled with a simulation engine. We put our quadruped robot through specific tests, in
its pitch position.
order to evaluate its pitch position.
Having matched the hardware configuration to the software simulation, we could
adjust the pitch position of the actual robot to influence the model in the simulation using
the Pybullet Physics Engine. Figure 19a displays the ideal functioning of our simulation
model. The sensor’s roll and pitch angle data were successfully acquired during our hard-
Robotics 2023, 12, x FOR PEER REVIEW 19 of 24
Robotics 2023, 12, 28 18 of 23

successfully
Havingacquired
matchedduring our hardware
the hardware orientation
configuration test.software
to the Figure 20b shows a graph
simulation, de-
we could
picting the roll–pitch angle over time, as measured by our sensors. The roll orientation
adjust the pitch position of the actual robot to influence the model in the simulation using of
our simulation model is displayed in Figure 20c. The related roll–pitch vs.
the Pybullet Physics Engine. Figure 19a displays the ideal functioning of our simulationtime graph for
roll position
model. The is shown roll
sensor’s in Figure 20d. Figure
and pitch 20e shows
angle data how the simulation
were successfully acquiredmodel
duringstabi-
our
lized when we positioned the gyroscope sensor to induce roll. The core coordinate
hardware orientation test. We constructed a roll–pitch vs. time graph (Figure 19b), based system
was
on themaintained
informationin its initial location,
gathered while the
by our sensors. rest
Our of the model
simulation counteracted
model’s the roll by
pitch orientation is
shifting the coxa angle. In this way, the simulated model can maintain a level stance
depicted in Figure 19c. We set the robot’s pitch orientation on the hardware test-bench. The rela-
tive
pitchtoposition
the reference plane.
roll–pitch vs. time graph is depicted in Figure 19d.

(a) (b)

(c) (d)

(e)
Figure
Figure 19.
19. (a)
(a)Ideal
Idealstate
stateposition
positionofofrobot inin
robot side view;
side (b)(b)
view; Ideal state
Ideal graph;
state (c) Pitch
graph; 9 of9robot;
(c) Pitch (d)
of robot;
Roll–Pitch vs. time
(d) Roll–Pitch graph;
vs. time andand
graph; (e) Self-Balanced position.
(e) Self-Balanced position.

With hardware and software synchronization, we not only could test how the hardware
setup manipulates the simulation model, but also vice versa. We used Proportional Integral
Derivative with our kinematics equations. While configuring our robot into a pitch position,
the simulated model moved such that its center of mass took precedence over the initial
pitch. Figure 19e demonstrates how the simulated model adjusted its femur and tibia to
counteract the pitch position. The central body always lies flat against the reference plane in
the virtual world. Viewed from the front, our simulation model in its ideal state is depicted
in Figure 20a. The sensor’s roll and pitch angle data were successfully acquired during our
hardware orientation test. Figure 20b shows a graph depicting the roll–pitch angle over
time, as measured by
(a)our sensors. The roll orientation of our simulation
(b) model is displayed
in Figure 20c. The related roll–pitch vs. time graph for roll position is shown in Figure 20d.
Robotics 2023, 12, 28 19 of 23

Figure 20e shows how the simulation model stabilized when we positioned the gyroscope
(e)
sensor to induce roll. The core coordinate system was maintained in its initial location,
Figure
while 19.
the(a)
restIdeal state
of the position
model of robot in side
counteracted the view;
roll by(b)shifting
Ideal state
thegraph; (c) Pitch
coxa angle. In9this
of robot;
way, (d)
the
Roll–Pitch vs. time graph; and (e) Self-Balanced position.
simulated model can maintain a level stance relative to the reference plane.

Robotics 2023, 12, x FOR PEER REVIEW 20 of 24

(a) (b)

(c) (d)

(e)
Figure
Figure 20.
20.(a)
(a)Ideal
Idealstate
stateposition
positionofofthe
therobot
robotin
infront
frontview;
view;(b)
(b)Ideal
Idealstate
stategraph;
graph;(c)
(c)Roll
Rollorientations
orientations
of
of the
the robot;
robot; (d)
(d) Roll–Pitch
Roll–Pitch vs.
vs. time
time graph;
graph; and
and (e)
(e) Self-Balanced
Self-Balanced position.
position.

We
We constructed a real robot
robot using
using 3D-printed
3D-printed components
components and and the
the six-bar
six-bar connection
connection
system.
system. When we assessed
assessed the
therobot’s
robot’strot
trotgait
gaitpattern,
pattern,itsitswalking
walking was
was essentially
essentially flaw-
flawless.
less.
FigureFigure 21 clearly
21 clearly depictsdepicts the motion
the motion of gait
of the trot the pattern.
trot gaitAccording
pattern. According to the test,
to the flat terrain flat
terrain test, we simultaneously
we simultaneously recorded therecorded the roll
roll vs. pitch vs. pitch
sensor data sensor data During
as a graph. as a graph.
the During
trotting
the trotting
period, period,
the robot the robot
recorded recorded roll
a minimum a minimum −5◦angle
angle of roll and aof −5° and aroll
maximum maximum
angle of roll16◦ .
angle of 16°. The
The minimum minimum
pitch pitch
angle was ◦
−angle was −11°,
11 , while while the
the highest pitchhighest
anglepitch ◦
was 9angle
. Thewas
angle9°. data
The
angle data are in
are displayed displayed in The
Figure 22. Figure 22. The results
gyroscope gyroscope results
of the of the
standing were −
standing
robot 1.2◦ were
robot pitch
and 0.5 ◦ roll. Some photographs of our robot are shown in Figure 23.
−1.2° pitch and 0.5° roll. Some photographs of our robot are shown in Figure 23.
 system. When we assessed the robot’s trot gait pattern, its walking was essentially flaw-
less. Figure 21 clearly depicts the motion of the trot gait pattern. According to the flat
terrain test, we simultaneously recorded the roll vs. pitch sensor data as a graph. During
the trotting period, the robot recorded a minimum roll angle of −5° and a maximum roll
angle of 16°. The minimum pitch angle was −11°, while the highest pitch angle was 9°. The
Robotics 2023, 12, 28
angle data are displayed in Figure 22. The gyroscope results of the standing robot20were
of 23

−1.2° pitch and 0.5° roll. Some photographs of our robot are shown in Figure 23.

Robotics 2023, 12, x FOR PEER REVIEW 21 of 24

Figure
Robotics 2023, 12,Figure 21. Still
21.
x FOR PEER pictures
REVIEW
Still picturesofofadopted watt
adopted six-bar
watt linkage
six-bar mechanism-based
linkage quadruped
mechanism-based robot.robot.
quadruped Sequence
Se- 21 of
1–8 shows
quence 1–8the trot the
shows gaittrot
transition on flat terrain.
gait transition on flat terrain.

Figure 22. Roll

Figure Figure
Roll pitch
pitch 22.extracted
data
data Roll pitchfrom
extracted data running
from extracted fromgait,
running trot
trot running
gait, trot gait,
according
according according
to Figure
to Figure to Figure
21 (real
21 (real robot21flat
robot (real robot f
flat
terrain
terrain test).
test). terrain test).

Figure 23. Still pictures of the

Figure 23. Stillimplemented
pictures of theversion of Ironversion
implemented Dog mini.
of Iron Dog mini.
Figure 23. Still pictures of the implemented version of Iron Dog mini.
Using the PybulletUsingphysics engine
the Pybullet simulation,
physics enginewe achievedwe
simulation, dynamic
achieved simulation re-
dynamic simulation r
sults.Using
Several translational
the sults.
Pybullet and
physics
Several rotational movements
engine simulation,
translational and rotational were demonstrated.
wemovements
achieved dynamic Kinematics (in
simulation Kinematics
were demonstrated. re- (
particular,
sults. Forward
Several and Inverse)
particular,
translational Forward aided usmovements
in achieving
and Inverse)
and rotational aided us this motion. this
in achieving
were demonstrated.Forward
motion. kinematics
Forward
Kinematics (in kinemati
provide theForward
particular, positions
provide ofthe
and the end effectors,
positions
Inverse) of the
aided while
usend inversewhile
effectors,
in achieving kinematics
this inverseprovide
motion. kinematics
Forward the locations
provide
kinematics the locatio
of the joints.
provide Therefore, x, y,Therefore,
of the joints.
the positions of the andeffectors,
end z translation,
and zas
x, y, while well askinematics
translation,
inverse roll,
as pitch,
well as and
roll,yaw
provide themovement
pitch, and yaw moveme
locations
canthe
of bejoints.
easilyTherefore,
achieved using
can be easily these
x, y,achieved
and two mathematical
using theseastwo
z translation, well approaches.
mathematical
as roll, pitch, We have,
yawthus,
approaches.
and brieflythus, brief
We have,
movement
described the kinematics
described portion
the of the
kinematics mathematical
portion of theanalysis. During
mathematical
can be easily achieved using these two mathematical approaches. We have, thus, briefly the simulation
analysis. During time,
the simulatio
described the kinematics portion of the mathematical analysis. During the simulation a 14-poi
robot walking time,
was robot
achieved walking
throughwas achieved
the use of through
a trot the
gait use
patternof a
andtrot
a gait pattern
14-point and
Bezier
curve.robot
time, According Bezier
walking curve.
to the
was According
mathematical
achieved to the the
analysis,
through mathematical
gait
usepattern, analysis,
of a trot and gait pattern,
gaittrajectory
pattern aand
generation,
and trajectory
the
14-point gener
outcome of the tion, the outcome
simulation was of the simulation
satisfactory. Details was
of satisfactory.
the practical Details of the practical
implementation and implement
a
Bezier curve. According to the mathematical analysis, gait pattern, and trajectory genera-
simulation videotion and
can beasimulation
simulation
found video [35].
elsewhere can be found elsewhere [35].
tion, the outcome of the was satisfactory. Details of the practical implementa-
tion and a simulation video can be found elsewhere [35].
5. Conclusions and Future Work

5. Conclusions and This study on the low-cost development and simulation of a quadruped robot, co
Future Work
sidering a six-bar linkage mechanism, is expected to play an important role in the vigoro
This studydevelopment
on the low-cost development
of quadruped and
robots in simulation of a quadruped
the future. Within the scope ofrobot, con-
this investigation, w
sidering a six-bar linkage mechanism, is expected to play an important role in the vigorous
presented a process for computational simulation and its underlying mathematical r
development oftionale.
quadruped robots in the
We developed an future. Within
innovative the scope
concept, of thisthorough
providing investigation, we
explanations of t
Robotics 2023, 12, 28 21 of 23

5. Conclusions and Future Work

This study on the low-cost development and simulation of a quadruped robot, consid-
ering a six-bar linkage mechanism, is expected to play an important role in the vigorous
development of quadruped robots in the future. Within the scope of this investigation,
we presented a process for computational simulation and its underlying mathematical
rationale. We developed an innovative concept, providing thorough explanations of the
steps followed to generate the simulation model. A cutting-edge Watt six-bar linkage
system was added to redistribute the weight better. Overall, we have developed a relatively
affordable quadruped robot, allowing researchers to more easily pursue careers in this
field. Our four-legged robot costs only USD 350, which is significantly less than other
quadruped robots such as the Spot (USD 74,500), Unitree a1 (USD 15,000), and Xiaomi
Cyberdog (USD 1599). Compared with existing robots, the proposed robot is lightweight,
can go through rough terrain, and can carry loads. We also used relatively cheap servo
motors and 3D-printed components.
In comparison to more traditional methods, including the belt-pulley approach utilized
in previous quadruped robots, we opted to use the Watt six-bar linkage mechanism, due
to its intuitive design. We provided in-depth kinematic analysis data. We may consider
training the robot’s gait using this model as a template. With any luck, our upcoming study
will demonstrate the real-time software and hardware synchronization system, including
a reinforcement learning technique. One of the goals of this research is to use AI in a
simulation environment, then take the results and use them to fine-tune a physical robot.
In Table 6 we exhibited the cost breakdown table.

Table 6. Cost Breakdown table.

Component Per Unit Cost (USD) Quantity Total Cost (USD)

SPT5535LV Actuator-Servo 15 12 Pieces 180
Metal Servo Arm 0.50 4 Pieces 2
Metal Servo Horn 0.40 4 Pieces 1.6
Hex spacer 0.10 30 Pieces 3
M3 Screws 0.05 85 Pieces 4.3
Brass Inserts 0.02 85 pieces 1.7
16Awg Silicon wire 1.5 (per yard) 5 yards 7.5
3D printing Abs material 21.99 (per kg) 2 kg 43.98
Raspberry Pi 4b 40 1 Pieces 40
Bluetooth Joystick Controller 16 1 Pieces 16
Arduino Pro Mega 8 1 Pieces 8
41 A Buck Converter 5 1 Pieces 5
7.4 V 2200 mAh Battery 18 1 Pieces 12
IMU 6050 0.6 1 Pieces 0.6
Total Cost 332.18$

Throughout this research, we designed and built our quadruped robot and developed
a kinematic method for its simulation in the Pybullet Physics Engine. We used 3D printing
technology to test the model in the actual world when the simulation was complete. With
the help of our simulated model, we could correct most issues, including improper scaling
and changes that would have harmed the robot. The use of a simulation model makes it
straightforward to examine the kinematic model and create a gait. Using the simulation,
we can refine our robot model to keep up with the real robot. Our long-term goal is to
Robotics 2023, 12, 28 22 of 23

advance robotics by creating a low-cost, highly developed quadruped robot that the next
generation of scientists and engineers can easily use.

Author Contributions: Methodology, M.H.R.; Formal analysis, S.B.A., T.D.M., M.F.U. and M.H.;
Investigation, M.H.R. and T.D.M.; Writing—original draft, M.H.R.; Visualization, M.H.R.; Supervision,
S.B.A.; Project administration, M.H.R.; Funding acquisition, S.B.A. and M.H. All authors have read
and agreed to the published version of the manuscript.
Funding: This research received no external funding.
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: Data is available upon request from the authors.
Acknowledgments: We would like to express our appreciation and gratitude to Fayed Al Monir’s
help with the mathematics and (FEM), (FoS) testing of our research work.
Conflicts of Interest: The authors declare no conflict of interest.

Abbreviations
The Following abbreviations are used in this research article.
URDF Unified Robot Description Format
DoF Degrees Of Freedom
FEM Finite Element Method
FoS Factor of Safety
MIT Massachusetts Institute of Technology
CAD Computer Aided Design
STL Stereolithography
PID Proportional Integral Derivative
IMU Inertial Measurement Unit
FR Front Right
FL Front Left
BR Back Right
BL Back Left
g gram
Cm Centimeter

References
1. Yao, L.; Hao, Y.; Lu, Z. Design and driving model for the quadruped robot: An elucidating draft. Adv. Mech. Eng. 2021,
13, 16878140211009035. [CrossRef]
2. Meng, X.; Wang, S.; Cao, Z.; Zhang, L. A review of quadruped robots and environment perception. In Proceedings of the 2016
35th Chinese Control Conference (CCC), Chengdu, China, 27–29 July 2016; IEEE: New York, NY, USA, 2016.
3. Biswal, P.; Mohanty, P.K. Mohanty. Development of quadruped walking robots: A review. Ain Shams Eng. J. 2021, 12, 2017–2031.
[CrossRef]
4. Todd, D.J. Walking Machines: An Introduction to Legged Robots; Springer Science & Business Media: Berlin/Heidelberg, Germany, 2013.
5. Rong, X.; Li, Y.; Ruan, J.; Li, B. Design, and simulation for a hydraulic actuated quadruped robot. J. Mech. Sci. Technol. 2012,
26, 1171–1177. [CrossRef]
6. Yang, K.; Zhou, L.; Rong, X.; Li, Y. Onboard hydraulic system controller design for quadruped robot driven by gasoline engine.
Mechatronics 2018, 52, 36–48. [CrossRef]
7. Atique, M.U.; Sarker, R.I.; Ahad, A.R. Development of an 8DOF quadruped robot and implementation of Inverse Kinematics
using Denavit-Hartenberg convention. Heliyon 2018, 4, e01053. [CrossRef]
8. Gor, M.M.; Pathak, P.M.; Samantaray, A.K.; Alam, K.; Kumar, P.; Anand, D.; Vijay, P.; Sarkar, R.; Yang, J.-M.; Kwak, S.W.
Development of a compliant legged quadruped robot. Sādhanā 2018, 43, 102. [CrossRef]
9. WildCat-The World’s Fastest Quadruped Robot. Available online: https://www.bostondynamics.com/wildcat (accessed on
14 March 2019).
10. Ananthanarayanan, A.; Azadi, M.; Kim, S. Towards a bio-inspired leg design for high-speed running. Bioinspir. Biomim. 2012,
7, 046005. [CrossRef]
11. Spröwitz, A.; Tuleu, A.; Vespignani, M.; Ajallooeian, M.; Badri, E.; Ijspeert, A.J. Towards Dynamic Trot Gait Locomotion-Design,
Control, and Experiments with Cheetah-cub, a Compliant Quadruped Robot. Int. J. Robot. Res. 2013, 32, 932–950. [CrossRef]
Robotics 2023, 12, 28 23 of 23

12. Rosendo, A.; Liu, X.; Nakatsu, S.; Shimizu, M. A combined CPG-stretch reflex study on a musculoskeletal pneumatic quadruped.
In Biomimetic and Biohybrid Systems, Proceedings of the Third International Conference, Living Machines 2014, Milan, Italy, 30 July–1
August 2014; Springer International Publishing: Cham, Switzerland, 2014.
13. Lu, M.; Jing, B.; Duan, H.; Gao, G. Design of a Small Quadruped Robot with Parallel Legs. Complexity 2022, 2022, 9663746.
[CrossRef]
14. Pinto, V.; Soares, I.; Rocha, M.; Lima, J.; Gonçalves, J.; Costa, P. Design, modeling, and control of an autonomous legged-wheeled
hybrid robotic vehicle with non-rigid joints. Appl. Sci. 2021, 11, 6116. [CrossRef]
15. Rahman, M.H.; Islam, M.M.; Al Monir, M.F.; Alam, S.B.; Rahman, M.M.; Shidujaman, M.; Islam, R. Kinematics analysis of a
quadruped robot: Simulation and Evaluation. In Proceedings of the 2022 2nd International Conference on Image Processing and
Robotics (ICIPRob), Colombo, Sri Lanka, 12 March 2022; IEEE: New York, NY, USA; pp. 1–6.
16. Biswal, P.; Prases, K. Mohanty. Kinematic and Dynamic Modeling of a Quadruped Robot. Machines, Mechanism and Robotics; Springer:
Singapore, 2022; pp. 369–378.
17. Morlando, V.; Teimoorzadeh, A.; Ruggiero, F. Whole-body control with disturbance rejection through a momentum-based
observer for quadruped robots. Mech. Mach. Theory 2021, 164, 104412. [CrossRef]
18. Yi, S. Reliable gait planning and control for miniaturized quadruped robot pet. Mechatronics 2010, 20, 485–495. [CrossRef]
19. Ho, T.; Lee, S. Piezoelectrically actuated biomimetic self-contained quadruped bounding robot. J. Bionic Eng. 2009, 6, 29–36.
[CrossRef]
20. Sun, Z.; Zhu, Z.; Zhang, G.; Li, Y.; Rong, X. Design and Control of a Quadruped Robot with Changeable Configuration. In
Proceedings of the International Conference on Intelligent Robotics and Applications, Harbin, China, 1–3 August 2022; Springer:
Cham, Switzerland, 2021.
21. Shi, Y.; Li, S.; Guo, M.; Yang, Y.; Xia, D.; Luo, X. Structural design, simulation and experiment of quadruped robot. Appl. Sci. 2021,
11, 10705. [CrossRef]
22. Cong, Q.; Shi, X.; Wang, J.; Xiong, Y.; Su, B.; Xu, W.; Liu, H.; Zhou, K.; Jiang, L.; Tian, W. Stability Study and Simulation of
Quadruped Robots with Variable Parameters. Appl. Bionics Biomech. 2022, 2022, 9968042. [CrossRef]
23. Rodinò, S.; Matteo Curcio, E.; di Bella, A.; Persampieri, M.; Funaro, M.; Carbone, G. Design, simulation, and preliminary
validation of a four-legged robot. Machines 2020, 8, 82. [CrossRef]
24. Takei, Y.; Tazawa, R.; Kaimai, T.; Morishita, K.; Saito, K. Dynamic simulation of non-programmed gait generation of quadruped
robot. Artif. Life Robot. 2022, 27, 480–486. [CrossRef]
25. Bledt, G.; Powell, M.J.; Katz, B.; Di Carlo, J.; Wensing, P.M.; Kim, S. MIT Cheetah 3: Design and control of a robust, dynamic
quadruped robot. In Proceedings of the 2018 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS),
Madrid, Spain, 1–5 October 2018; IEEE: New York, NY, USA, 2018.
26. Raibert, M.; Blankespoor, K.; Nelson, G.; Playter, R. BigDog, the rough-terrain quadruped robot. IFAC Proc. Vol. 2008,
41, 10822–10825. [CrossRef]
27. Zong, M.; Yang, Y.; Fu, M.; Wang, S.; Deng, Z. Dynamic modeling and kinematic analysis of Frog robot. In Proceedings of the
32nd Chinese Control Conference, Xi’an, China, 26–28 July 2013; IEEE: New York, NY, USA; pp. 5848–5853.
28. Raibert, M. Alphadog, the rough-terrain robot. In Adaptive Mobile Robotics; World Scientific: Singapore, 2012; p. 7.
29. Semini, C.; Tsagarakis, N.G.; Guglielmino, E.; Focchi, M.; Cannella, F.; Caldwell, D.G. Design of HyQ-a hydraulically and
electrically actuated quadruped robot. Proc. Inst. Mech. Eng. Part I J. Syst. Control Eng. 2011, 225, 831–849. [CrossRef]
30. Chen, X.; Gao, F.; Qi, C.; Zhao, X. Spring parameters design to increase the loading capability of a hydraulic quadruped robot. In
Proceedings of the 2013 International Conference on Advanced Mechatronic Systems, Luoyang, China, 25–27 September 2013;
IEEE: New York, NY, USA, 2013.
31. Hutter, M.; Gehring, C.; Jud, D.; Lauber, A.; Bellicoso, C.D.; Tsounis, V.; Hwangbo, J.; Bodie, K.; Fankhauser, P.; Bloesch, M.; et al.
Anymal-a highly mobile and dynamic quadrupedal robot. In Proceedings of the 2016 IEEE/RSJ International Conference on
Intelligent Robots and Systems (IROS), Daejeon, Republic of Korea, 9–14 October 2016; IEEE: New York, NY, USA, 2016.
32. Ackerman, E. Boston Dynamics’ SpotMini Is all Electric, Agile, and has a Capable Face-Arm; IEEE Spectrum: New York, NY, USA, 2016.
33. Tirumala, S.; Sagi, A.; Paigwar, K.; Joglekar, A.; Bhatnagar, S.; Ghosal, A.; Bharadwaj, A.; Kolathaya, S. Gait library synthesis for
quadruped robots via augmented random search. arXiv 2019, arXiv:1912.12907.
34. Oruç, H.; George, M. Phillips. q-Bernstein polynomials and Bézier curves. J. Comput. Appl. Math. 2003, 151, 1–12. [CrossRef]
35. Rahman, M.H. Practical Implementation and Simulation Video of Iron Dog Mini. Available online: https://github.com/
irondogmini (accessed on 7 January 2023).

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.