ABSTRACT

A bipedal walking robot is a type of humanoid robot that mimics human

behaviour and may be programmed to do certain jobs. A prototype robot is
developed in this study to serve as a test bed for the physical locomotion that is
utilised to regulate robot movements such as moving forward, backward,
turning left and right, getting up from front and back, and rolling over from left
and right. The bipedal robot is also described in the study, as well as how the
movement steps are produced and how it is detected when it falls down. A
remote controller can also be used to control the robot's movement. In a
dangerous situation, this bipedal robot can aid humans with chores or activities.
This could eliminate the possibility of human injury or death.
1. INTRODUCTION
From automated industrial lines to simple everyday use of electronic gadgets,
the mobile robot has become a passion for individuals who are robot enthusiasts
in the twenty-first century. Hobbyists have created robots for games such as
soccer robot, sumo robot, and others. Robots have also shown to be quite useful
in the workplace, and they were created with the intention of assisting humans.
Hexapods, Quadrupeds, Wheeled robots, and other sorts of robots exist. To
emulate human movements, a bipedal walking robot with human traits was
chosen. Bipedal robots will be far more efficient in a human setting than any
other sort of robot yet conceived.
It is capable of completing tasks that would be too difficult or risky for a
human to perform. Such hazardous applications include fire rescue operations,
toxic gases or chemicals, explosives such as land mines, and assisting humans
in difficult tasks that they are unable to complete.
Humanoid robot research is gaining traction in recent years, with a particular
focus on applications in built-for-human environments and human-robot
interaction. Because of their anthropomorphism, human-friendly design, and
locomotion applicability, humanoid robots are the sort of robot that can
practically cohabit with humans in a built-for-human environment.
Humanoid robots differ from other types of robots in that their physical
structures are designed to closely resemble those of humans. Humanoid robots
are intended to cohabit and interact with people in areas where humans work
and live since their shape has many basic physical traits with humans. They
could also be used to replace humans in dangerous situations or at catastrophe
sites.
These requirements make it necessary for humanoid robots to perform a variety
of complex motions, such as walking, ascending stairs, avoiding objects,
creeping, and so on. Humanoid robots must be able to perceive and perform
human-like motion in order to cooperate with people. Robot researchers have
been working hard over the last decade to create anthropomorphic humanoid
robots that can think intelligently and replicate human behaviour.
Sophia is a social humanoid robot developed by Hong Kong-based
Company Hanson Robotics. Sophia was activated on February 14, 2016 and
made her first public appearance at South by Southwest Festival (SXSW) in
mid-March 2016 in Austin, Texas, United States.
Sophia has been covered by media around the globe and has participated in
many high-profile interviews. In October 2017, Sophia "became" a Saudi
Arabian citizen, the first robot to receive citizenship of any country. In
November 2017, Sophia was named the United Nations Development
Programme's first ever Innovation Champion, and is the first non-human to be
given any United Nation title.
Sophia was first activated on February 14, 2016. The robot, modelled after the
ancient Egyptian Queen Nefertiti, Audrey Hepburn, and her inventor's wife,
Amanda Hanson, is known for human-like appearance and behavior compared
to previous robotic variants.
As of 2018, Sophia's architecture includes scripting software, a chat system,
and Open cog, an AI system designed for general reasoning. Sophia imitates
human gestures and facial expressions and is able to answer certain questions
and to make simple conversations on predefined topics (e.g. on the
weather). Sophia uses speech recognition technology from Alphabet Inc. (parent
company of Google) and is "designed to get smarter over time".
Her speech synthesis ability is provided by Cereproc's Text-to-Speech engine
and also allows her to sing. Sophia's intelligence software is designed by
Hanson Robotics. The AI program analyses conversations and extracts data that
allows it to improve responses in the future.
Hanson designed Sophia to be a suitable companion for the elderly at nursing
homes, or to help crowd’s at large events or parks. He has said that he hopes
that the robot can ultimately interact with other humans sufficiently to
gain social skills. Sophia is marketed as a "social robot" that can mimic social
behaviour and induce feelings of love in humans.
Sophia has at least nine robot humanoid "siblings" who were also created by
Hanson Robotics. Fellow Hanson robots are Alice, Albert Einstein
Hubo, BINA48, Han, Jules, Professor Einstein, Philip K. Dick Android, Zeno,
and Joey Chaos. Around 2019–20, Hanson released "Little Sophia" as a
companion that could teach children how to code, including support for Python,
Blockly, and Raspberry Pi.

Fig1.Sophia Humanoid Robot

Fig 2. Biped Robots

1.1. Disturbance of the Terrain
Current robotics research is beginning to expand applications to include
mobility in unstructured situations with unpredictable terrain. This is a difficult
expansion since a robot that moves precisely in a controlled environment may
fall after its first experience with a terrain disturbance.
Even though various studies have been conducted to improve robustness to
ground variation, bipedal robots travelling on tough terrains still face substantial
limitations. Bipedal robots could only manage unplanned impediments that
were less than 6% of leg length until the early 2010s.
When compared to normal impediments in ordinary life, such as the height of
steps in a building or the curb height of a sidewalk on a city street, this figure
seems absurdly low. A priori knowledge of ground height fluctuations more
than a few millimetres is also required.
These robots are limited to a walking pace of about 1.0 m/s on average. Uneven
terrains have traditionally been travelled at slow walking rates to preserve
stability and, if relevant, to assure perception accuracy. Without any knowledge
of the ground, the fastest walking pace rarely exceeds 1 m/s.
Since its inception in 2004 [18], HONDA has been producing HRP-series
robots, with the HRP2 becoming a successful platform in the study field of
humanoid robotics. To overcome the hard terrain walking problem, Takubo et al
proposed a Zero Moment Point (ZMP) based criteria.
HRP-2 was used to validate the newly proposed approach. Takubo et al. used
the terms "step up" and "step down" to represent the variance of normal flat
terrain in their investigation. The walking pattern's landing time was changed to
correspond to the "step up" or "step down" state.
The studies confirmed the efficacy of the proposed methods, despite the fact
that the “step” heights were less than 10 cm. To increase the practicability of
humanoid robots, HRP-2 was replaced by HRP-4C and HRP-4. Similar to HRP-
2, the processor HRP-4(C) model is completely actuated, with degrees of
freedom of robots corresponding to the actuation system.
By simplifying HRP-4C as a single linear inverted pendulum with ZMP delay,
the actuation system successfully applied posture/force control to stable HRP-
4C. HRP-4C walked well on the uneven pavement with modest alterations to the
set trajectory. The pavement had a maximum slope of roughly 3 degrees. HRP-
4C could walk across the pavement with a present flat ground walking pattern
and a walking speed of 0.2 m/s without knowing the outdoor ground profile.
Fig3.Honda HRP series robot

Fig4.Honda Asimo robots

Fig5. Development of Asimo
Caron et al made a breakthrough in humanoid robot dynamic stair climbing in
2019. Two new components were added to the stability controller for climbing
stairs: quadratic programming-based wrench distribution and a whole-body
admittance controller. Untethered stair climbing had never been done before in
order to prevent robot falls and secure their safety.
1.2. Specifications for Bipeds
This bipedal robot has six Tower Pro SG-5010 servo motors (Operating Speed:
0.20sec/60 (4.8V), 0.16sec/60 (6.0V), Stall Torque: 5.2kg*cm (4.8V), 6.5kg*cm
(6.0V), Rotational Range: 180) in each leg. The robot's structure is built of 1
mm thick aluminium sheet brackets.
To support and attach all of the components together, including the electrical
hardware, hardware fasteners such as standoffs (M3 size), screws, and nuts are
employed. Acceleration sensor (3g), two axes with sensitivity (440 mV/g) to
identify the robot when it falls down and automatically get up.
The movement of the robot allows it to go forward, backward, turn left and
right, and get up from both front and rear. A 7.4VDC Lithium Polymer (Li-Po)
battery pack powers the robot. A 9VDC alkaline battery is used to power a
remote controller that controls the robot's motions.
2. LITERATURE REVIEW
One of the most intriguing and pressing difficulties in biped humanoid robot
research is achieving great locomotion robustness. This paper gives a quick
review of work and methods on bipedal robots' robust walking. To reject terrain
disturbances and impulsive force disturbances, various robust walking strategies
have been presented thus far.
The proposed strategies improve the resilience and adaptivity of robots by a
substantial margin when applied to real robots. Bipedal robots have so far been
able to navigate unfamiliar terrains with ground variation reaching 20% of leg
length. In comparison to decades ago, the height of obstacles has increased by
more than threefold.
In the case of an unanticipated external force, bipedal robots can regain
equilibrium not only in a stationary condition, but also when walking. Biped
running, on the other hand, is underdeveloped in comparison to robust walking.
Even yet, the top running speed is less than 3.0 m/s, not to mention the system's
lack of robustness to major shocks.
Based on the groping result, a humanoid robot adjusts its orientation and
movement direction. During operation in a human context, this motion is
critical to avoid colliding with any object, particularly a wall. The correction
method consists of two parts: correction of distance and angle, as well as
adjustment of the robot's orientation and locomotion direction.
The findings of the experiments showed that the humanoid robot could
recognise wall orientation and correct locomotion direction by creating
appropriate leg trajectories. The suggested approach worked well in conjunction
with the groping locomotion algorithm to detect and avoid obstacles in the
rectification region, improving the groping locomotion algorithm's performance.
Autonomous motions of the robot's manipulators are demonstrated in the
experiments. These results meet the goal of this study, which is to design an
autonomous system that can avoid obstacles during groping locomotion.
2.1. Front wall groping
The position of the wall confronting the robot in groping front wall presents the
chance of collision during the robot's orientation adjustment. As a result,
adjusting the robot's distance was as simple as defining a trajectory for the legs
to move in a retrograde direction.
2.2. Right-side wall groping
Correction of distance comprises leg trajectory generation to walk side-step
away from the wall when groping the right-side wall. However, even if the
groping angle I is 0Id45°, the robot can still crash with the wall. In this example,
the robot will take one stride backwards before beginning to walk sideways. If
the groping angle I is 45°Id90°, the robot will eventually walk side-step away
from the wall to rectify its position.
2.3. Obstacle Avoidance Motion
The ability to perceive and avoid obstacles is unavoidably significant in the
application of humanoid robots to human environments. The obstacle avoidance
approach presented in this study involves adding an appropriate algorithm to the
humanoid robot's control system to recognise and avoid impediments that occur
within the corrective region of groping locomotion.
The suggested obstacle avoidance algorithm was tested on a bipedal humanoid
robot whose arms were equipped with six-axis force sensors that recognised the
presence of barriers physically and generated an appropriate trajectory to avoid
them.

Fig6.Obstacle avoid walking robot

2.4 Avoiding a high hazard
When avoiding a high obstacle, the height of the obstacle is roughly the same.
The obstacle-avoidance algorithm's notion is based on the groping findings and
the trajectory creation of the humanoid robot's legs.
Interpolation utilising polynomial equations determines leg positions, and each
leg-joint position is determined using angle data from the invective
computation. Leg positions are determined using polynomial equations, and
each leg-joint position is determined using angle data derived from the inverse
kinematics required to move the legs to the desired locations.
In general, obstacle avoidance is carried out after the robot's distance from the
wall has been corrected and before going to the correct angle. The left arm will
search for and detect any obstacle inside the correction angle's area and up to
the arm's maximum length while inspecting the obstruction to the left, in order
to instruct the robot's system whether to proceed with the correction or to go on
to the next obstacle avoidance step.
If an obstacle is identified, the robot will rotate to the back-left, reorienting
itself to face the obstacle. The robot will then conduct the "confirm obstacle"
operation on a regular basis to ensure that the barrier is still present. The robot
will walk forward if no obstacles are found. However, if an impediment is
detected, the robot will walk side-step to its left side and repeat the confirmation
process until no obstruction is found.
2.5. Low-obstacle avoidance
When avoiding a low obstruction, keep in mind that the height of the obstacle is
roughly equal to the distance between your foot and knee. With the assistance of
both arms, the robot conducts motion to hop over the obstacle. This motion has
nothing to do with the groping mode of locomotion.
The robot's vision sensor or sensors at the bottom of its feet may detect the
impediment.

Fig7.Low Obstacle avoiding robot

2.6. Mechanical Structure
The kinematic structure of any bipedal humanoid is its distinguishing trait.
Initially, due to restricted technology, only a basic representation of the
humanoid figure was possible, frequently including solely of the legs. This is
not a drawback because the design concept may favour simpler forms that allow
certain assumptions to be made and utilised in the control component.
The MIT 3D Biped and ATRIAS are excellent examples of this methodology.
The 3D positioning function can be performed without a human-like shape by
extending a leg and rotating the hip. The spring-loaded inverted pendulum
(SLIP) control template is unique to both robots.
The MIT 3D Biped uses linear thrusters, whereas ATRIAS uses spring plates to
connect a pair of rotary actuators to the anterior and posterior thigh members of
a 4-bar parallel linkage. Both structures accomplish a hip-concentrated mass
and low-inertia legs; nevertheless, ATRIAS wastes a large amount of energy
due to the antagonistic actuation regime resulting from the leg design.
The MIT 3D Biped uses linear thrusters, whereas ATRIAS uses spring plates to
connect a pair of rotary actuators to the anterior and posterior thigh members of
a 4-bar parallel linkage. Both structures accomplish a hip-concentrated mass
and low-inertia legs; nevertheless, ATRIAS wastes a large amount of energy
due to the antagonistic actuation regime resulting from the leg design.
Despite their striking appearance, these robots are primarily utilised as
demonstrators for the study of dynamic bipedal mobility. Robots with a full
humanoid body design have been constructed for more widespread uses. In
terms of kinematics, prominent humanoids such as ASIMO, the HRP series
HUBO, REEM-C, and TALOS are all comparable.
A three-degree-of-freedom (DoF) hip simulates a spherical joint, one DoF for
knee bending, and two DoF for an ankle ball joint make up the legs. As a result,
six actuators are sufficient to approximate the form and functionality of a
human leg. This structure also makes it possible to solve the inverse kinematics
in closed form. To prevent singularities, joint restrictions, or to achieve a
specific configuration, further joints (for example, a toe) can be added. The
robots in question use servomotors in a serial format with rotary joints.
The structure is well-known in the scientific community and has undergone
extensive research, as indicated by the large number of robots that use it. Its
strength resides in its simple mechanics and control, however this comes at the
cost of restricted functionality. Because rotary actuators are usually installed
immediately at the joint, the ones closest to the chain's origin must bear the one.
 This has two effects: increased leg inertia and blowback or elasticity
accumulation from each joint. Both contribute to positioning mistakes that must
be compensated for by the control mechanism. Reduced limb rigidity and
greater leg inertia can have a significant impact on the system's overall dynamic
performance, putting higher demands on actuators in terms of quality and power
output.
As in an earlier version, this problem can be solved by attaching the actuators
off-axis, closest to the root of the connection, and employing a lightweight
coupling or transmission (e.g. synchronous belts). By putting the actuators off-
axis, you can create novel joint designs with parallel connections. Due to the
mechanical drive coupling, parallel mechanisms have a higher stiffness, which
averages out the actuation error. Due to the matings of the links, parallel
solutions have a limited workspace as well as higher mechanical and
computational complexity.
Some systems can be used for individual joints, while others can be used for the
entire leg. The limited range of motion and lack of a knee joint of a Stewart
platform type leg may bring benefits in terms of loading capacities, but they
may not convert to human-like capabilities.
Because it is the root of the leg, a 3-DoF hip will gain the least from the
addition of a parallel structure. Even a little reduction in inertia necessitates a
large investment in mechanical development. Because of these mechanical
complexities, yawing in the hip is still commonly done using a separate rotary
actuator, whose backlash has only a minor effect on the placement of the entire
leg.
With two actuators, hip abduction/adduction, flexion and extension can be
achieved, with the sum of their movements resulting in pitching and differential
movement resulting in hip rolling. Because a wide range of motion in the hip is
desirable, only a few robots have used parallel linkages in the hip to date, such
as the RHP-2.
A parallel knee joint is relatively simple to implement since there is generally
enough space in the thigh link to install the actuator and transmission. A rotary
actuator in a crank-lever mechanism, as seen in WALK-MAN and the modern
ASIMO, or a linear actuator, which pulls and pushes the shank link, are two
possibilities.
The needed knee torque can be greatly reduced when combined with a non-
trivial linking mechanism, resulting in a larger torque-to-weight ratio and
consequently improved performance. The main benefits of employing parallel
linkages come from a 2-DoF ankle joint, as the improved stiffness complements
a statically stable upright posture without requiring significant torque
consumption.
The shank, like the thigh, has plenty of room for the actuators and couplings
that allow the ankle to pitch and roll around a Cardan joint. As a result, a
number of small and attractive solutions have been developed. This comprises
the ASIMO and CogIMon crank-lever mechanisms, as well as the linear
actuator system.
Aside from joint-specific solutions, only a few researchers approached the
kinematic structure in a novel way, created a hybrid serial-parallel leg in which
the hip houses five of the six leg actuators. This reduces leg inertia and, as a
result, the needed joint velocities and torques, which are also more consistent.
When compared to a reference robot using a serial kinematic chain, the
structure was used to produce a Bipedal robot that achieved a 15% reduction in
the distance from origin to CoM (scaled to leg length). The control theory has
been simplified by making assumptions such as walking solely on flat ground.
In a similar way, this can be used in the mechanical part.
The NimbRo-OP2(X) robots use a five-degree-of-freedom serial-parallel leg
structure with two 4-bar connections in the sagittal plane and a serial chain in
the lateral direction. The 4-bar pantographs keep the foot horizontal to the waist
and constrain leg orientation. When there is no tilting, the feet are basically
parallel to the ground.
The actuation technique is fundamentally different, with the angle of the thigh
and shank being adjusted instead of the hip and knee joints. Actuators can be
synchronised and inserted at any pantograph axis to improve performance. A
leg with this construction can also be totally operated at the hip, minimising
inertia, and augmented by a spring to reduce torque requirements.
Biomimetic is a distinct design approach that is attracting a growing number of
scholars. Because humanoid robots are meant to imitate people, it's only
reasonable to use not just the concept of a humanoid body plan, but also its
implementation.
Biarticular actuation in a musculoskeletal structure minimises control
bandwidth requirements and improves efficiency by allowing coordination and
energy transmission between two joints. The use of tendons and flexible parts
enhances limb compliance and shock-loading.
Tensegrity, or the co-location of tendons with the skeleton, decreases stress on
structural sections and allows for the use of lighter elements, resulting in greater
dynamic performance. With several tendons passing through various joints, it's
vital to reduce friction or account for it.
Humanoids that are completely biomimetic are still rare. Kenshiro and his
descendant Kengoro are highly complex humanoids with a human-like bodily
structure, including anatomically correct muscle arrangement and skeleton
structure. The potential of such robots has yet to be explored because to the
great complexity in the control and synchronisation of numerous "muscles."
2.7. ACTUATION
Designing a humanoid robot necessitates a holistic approach, in which the
robot's structure is dependent on the actuator used and vice versa. Electrical
actuators with high-ratio reducers have been the standard since the mid-1980s,
as they provide a good balance of torque, speed, and size.
Aside from performance, using electrical actuators simplifies control due to the
nearly linear input-to-output relationship and the availability of power storage
and distribution systems. Because of the high transmission ratio required to
deliver the needed torque, a high-gain control regime is required.
As humanoid robots are expected to interact with their surroundings more and
more, force-feedback control becomes more important. This doesn't work well
with the high-reduction setup's inflexible actuation, which interferes with the
current/torque connection and isn't shock-resistant.
This problem is solved by incorporating an elastic element into the actuator's
structure, resulting in a series elastic actuator (SEA), similar to the one
employed in NASA's Valkyrie. The spring element has various functions: it
absorbs impacts, stores energy, and uses Hooke's law to adjust the output
torque.
WALK-MAN is outfitted with the most cutting-edge SEA technology,
including active and passive adaptability. Existing humanoids can be made
more compliant by adding an elastic element to the actuator's output shaft.
Brushless DC (BLDC) motor manufacture and control have advanced to the
point where a new method is possible. Katz  effectively integrates a single-stage
1:6 planetary reduction implanted inside the stator bore with the high torque
density of a ‘pancake' BLDC motor.
Because the commutation is done electronically utilising forward and inverse
Park and Clarke transforms, the output torque may be sensed and controlled
directly via the Q-axis current. These Quasi-Direct Drive (QDD) modules, when
combined with an embedded impedance controller, enable robust, compliant,
and proprioceptive actuation with high peak power for dynamic locomotion.
These actuators were designed for the quadruped MIT Mini Cheetah, although
they can also be used on bipeds. Changes to these actuators (e.g., a longer
stator, a different reduction) with better specifications are on the market, thus
compliant and dynamically capable humanoid robots are only a matter of time.
Electrical motors are normally employed for rotary motion; however, a
translational (e.g. ball screw) mechanism can be employed to provide linear
motion. They can also be made as direct drive linear motors, which are basically
rotary motors that are ‘unrolled' on a guide way.
Magnetic saturation and thermal constraints limit electrical motors in high-
power applications in one way or another. The greater the motor, the higher the
power requirements. They're also rather fragile, as overloading them or using
them in sensitive situations (such as water or dirt) will swiftly kill them.
Hydraulics, on the other hand, are immune to these drawbacks because they
have a high power density, scale well, and can stop under huge loads without
causing damage, even in severe conditions. They're commonly made as pistons,
with linear motion controlled by a moveable cylinder that separates two
chambers separated by an incompressible liquid (typically oil).
When developing the first statically and dynamically moving bipeds, roboticists
were drawn to hydraulics because of the high-power specifications that could be
achieved. Boston Dynamics is currently in the forefront of this technology's
application in humanoid robotics. PETMAN, their first humanoid, had human-
like morphology, range of motion, strength, and walking.
ATLAS (together with ATLAS-DRC) was designed as a direct descendent to
examine navigation in unstructured terrain. Both robots were amazing at the
time, but they also demonstrated the limitations of the hydraulic system: power
storage and distribution were more difficult than with electrical systems because
the robots were tethered.
Due of the created noise, everyone in the area of the robot was obliged to wear a
hearing protector. With ATLAS-Unplugged, this problem was handled by
upgrading the hydraulic power unit to provide variable pressure set-points,
miniaturising it, and installing it on board the robot.
At the cost of a weight increase from 152 to 182 kg, this change gave ATLAS
full autonomy and substantially quieter operation. The new ATLAS robots are
the result of years of research and development in the field of mobile hydraulic
technology.
ATLAS was able to achieve a human-like size and weight (80 kg) by
miniaturising the entire structure and components (particularly the power unit).
It was also capable of beautiful and forceful human-like movement, such as
gymnastic routines and dance.
The road taken by Boston Dynamics demonstrates that the entrance point is
steep and that taking full use of the benefits given by hydraulic actuation
necessitates a synergy of interdisciplinary technological knowledge. Because the
technology is yet unavailable for commercial use, any business interested in
implementing a hydraulic system will encounter comparable challenges.
Multiple contestants used the ATLAS platform during the DARPA Robotics
Challenge (DRC), which was highly robust and did not require maintenance for
months, but repairs or upgrades had to be done by the manufacturer, which
slowed progress.
Due to a lack of access to technology combined with the unfamiliarity and
inconvenient nature of working with it (noise, leakages), only a few platforms
have been developed. Hydra and TaeMu are two of the most recent full-size
humanoids, with different actuation schemes.
Hydra uses electro-hydrostatic actuators (EHAs), each with its own pump,
instead of the conventional central pump design with servo valves used by
TaeMu to adjust pressure across the robot. Hydra's solution incorporates the
advantages of both electric and hydraulic actuation: back drivability, impact
resistance, controllability, and power density, but it still needs to be size and
weight optimised.
Oil is still channelled through flexible hoses in hydraulic systems, which lowers
joint motion and leads to leaks over time. It is feasible to incorporate oil
channels into the structural components of robots, but this must be done at the
design stage, putting further limits to the process.
Even the design and construction of the few humanoids listed required
specialised solutions, as shown. This could be due to the current condition of
market options, namely the lack of standardised and modular solutions for
hydraulically actuated robots.
2.8. Walking Pattern Generation
BIPED HUMANOID ROBOT PLATFORM, KHR-2 KHR-2 is a biped
humanoid robot developed in 2003. It has been utilized as a test robot platform
to develop a walking control algorithm for the authors’ biped humanoid robots,
KHR-3(HUBO) and Albert HUBO. The height, weight, and total number of
degrees of freedom of KHR-2 are 56 kg, 120 cm, and 41 (6 for each leg, 4 for
each arm, 7 for each hand, 1 for torso, and 6 for head), respectively. All joint
actuators are brushed DC motors with harmonic reduction gears or planetary
gears.
The authors realized a self-contained system by putting all mechanical and
electronic parts into the robot body. Hence, KHR-2 is tele-operated via a
wireless LAN (Local Area Network). Electrical circuit boards of joint motor
controllers and sensory devices were efficiently designed for minimum energy
consumption.
Aluminium was used as the body frame material. The thickness and size were
also minimized so as to reduce the weight within an allowable range. For
human-like appearance, the ratio of each body part corresponds with the human
ratio.
The degrees of freedom and dimensions are summarized. The control system
architecture of KHR-2 is a distributed control system. The main computer is
installed in the torso, and sub-controllers such as joint motor controllers and
sensory devices are distributed throughout the body.
Communication between the main computer and the sub-controllers is achieved
by using a CAN (Controller Area Network) protocol. The specifications and
descriptions of the sub-controllers including sensors are presented.
2.9. BIPED WALKING AND ONLINE CONTROL ALGORITHM
CONSIDERING FLOOR CONDITION
In this section, a suitable walking pattern design for dynamic biped walking
and an online control technique considering the floor conditions are described.
To make the robot walk, a gait trajectory is designed offline. In the biped
robotics research field, the gait trajectory, otherwise known as the walking
pattern, generates the relative position trajectories of two feet with respect to the
pelvis centre.
Even a well-designed walking pattern cannot prevent the robot from falling
down as a result of large upper body motions, vibrations of the body parts, and
an uneven floor. Therefore, an online walking control algorithm composed of
various online controllers is essential to maintain the dynamic balance in real-
time.
These online controllers can finely compensate the joints trajectories.
Realization of biped walking. Walking Pattern Generation for the design of the
walking pattern, the authors considered the following four design factors.
1) Walking cycle (2 × step time)
2) Lateral swing amplitude of the pelvis
3) Double support ratio
4) Forward landing position ratio of the pelvis

Fig8.Forward landing position ratio of the pelvis

Fig9.Walking Cycle
Fig10.Forward Landing Position of the pelvis
3. MANUFACTURING
Robotics manufacturing materials and processes are just as important as the
technology used in them. The structure's performance is influenced by the
materials used in terms of weight, rigidity, and maintenance. Metal alloys are
the preferred material for most humanoids, due to their great stiffness,
machinability, and heat dissipation qualities.
Subtractive (CNC-milling) or additive manufacturing (casting) are employed for
this, each with its own set of advantages and disadvantages. The iCub humanoid
is a prime example of a robot built using a subtractive technique, with milling
accounting for the vast majority of the 5000 parts. Milling limits the shape that
can be achieved because the cutting tool must be able to follow a pre-
programmed path.
3.1. MATERIALS REQUIRED

Here is the list of all the components and parts required to make your very own
Arduino powered Bipedal robot. All parts should be commonly available and
easy to find.

ELECTRONICS:

 Arduino Uno x 1

 Towerpro MG995 servo motor x 6

Perfboard (similar size to the Arduino)

  Male and Female header pins (about 20 of each)

 Jumper Wires (10 pieces)

 MPU6050 IMU (optional)

 Ultrasonic sensor (optional)

HARDWARE:

 Skateboard Bearing (8x19x7mm)

 M4 nuts and bolts

 3D printer filament (in case you don't own a 3D printer, there should be a

3D printer in a local workspace or the prints can be done online for quite cheap)

Fig11.Hardware Components
3.2. Design Process

The humanoid legs were designed in Autodesk's free to use Fusion 360 3d

modelling software. I began by importing the servo motors into the design and built
the legs around them. I designed brackets for the servo motor which provides a
second pivot point diametrically opposite to the servo motor's shaft.

Having dual shafts on either end of the motor gives structural stability to the design
and eliminates any skewing that may occur when the legs are made to take some
load. The links were designed to hold a bearing while the brackets used a bolt for
the shaft.

Once the links were mounted to the shafts using a nut, the bearing would provide a
smooth and robust pivot point on the opposite side of the servo motor shaft.
Another goal while designing the biped was to keep the model as compact as
possible to make maximum use of the torque provided by the servo motors.

The dimensions of the links were made to achieve a large range of motion while
minimizing the overall length. Making them too short would make the brackets
collide, reducing the range of motion, and making it too long would exert
unnecessary torque on the actuators.
Fig12.Design Process of Body Parts

3.3. ROLE OF ARDUINO

An Arduino Uno was used in this project. The Arduino was responsible to
compute the motion paths of the various gaits that were tested and instructed the
actuators to move to precise angles at precise speeds to create a smooth walking
motion. An Arduino is a great choice for developing projects because of its
versatility. It provides a bunch of IO pins and also provides interfaces such as
serial, I2C, and SPI to communicate with other microcontrollers and sensors.

The Arduino also provides a great platform for rapid prototyping and testing
and also gives developers room for improvements and expandability. In this
project, further versions will include an Inertial Measurement Unit for motion
processing such as fall detection and dynamic locomotion in uneven terrain and
a distance measuring sensor to avoid obstacles. The Arduino IDE was used for
this project. (Arduino also provides a web-based IDE)
Fig13.Arduino

3.4.3D Printed Parts

Ultrasonic sensor Ultrasonic sensor (optional)Skateboard Bearing Skateboard

Bearing (8x19x7mm)M4 nuts and bolts M4 nuts and bolts3D printer lament 3D
printer lament. The parts required for this project had to be custom designed
therefore a 3D printer was used to print them out. The prints were made at 40%
inll, 2 perimeters, 0.4mm nozzle, and a layer height of 0.1mm with PLA.

 foot servo holder x 1

 foot servo holder mirror x 1
 knee servo holder x 1
 knee servo holder mirror x 1
 foot servo holder x 1
 bearing link bearing link x 2
 servo horn link x 2
 foot link x 2
 bridge x 1
 electronics mount x 1
 electronics spacer x 8 (optional)
 space servo horn space x 12 (optional)
Fig 14.3D Printed Part
3.5. Preparing the Servo Brackets
Once all of the parts are printed you can begin by setting up the servos and the
servo brackets. First push in a bearing into the knee servo holder. The t should
be snug but I would recommend sanding the inner surface of the hole a bit
instead of forcing the bearing which may risk breaking the part. Then pass an
M4 bolt through the hole and tighten it using a nut. Next, grab the foot link and
attach a circular servo horn to it using the supplied screws. Attach the foot link
to the knee servo holder using the screws you will use to also attach the servo
motor. Make sure to align the motor so that the shaft is on the same side of the
bolt you had attached earlier. Finally secure the servo with the rest of the nuts
and bolts. Do the same with the hip servo holder and foot servo holder. With
this, you should have three servo motor and their corresponding brackets.
 Fig15.Servo brackets
3.6. Making the Link Pieces
Once the brackets are assembled, start making the links. To make the bearing
link, once again lightly sand the inner surface of the holes for the bearing then
push the bearing into the hole on both sides. Make sure to push the bearing in
till one side is us. To build the servo horn link, grab two circular servo horns
and the supplied screws. Place the horns on the
3D print and line up the holes, next screw the horn onto the 3D print by
attaching the screw from the 3D print side. I recommend using a 3D printed
servo horn spacer for these screws. Once the links are built you can begin the
assemble the leg.
Fig16.Link pieces
3.6. Assembling the Legs
Once the links and brackets are assembled, you can combine them to build the
leg of the robot. First, use the servo horn link to attach the hip servo bracket and
knee servo bracket together. Note: Note: Don't screw the horn to the servo just
yet as there is a setup stage in the following stage and it'll be an inconvenience
if the horn were screwed onto the servo motor. On the opposite side mount the
bearing link onto the protruding bolts using nuts. Lastly, attach the foot servo
bracket by inserting the protruding bolt through the bearing on the knee servo
holder. And x the servo shaft to the servo horn connected to the knee servo
holder on the other side. This may be a tricky task and I would recommend a
second pair of hands for this.
Fig17.Assembling the legs
3.7. Custom PCB and Wiring
This is an optional step. To make the wiring neater I decided to make a custom
PCB using perf board and header pins. The PCB contains ports to directly
connect the servo motor wires. In addition, I also left extra ports in case I
wanted to expand and add other sensors such as Inertial Measurement Units or
ultrasonic distance sensors. It also contains a port for the external power source
required to power the servo motors. A jumper connection is used to switch
between USB and external power for the Arduino. Mount the Arduino and PCB
to either side of the electronics mount using screws and the 3D printed spacers.
Left Hip >> pin 9 Right Hip >> pin 8 Left Knee >> pin 7 Right Knee >> pin 6
Left Foot >> pin 5 Right Foot >> pin 4
To make the PCB follow the same order as above by using the ports on the PCB
from right to left with the IMU port facing up. And use regular male to female
jumper wires to connect the PCB to the Arduino using the above pin numbers.
Make sure to also connect the ground pin and create the same ground potential
and VIN pin for when you decide to run it without USB power.
Fig18.Custom PCB Wiring
Assembling the Body
3.8. Initial Setup
Once the two legs and the electronics are assembled, combine them together to
build the robot body. Use the bridge piece to link the two legs together. Use the
same mounting holes on the hip servo holder and nuts and bolts that hold the
servo motor. Finally, connect the electronics mount to the bridge. Line up the
holes on the bridge and electronics mount and use M4 nuts and bolts to make
the joint.
Fig19.Initial Setup
3.9. Programming Part
 "Central position" of each servo motor has to be manually adjusted to
align with the legs. To achieve this remove the servo horns from each
servo and run the initial_setup.ino sketch. Once the motors have settled in
their central position reattach the horns such that the legs are perfectly
straight and the foot is perfectly parallel to the ground. If this is the case
you are in luck. If not open the constants.h file found on the adjacent tab
and modify the servo offset values (lines 1-6) till the legs are perfectly
aligned and the foot is flat.

Fig20.Programming initial setup

Programming the Arduino
Fig21.Programming of arduino
3.10.Final Outcome
The biped can take steps that vary from 10 to 2 cms long without tipping over.
The speed too can be varied while keeping the gait balanced. This biped
combined with the power of the Arduino provides a robust platform to
experiment with various other gaits and other objectives such as jumping or
balancing while kicking a ball. Try to change the motion paths of the legs to
create your own gaits and discover how various gaits affects the performance of
the robot. Sensors such as an IMU and distance sensor can be added to the
system to increase its functionalities while force sensors can be added to the
legs to experiment with dynamic locomotion on uneven surfaces.
Fig21.Final Outcome

4. Analysis
 Analysis can even be done in two similar ways as modelling is done,
one is the mathematical analysis and the other is the software analysis. In
whichever way it is done, the two basic analysis to be done on a mechanical
system is the kinematic and dynamic analysis be it reverse or direct.
Mathematical analysis need a higher order mathematics to deal with, on the
other hand software analysis can be done in an easy way in comparison with the
mathematics involved. Software: So, both the kinematic and the dynamic
analysis for the humanoid robot model is carried on in the CAD software
Pro/Engineer.
For the analysis, the foot of the humanoid robot is modelled to move on a
curvature of third order polynomial and the hand to perform salute. In the
kinematic analysis, while the foot moves on the pre-defined curved path and
saluting, the angles subtended by the other joints in the humanoid robot for
every time instance is recorded. This recorded data can be accessed in the form
of a graph or as a set of values. This data is used while writing the code in
Arduino IDE for the humanoid to perform a balanced walking gait. The data
that is recorded while performing the kinematic analysis is fed to the particular
joints in the model to perform the dynamic analysis on the model under
gravitational force.
The axial moment acting on every joint at different time instances are recorded.
This data that is accessed in the form of a graph or a set of values is utilized in
estimating the torque required at each and every joint for the humanoid robot to
traverse the pre-defined contour. Fabrication Based on the analysis, the
humanoid robot is fabricated to bring it live onto the earth. The fabrication part
is divided into mechanical aspects, electrical and electronic aspects and
computer programming.
Mechanical aspects in this the skeleton of the humanoid robot is build. Basically
the material used for fabricating the skeleton is Aluminium sheet of thickness
around 1.2 mm, because Aluminium is a metal with properties like low density,
easy machinability etc. However, Aluminium being ductile, easily deforms
under medium loads. The compactness of the humanoid robot can cope up with
this disadvantage by using an Aluminium alloy instead of pure Aluminium or a
bit more thickened sheet at the sites prone to deformation. Fabrication of the
skeletal system is done by first cutting the Aluminium alloy sheet according to
the desired dimension with the help of hacksaw.
The holes are drilled through the frames for bolts to fit in with the help of
manual hand drill. These frames are fastened with nuts and bolts with the help
of screw driver and a cutting plier. The servo motors are attached to the
mechanically fastened frames with the help of L-clamps, nuts and bolts. The
mechanical nuts and bolts with the help of screw driver and a cutting plier. The
servo motors are attached to the mechanically fastened frames with the help of
L-clamps, nuts and bolts. The mechanical framework is successfully actuated by
selecting appropriate actuators based on the torque estimated in the analysis.
5. RESULTS AND DISCUSSION

5.1. Software Analysis

 Kinematic Analysis
It can be observed for the derived and measured data that there is no much
deviation from the simulated results. The hip motors are varied for 5 degrees in
simulation as well as in coding. (Table 1) The robot falls in simulation but
practically it can go up to 8 degrees more. This can be due to the reduced
weight of the links during 3d printing where various infill techniques are used.
Thus the 3d printed model in more stable than the simulated model with
maximum error of 5%. Similar results can be seen varying thigh motors. (Table
2) The angular positions of the joints that are recorded with respect to time
while the foot traverses on the pre-defined curved path for a left ankle joint,
right ankle joint, left hip joints, right hip joints, are

Table1.Deviation between simulation & Prototype for stability for hip motors.
Table2. . Error analysis for thigh motors
6. CONCLUSION AND FUTURE SCOPE
This work described a biped humanoid robot walking control method that
takes into account an uneven and sloped floor. The online controllers in the
author's prior algorithm were constructed without taking into account the
local and global inclinations of the floor. That is, it was expected that the
floor was rather flat.
As a result, six online controllers were constructed and added to the prior
algorithm (upright pose controller, landing angular momentum controller,
landing shock absorber, landing time controller, landing position controller,
and vibration reduction controller). Simple mathematical models and
experiments were used to construct these online controllers, and then
appropriate activation times were determined.
It is required to produce a more human-like sole that is ground shape
adaptive and ground reaction force absorptive as part of future research. The
broad and sensitive skin of a human sole absorbs landing impact and adjusts
to uneven ground.
7. REFERENCES
[1] Robotics. (2014). Wikipedia, the free encyclopedia. Retrieved November
11th, 2014, from http://en.wikipedia.org/wiki/Robotics.
[2] Humanoid robot. (2014). Wikipedia, the free encyclopedia. Retrieved
November 13th, 2014, from http://en.wikipedia.org/wiki/Humanoid_robot.
[3] Rodney A. Brooks, Cynthia Breazeal, Matthew Marjanovi´c, Brian
Scassellati and Matthew M. Williamson (1999). The Cog Project: Building a
Humanoid Robot, C. Nehaniv (Ed.): Computation for Metaphors, Analogy
and Agents, LNCS 1562, 52-87.
Retrieved from http://www.ai.mit.edu/projects/cog/.
[4] Ill-Woo Park, Jung-Yup Kim, Jungho Lee and Jun-Ho Oh, (2005).
Mechanical Design of Humanoid Robot Platform KHR-3. Proceedings of 5th
IEEE-RAS International Conference on Humanoid Robots. DOI:0-7803-
9320-1/05.
[5] W. T. Miller III. Real-time neural network control of a biped walking
robot. IEEE Control Systems Magazine. 14(1): 41-48 (1994)
[6] C. L. Shih.: Ascending and descending stairs for a biped robot. IEEE
Transactions on Systems, Man and Cybernetics. 29(3): 255-268 (1999)
[7] Jacky Baltes and Thomas Br¨aunl. HuroSot Laws of the Game.
University of Manitoba, Winnipeg, Canada, May 2004.
http://www.fira.net/hurosot.
[8] Robocop Federation. Robocop humanoid league 2002. WWW,
November 2001.
http://www.robocup.org/regulations/humanoid/rule humanoid.htm.
[9] Fuminori Yamasaki, Tatsuya Matsui, Takahiro Miyashita, and Hiroaki
Kitano. Pino the humanoid: A basic architecture. In Peter Stone, Tucker
Balch, and Gerhard Kraetszchmar, editors, Robo
[10] M.Vukobratovic, B. Borovac, D. Surla, D. Stokic.: Biped locomotion.
Springer-Verlag (1990)
[11] K. Nishiwaki, S. Kagami, Y. Kuniyoshi, M. Inaba, and H. Inoue.:
Online generation of humanoid walking motion based on fast generation
method of motion pattern that follows desired ZMP. Paper presented at the
IEEE/RSJ international conference on intelligent robots and systems,
Lausanne, Switzerland, 30 Sep. 5 Oct. 2002
[12] D. Hobbelen and M. Wisse, “A disturbance rejection measure for limit
cycle walkers: The gait sensitivity norm,” IEEE Transactions on Robotics,
vol. 23, no. 6, pp. 1213 –1224, Dec. 2007.
 [13] “Swing-leg retraction for limit cycle walkers improves disturbance
rejection,” IEEE Transactions on Robotics, vol. 24, no. 2, pp. 377 –389,april
2008.
[14] M. Wisse, A. L. Schwab, R. Q. van der Linde, and F. C. T. van der
Helm, “How to keep from falling forward: Elementary swing leg action for
passive dynamic walkers,” IEEE Transactions on Robotics, vol. 21, no. 3,
pp. 393–401, June 2005
[15] R. Goebel, R.G. Sanfelice, and A.R. Teel. Hybrid dynamical systems.
Princeton University Press, 2012.
[16] A. D. Ames, K. Galloway, and J. W. Grizzle. Control Lyapunov
functions and hybrid zero dynamics. In 51st IEEE Conference on Decision
and Control, 2012
[17] Wonderfulengineering.com/history-of-humanoid-robots-in-pictures-
1868-1970
[18]J. Yamaguchi, A. Takanishi, I. Kato.: Development of a biped walking
robot compensating for threeaxis moment by trunk motion. Paper presented
at the IEEE/RSJ international conference on intelligent robots and systems,
Yokohama, Japan, 26-30 July 1993
[19] K. Nagasaka, H. Inoue, M. Inaba. : Dynamic walking pattern
generation for a humanoid robot based on optimal gradient method. Paper
presented at the IEEE international conference on systems, man, and
cybernetics, 12-15 Oct. 1999
[20] Y. Sakagami, R. Watanabe, C. Aoyama, S. Matsunaga, N. Higaki, K.
Fujimura.: The intelligent ASIMO: System overview and integration. Paper
presented at the IEEE/RSJ international conference on intelligent robots and
systems, Lausanne, Switzerland, 30 Sep.- 5 Oct. 2002