DESIGN AND IMPLEMENTATION OF STEWART PLATFORM

ABSTRACT

A Stewart platform is a type of parallel manipulator that has six prismatic

actuators, commonly hydraulic jacks or electric linear actuators, attached in pairs
to three positions on the platform's baseplate, crossing over to three mounting
points on a top plate. All 12 connections are made via universal joints. Devices
placed on the top plate can be moved in the six degrees of freedom in which it is
possible for a freely-suspended body to move: three linear movements x, y, z
(lateral, longitudinal, and vertical), and the three rotations. The platform was then
built using servo motors, magnetic aluminium channels, and acrylic plates. Control
of the Stewart Platform is achieved using an Arduino
 1 INTRODUCTION

Stewart Platform is one of the most popular parallel manipulators. It is

a six degrees-of-freedom positioning system that consists of a top plate, a
bottom plate, and six extensible legs connecting the top plate to the bottom
plate. A general Stewart Platform is shown in Figure 1.1.

Figure 1.1. General Stewart Platform

The Stewart Platform was originally proposed and presented to

academia in 1965 as a flight simulator by Stewart [1]. This structure
consisted of three linear actuators in parallel. Gough [2] had earlier
suggested a structure similar to Stewart's model as a tire- testing machine. In
his system there are six actuators in parallel resulting in a fully parallel
actuated mechanism as shown in Figure 1.3. Gough was the first one
realizing benefits of this kind of a manipulating structure; however,
research on this subject began with Stewart's paper. Therefore, it is a
tradition to call the structure as Stewart Platform, or sometimes it is also
referred as Stewart-Gough Platform [3].

Figure 1.3. The first octahedral hexapod

or the original Gough platform

From the time it was proposed, there had been no interest to this
mechanism for about 15 years until Hunt stated the advantages of using
parallel manipulators. After 1983 researchers acknowledged its high load
capacity and precise positioning capabilities, and started to make detailed
analysis of the structure. Today's popular parallel manipulator known as
Stewart Platform reached its generalized form which consists of six linear
actuators that are connected to the base with spherical or universal joints
and to the moving platform with spherical joints [4]. From 1980's onward,
Stewart Platform, gained popularity mainly because of the advantages
offered by parallel manipulators over the serial ones [3].
 Figure 1.4. The first flight simulator based on an octahedral hexapod in
mid 1960s (courtesy of Klaus Cappel)

Parallel manipulators have been used since 1980's when high load carrying and
precise positioning capability is needed. Since the time it was proposed, it has
been used as a flight simulator. Besides, a wide variety of applications have
benefited from this design. A few of the industries using Stewart Platform
design include aerospace and defense, automotive, transportation, machine tool
technology, and recently the platform is used in medical applications for its
precise positioning capability.
OBJECTIVE

The Stewart Platform is a widely accepted design as a motion control

device. This is largely because of the system's high load capacity and
accurate positioning capability. Stewart Platform provides a large amount
of rigidity that enables the system to provide a significant source of
positional certainty.
This work aims to develop a Stewart Platform mechanism with the
following capabilities:
▪ Six degrees of freedom

▪ Working in any orientation

▪ No moving cables, ease of setup and reduced friction

▪ Load capacity of 200 gr.

▪ Repeatability of ±0.5 µm

▪ Significantly smaller and stiffer package

Stewart Platform

The Stewart Platform is a common type of parallel

manipulator and possesses six degrees of freedom. In their
paper, Cruz, Ferreira, and Sequeira [5] state that “the
generic Stewart-Gough platform is composed of two rigid
bodies connected through a number of prismatic actuators
as in a parallel arrangement of kinematic chains. Usually
six actuators are used, pairing arbitrary points in the two
bodies.”

The actuators are connected to the rigid bodies by spherical

or universal joints. This would mean the Stewart Platform is
a spherical-prismatic-spherical, or SPS, robot. Although the
joint positions can be arbitrary, evenly spacing them results
in special cases. When the joints are spaced 60° around the
base and the moving platform, this is called a 6-6
configuration. When a pair of actuators share the same joint
spaced 120° on the moving platform, this is called a 6-3
configuration. 120° spacing on top and bottom platforms
results in a 3-3 configuration as shown in Figure 1.2.
(a) Type 6-6 (b) Type 6-3 (c) Type 3-3

Figure 1.2: Stewart Platform Configurations

 Geometry Stability

In a literature survey, a true type 6-6 Stewart Platform was

never found to be used in practice. Typically, platforms
were designed as a mix between the 6-6 and 3-3
configuration. Instead of pairs of actuators sharing the same
joint, each actuator had its own joint like in a 6-6
configuration; pairs of joints were separated by a small gap,
resulting in a geometry visually similar to Type 3-3. A
diagram of this geometry is shown in Figure 1.3 and
example platforms are shown in Section 1.3.4.

Figure 1.3: Typical Stewart Platform Joint

Configuration
Further analysis reveals that a true 6-6 platform with 60°
spacing between all joints is unstable because there is no
force balance, allowing the top platform to rotate about its
axis. This is illustrated in Figure 1.4. When a small
disturbance occurs, the links cannot exert forces in opposite
directions resulting in a couple about the links. Spherical
joints cannot resist this moment, leading to a collapse. This
shortcoming of a true 6-6 configuration shows the value of
6-3, 3-3, and comparable geometries.
Figure 1.4: Unstable 6-6 Configuration

History

The early stages of robotics were dominated by serial

robots, which drew heavy inspiration from the human arm
and its utility. It was the contributions by Stewart and
Gough that brought parallel robots under the scope of
further research and development, especially in industry in
recent years. According to Merlet [15], Eric Gough
established a closed-loop kinematic mechanism and its
principles in 1947 and built a prototype for tire testing in
1955. In the 1960s, the rise of aeronautics created a demand
for flight simulators. D. Stewart published “A Platform
with Six Degrees of Freedom” in 1965, discussing a
triangular parallel mechanism design approach for the
simulators. While Stewart had important contributions to
flight simulator development, his published design was not
used in practice. On the other hand the tire-testing
mechanism designed by Gough, who was one of the
reviewers of Stewart’s paper, found heavy usage in industry.
Gough’s Tire Testing Machine is shown in Figure 1.5 and
used six jacks to exert combined loads to test Dunlop tires.
Regardless,
the platform has become known as the Stewart-Gough
Platform, or more informally as the Stewart Platform.

Figure 1.5: Tire Testing Machine Application of Stewart

Platform [13]
 Kinematic Equations

The Stewart Platform kinematics are derived from analysis of

its six kinematic chains. The following derivation is drawn
from Dr. Saeed Niku’s inverse kinematic derivation of the
Stewart Platform, found in the 3rd edition of his book,
Introduction to Robotics: Analysis, Control, Applications
[17].

Figure 1.6: Home Position of Stewart Platform Robot

A fixed reference frame xyz is placed at the center of the

platform base. The z -axis is normal to the base and the x -
axis is parallel to it, pointed towards a spherical joint. A
moving reference frame noa is similarly placed at the center
of the moving platform with the a-axis normal to the moving
platform face and the n-axis pointed towards a spherical
joint. The spherical joints are assumed to be connected
through the linear actuator. Frame noa rotates by θ,φ,ψ
respectively.

Four parts compose each kinematic chain C: A connects the

origin of the fixed reference frame and base spherical joint
at an angle from x. B connects the origin of the moving
reference frame and moving platform spherical joint at an
angle from
n. L connects the spherical joints. P connects the origins of
the two frames, and is common for all six chains. The
purpose of the inverse kinematics is to determine the length
L of each linear actuator, given the position P and
orientation of frame noa. The vector equation to determine
each actuator length is
 Li = P + Bi − Ai (1.1)

Applications

There are several applications suited for a Stewart Platform.

Flight simulators make use of the platforms 6 DOF to
simulate orientation and motion. The Stewart Platform can
accurately orient to the correct roll, pitch and yaw, and their
high-dynamic capability can create linear and nonlinear
forces, like a bump. Shake tables make use of this capability
to create high-acceleration, low-displacement moves.
Fiber-optic equipment can be aligned using a Stewart
Platform because they must be accurately positioned and
oriented, enabling the light to refract and the optic signal to
operate
properly. In astronomy, Stewart Platforms are used in
spherical radio telescopes and trajectory tracking, such as
the one shown in Figure 1.7.

Figure 1.7: AMiBA Radio Telescope in Hawaii [12]

Stewart Platform Specifications

The application of this Stewart Platform is for education and

experimentation to pro- mote the ”Learn By Doing”
experience in regards to parallel robots. Thus, there are not
external deterministic requirements or specifications that
the Stewart Platform must meet. Instead, an internal
specification will be set. Quantitative specifications are
found in Table 1.1.
Table 1.1: Stewart Platform Specifications

Metric Specificatio Actual

n Value
Mechanical Properties
Weight < 40 lb 23.7 lb
Lengt < 24 in 21
h < 24 in in
Width < 36 in 21
Heigh in
t 31
in
Range of Motion
Θ > 30° 44°
φ > 30° 54°
ψ > 60° 90°
X 6 in 13.5 in
6 in 14.5 in
Y 6 in 25.0 in

Z
Accurac
y
Angular < 1°
Positional 2° 0.25 in
0.25 in

19
The mechanical properties were set to allow the robot to
fit comfortably on a desk surface, and be moved by one
person. The specified range of motion demonstrates that
the Stewart Platform can properly utilize each of its six
degrees of freedom. The robot is not being used for
tooling or processing applications, and does not need to
be extremely accurate; it must be accurate enough to
pass visual observation by the user.

There are also some qualitative specifications set for the

Stewart Platform. The platform should be safe for
students to operate and experiment with. It should be
designed to be easy to manufacture using Cal Poly
Machine Shop resources. Chosen vendor and custom
parts used should be readily available or easy to replace,
keeping the Stewart Platform serviceable. It would be
educationally beneficial to reconfigure the robot.
Typically, the geometry of a robot is optimized to solve
a problem or complete a task. For this lab purpose,
exploring multiple configurations and their benefits and
limitatio

CHAPTER 2

LITERATURE SURVEY

Forward Kinematics

20
 The most challenging problem concerning the Stewart problem has
been the forward kinematics analysis since late 1980's. The problem is to
determine the assembly configurations when base points, platform
geometry and link lengths are given. In other words, it is the problem of
solving the following kinematics relation
|| t + R* p – b ||2 = L 2

21
6i =

22
1 to 6

23
(2.1)
i i i

where bi

24
denotes the

25
ith

26
base point,

27
pi the

28
ith

29
platform point, R the rotation, t the

translation matrices for the given link lengths,

30
Li . Analytical solution for the general

case is quite difficult due to highly nonlinear equations with multiple

solutions. Therefore, researchers constructed Stewart Platforms with
special geometries to simplify the problem.

Analytical Solutions for Special Cases

Merging of base connection points and/or platform connection

points can lead to more simplified kinematics equations for a forward
kinematics analysis. The platforms with special geometries are
characterized by their m base points and n platform points. A 6-6
Stewart Platform is the general case, and 3-3 one is the simplest case.
Although 3-3 is the simplest case, the approach which is based on
solution of the input-output relations for the spherical joints makes the
problem more difficult [7-9]. Another approach uses the idea that if the
joint centers of the pairs of the adjacent limbs are coincident, the
hexagonal structure of the platform will be reduced to a triangle and the
platform can be put into a form which is isomorphic with those of
triple arm

31
mechanism [10-16]. A third approach converts the platform to an
equivalent serial mechanism, and the constraints on the joint angles are
utilized to derive the remaining equations for the forward kinematics [17-
20]. These approaches on specialized platforms enabled closed form
solutions for the simple 3-3 case and for more complicated 5-5 and 6-4
cases. Beside the coincident connection points, some researchers [21-23]
proposed the use of angular constraints between six pairs of points, lines,
and/or planes in the base and the moving platform to obtain the solution
of the problem in closed form.
Apart from the above-mentioned approaches, a decomposition
scheme was proposed by Nair and Maddocks [24] for the forward
kinematics problem. A linear dependent part and a nonlinear
independent part were suggested. Faugere and Lazard
[25] classified all the m - n cases according to the combinations of
connection points,

and found out the existence of 35 different classes.

Numerical Schemes

As discussed above, a closed form solution of the problem is very

difficult due to highly nonlinear equations with multiple solutions. It is
more advantageous to use computational techniques for practical cases
which need only a solution and if a good initial estimate is available from
a neighbor position [26-27]. Numerical approaches use some algebraic
and geometric elimination methods to simplify the kinematics equations

32
to obtain all the real solutions [28-30]. In order to find all the real roots,
Innocenti and Parenti-Castelli [31] used the analytical result of the 5-5
case by a unidimensional search over the value of the removed fictitious
leg length to obtain a numerical solution for the 5-5 case. Dasgupta and
Mruthyunjaya suggested an efficient 3 dimensional search and
verification algorithm based on pure geometric constraints [32].
Although numerical methods have computational advantages and
can be utilized to find the real solutions, they are not useful to predict
total number of the roots in the complex domain. Raghavan [33] was able
to find all distinct solutions in a complex domain by tracking 960 paths
which suggest the upper bound for the number of configurations for the
Stewart Platform to be 40.

33
Analytical Methods for General 6-6 Case

In order to reduce the total degree of the final polynomial system to

64, linearization of some equations, which include quadratic terms (that
are obtained from (2.7) by using a direction cosine matrix instead of
Euler angles) is utilized [34-38]. However, the results which were
published by Raghavan [33] contradict the validity of this approach.
Assuming the base plate as planar, the fundamental equations are
reduced to 40 degree univariate polynomial which is stated by Wen and
Liang [39].
Geometry based analytical methods have some drawbacks. They
cannot be applied to obtain all real and complex solutions, and to
determine the number of solutions available for 6-6 Stewart Platform.
Despite the fact that the closed form relation based on geometrical
considerations can provide some solutions, no analytical solution is
constructed for the general case. Hunt and Primrose [40] give
geometrical arguments to find the number of possible configurations in
special cases. Possible configurations for the general case were foreseen
to be 40, 48, 54 or 64. The idea suggested by Hunt and Primrose is
expected to provide valuable insight for general solution of the platform.
Wampler [41] and Husty [42] utilized Euler angles, and studied
parameters to obtain a mapping of spatial kinematics to eight quadratic
equations in eight dimensional image space. The possible positions of the
platform can be represented as intersections of six constraint manifolds
and the quadratic equations which ultimately lead to 40th univariate
polynomial. However, this approach cannot constitute an answer to the
number of solutions yet.

34
Other Approaches

The aforementioned methods are not suitable to obtain a real-time,

reliable and fast forward kinematics solution. A reliable and fast
approach should satisfy the question of selecting the actual one among
all the obtained results, and whether the solution is fast enough for the
real time applications. In order to satisfy both needs, Baron and Angeles
[43-45] suggested a redundant sensing method to make the resulting
procedure fast and robust to measurement noise. This method produces
estimates with about the same accuracy as a nonlinear procedure. In this
approach, the projection of the

35
motion of hip-attachment points onto their subspaces enables conversion
of the underlying direct kinematics to a linear algebraic system to resolve
the ambiguity. The linearization procedure, a polar least square estimate,
leads to a fast computation. Implementation of neural networks by Geng
and Haynes [46] constitutes yet another research effort for the forward
kinematics problem.

Dynamic Analysis

Dynamic analysis and derivation of dynamic formulation are quite

complicated due to the closed-loop structure and kinematics constraints
of the Stewart Platform manipulators. However, development and
analysis of dynamic models are the important trends in various study
fields on Stewart Platform. The importance of the dynamic model can
be illustrated in several different ways. A dynamic model can be used for
computer simulation of a robotic system without the need of a real
system to test various specified tasks. It is possible to achieve higher
performance by incorporating more structural system information for the
development of control strategies. Revealing all the joint reaction forces
and moments through the dynamic analysis is also necessary for sizing
the links, bearings, and actuators [47].
Several different methods have been studied to model the dynamics
of the Stewart Platform as a multi-body system, such as Newton–Euler
method, the Lagrange formulation, the principle of virtual work, and
Kane's method. Dasgupta and Mruthyunjaya [48] derived the complete

36
dynamic equations for the Stewart Platform through the pure Newton–
Euler approach. Nguyen and Pooran [49] and Lebret et al.
[50] developed Lagrange equations of motion, and the latter gave some
insight into the structure and properties of these equations. Lee and Geng
[51] studied the dynamics of a flexible Stewart Platform manipulator
using Lagrange formulation assuming platform to be rigid. The dynamic
formulation was established by a combination of screw theory with the
principle of virtual work by Gallardo et al [52].
Principle of virtual work formulation is based on the computation of
the energy of the whole system with the adoption of a generalized
coordinate framework [52]. To simplify this approach, theory of screws,
which is a way to express velocities and forces in three dimensional
space, combining both rotational and translational parts, was used.

37
 Wang and Gosselin [53] and Tsai [47] also used the principle of virtual
work to perform the dynamic analysis of the spatial six DOF parallel
manipulators with prismatic actuators.
Koekebakker et al. [54] provided the dynamic formulation of the
Stewart Platform through Kane's method. As Liu et al. [55] mentioned,
Kane's equation analyses the dynamics of a multibody system with N
bodies using the relation
F+F* =0

38
(2.2)

where F denotes the generalized active force, and F* is projection of

active and inertial forces on the generalized velocities. The calculation
of accelerations, partial velocities of mass centers, and partial angular
velocities of all links are required to apply the method.
Nguyen's equations, Lebret's model, and Koekebakker's formulation
all assume that origin of the coordinate system on the moving platform to
be the centre of mass of the moving platform. Furthermore, the mass
moment of inertia matrix is assumed to be diagonal. Liu et al. [55] also
derived a dynamic analysis with constraint equations based on Kane's
method. These constraints constitute a set of differential algebraic
equations which result in many numerical computational problems [56].
Additionally, Ji [57] studied the effect of leg inertia on the dynamics of
the Stewart Platform.
The application of Newton–Euler approach is straightforward.
However, this approach needs computation of all constraint forces and
moments at all joints. In addition, sometimes these computations are not
necessary for the simulation and control of the manipulator. Lagrange
formulation provides an orderly structure which can be expressed in
closed form. However, derivation becomes quite tedious due to large
amount of symbolic computations needed. Principle of virtual work is an
efficient approach for dynamic analysis of a Stewart Platform
manipulator. However, dynamic structure formulation is not explicit
[58]. In general, deriving the equations of motion for a parallel
manipulator results in a set of differential algebraic equations as
mentioned in [55]. In simulation and control, this formulation can cause
difficulties. Choosing the appropriate modeling method makes the
dependent variables be explicit functions of the integrable differential

39
equations to avoid the difficulties mentioned earlier. Accounting for the
parallel configuration of the Stewart Platform, and combining the
advantages of both Newton–Euler method and Lagrange formulation, the
explicit compact closed-form dynamic equations of Stewart Platform can
be derived in the task space [58].

40
 Flexure Joints

In order to further improve accuracy of the system, rather than using

spherical joints, interactions at joints should be improved to provide the
necessary motion in nanometer accuracy. There are several ways to
achieve this accuracy. One of them is to prevent the backlash and friction
in the joints. Flexure hinges is the best way to prevent backlash and
friction due to their characteristics. Today, flexure hinges are used in
many areas such as in accelerometers, gyroscopes, translation micro-
positioning stages, motion guides, piezoelectric actuators and motors,
high-accuracy alignment devices for optical fibers, high- precision
cameras and robotic micro-displacement mechanisms.
Mechanically assembled joints such as universal or ball joints
reduce the accuracy due to manufacturing errors. The monolithic
characteristics of the flexure joints help avoid manufacturing errors. This
characteristic brings easy manufacturing process and implies a very
compact design that can be used in the micro-assembly workstation
presented in [59-60]. From operation point of view, flexure joints clearly
reduce frictional losses. Therefore, they do not require lubrication, and
inaccuracies due to lubrication would be eliminated.
Flexure joints must be designed extremely carefully due to their
very sensitive force-displacement relationship. Because of this, high
dimensional accuracy during the fabrication and calibration after
fabrication process are needed. Flexure may also be sensitive to the
working temperature [61].
Young et al. [61] presented a new design tool and analysis for
parallel kinematics manipulator with flexure joints. The main difference

41
between flexure mechanisms and conventional joints is the consideration
of kinematics stability and the design issue.
Wei et al. [62] stated that the flexure hinges have a lot of advantages
compared to the others such as ball joints or universal ones. Because of
the fact that they are manufactured monolithically, they are very
compact in structurally. Besides, they have a lot of advantages like
having no backlash, no friction, no lubrication and no error due to
lubrication. However, they have a limited range of motion as they have
to flex without sustaining any plastic deformation at the joints.

42
 Figure 2.1. Wide range flexure-hinge based parallel manipulator

Hybrid Flexure Hinges

Flexure hinges with single-axis can be divided into two main

categories: leaf and notch type hinges [63]. Due to relative low rotation
precision and stress concentration, leaf type hinge is seldomly adopted.
In 1965, Paros and Weisbord [64] introduced the first notch hinge and
circular flexure hinge. The common feature of these two types is ease of
manufacture. Therefore, researchers turned their attention to other
configurations that could provide precision rotation in an even larger
angular range.
Smith et al. [65] presented a flexure hinge of elliptic cross-section,
the geometry of which is determined by ratio of the major and minor
axes. Likewise, Lobontiu et al.
[66] introduced an analytical model for corner-filleted flexure hinges that
are incorporated into planar amplification mechanisms. Later, they also
introduced the parabolic and hyperbolic hinges configurations [67].

43
Closed-form equations are formulated to characterize their compliance
both for the active rotation and all other in and out-of-plane motions.
To sum up, Gui-Min et al. [63] represented the compliance model of
the right circular hybrid flexure hinges. The close-form solutions were
provided to characterize the flexibility and precision of rotation. The
precision model with stress considerations were verified with the finite
element analysis. Their results show that the most suitable solutions for
large displacements and high accuracy are reached with the right circular
hybrid flexure hinges (Figure 2.2) rather than the right circular ones and
corner filleted hinges.

44
 Figure 2.2. Right circular hybrid flexure hinge

Two Axis Flexure Hinges with Axially Collocated Notches

According to Lobontiu and Garcia [68], the flexure needs to be

compliant in the bending direction and rigid for all other axes and
deformations. Constructively, a flexure hinge may have several
(multiple) sensitive axes. These sensitive axes define the rotations and
motions. The flexure hinges configurations were defined in [68] with
their compliant/sensitive axis as illustrated in Figure 2.3.

45
Figure 2.3. Flexure hinges' configurations with their
compliant/sensitive axis

46
 Apart from these, Paros and Weisbord [64] presented two-axis
circular flexure hinges which are designed in serial configuration.

Figure 2.4. Two-axis circular flexure hinges designed in serial

configuration

The serial design preserves the convenience of having each flexure

hinge designed according to the standard single-axis geometry. However,
it also requires the extra-length that is necessary to locate the two
flexures in a serial manner.
Two-axis flexure hinges with symmetric and axially-collocated
notches are presented in [68]. The compliance based formulation is
solved with an emphasis on the capacity of rotation. Stress, precision of
rotation and efficiency in terms of strain energy are calculated. Later, all
the calculations are performed again for the two axis flexure hinges with
parabolic notches. The parabolic-profiled two-axis flexure is also
compared with its constant rectangular cross-section counterpart in terms
of several performance criteria. Based on the analytical model the results
show that parabolic-profiled two-axis flexures give better performance
than their rectangular cross-section counterparts.

47
Design of Symmetric Conic-Section Flexure Hinges Based on
Closed-form Compliance Equations

The most widely used formulation for flexure hinge design was
proposed by Lobontiu et al. [69]. They derived closed form equations of
compliance for conic section (circular, elliptic, parabolic and hyperbolic)
flexure hinges.

48
Figure 2.5. Conic section flexure hinges

The analysis was performed in terms of two non dimensional

parameters. This allows performance comparison of elliptic, parabolic
and hyperbolic flexure hinges relative to circular flexure hinges. The
elliptic, parabolic and hyperbolic flexure hinges are more compliant (in
stated order) than the circular ones for large length to thickness ratios.
Hyperbolic flexures perform best in terms of preserving the center of
rotation position.

Workspace and Singularity Analysis

The aspect of the parallel manipulators make Stewart Platforms

superior compared to the serial manipulators due to their rigidity.
However, mechanism's number of degrees of freedom instantaneously
changes at some configurations which are referred as singularities in
kinematics. For parallel manipulators, one or more degrees of freedom
are gained at singular configurations. This causes loss in the rigidity of
the structure, and the mechanism cannot support force or movement in
certain directions [70]. Therefore, singularities must be avoided or

49
excluded from the workspace in order to improve the performance.
There has been ongoing research to find the singularities of Stewart
Platform, and many different solutions have been proposed.
When Hunt stated the advantages of such parallel manipulators in
1978, he also mentioned a singular configuration when the moving plate
rotates about a line intersected by all six legs [71]. In 1986, Fitcher [72]
found another singular configuration which occurs when the moving
plate is kept parallel to the base, and rotated 90 degrees about z-axis. In
1988, Merlet [73] used Grassmann Geometry to find

50
possible singular configurations. He found other possible singular
configurations in addition to the ones stated by Hunt and Fichter. All of
these approaches use geometric information; however, it is hard to
provide a unified relation this way. Singular configurations can be found
case by case if a modified or new Stewart Platform is designed. A
singularity equation is needed to create a unified relation [74]. In 1993,
Sefrioui and Gosselin [75] derived an analytic expression of the
singularity loci for a planar three DOF parallel manipulator. Until 1996
there had not been a deep study on the analytic expression of singular
configurations in six DOF parallel manipulators. In 1996, St-Onge and
Gosselin [76] derived an analytic equation directly from the property of
the determinant, which is a fourth-degree polynomial of the position
variables.
In addition to the research on the singular configurations,
characteristics of the singular configurations have been studied [74]. In
1990, Gosselin and Angeles [77] classified the singularities of the closed-
loop kinematics chain mechanisms into three categories in general form.
In 1991, Ma and Angeles [78] suggested another classification of the
singular configurations, and derived conditions for the architecture
singularity.
Other studies on singularity includes the work by Kim et al. [74]
who used extra sensors on Stewart Platform to reduce the complexity of
the position kinematics problem, and to find singular configurations. In
1999, they used extra sensors to simplify velocity equation as it was
done in the forward position kinematics case, and derived the singularity
equation directly from the velocity equation. Singular configurations can
also be found by forming the Jacobian symbolically. If determinant of
the Jacobian is set equaled to zero, singular configurations can be
extracted from this equation. However, since each Jacobian element is

51
quite complicated, Su et al. [79] proposed a simple singularity analysis
for Stewart Platform using Genetic Algorithm (GA). In this method, the
square of the determinant of Jacobian matrix is selected as the object
function, and the minimal value of the objective function is found in the
trajectory workspace using GA. If the minimum of this objective
function is zero, then there are certain singularities. Otherwise, there is
no singularity position for that Stewart Platform.

52
Figure 2.6. Hunt's singular configuration (Legs' vectors cross the same line in
space)

Figure 2.7. Fitcher's singular configuration (rotation ±90˚ in z

axis)

53
CHAPTER 3

METHODOLOGY

We’re working on a project where we explore robotic body language as a way

to augment/replace purely digital communication protocols. For this, we need to
move a cellphone with as many degrees of freedom as possible. A robot arm
would have been a good option, but for aesthetic reasons, we don’t want the
mechanism to draw too much attention to itself. After a bit of research, we’ve
settled on a structured called a Stewart Platform.

The Stewart Platform is a really amazing and versatile building block of

robotics. It exists in many sizes and is used for many purposes. Lots has been
written about it. But sometimes it’s hard to find a nice intro to the basic
operation of its mechanism.

After reading a couple of papers and trying out a couple of different

implementations, this is the one that worked for us.

Add TipAsk QuestionCommentDownload

Step 1: Hardware: Modeling, Cutting, Printing and Assembling

54
2 More Images

We designed the platform using servo motors because they are easier to find
and cheaper than linear actuators. The range of motion is a bit more limited, but
we were designing for expressiveness and not so much for range.

Before cutting metal and programming motors, Radamés designed a version of

our platform using Fusion 360 in order to get some intuition about the range of
movements and motor positioning.

The materials we used were:

55
- 1x RaspberryPi (with 5V/2A power supply)
- 6x Dynamixel AX-12A motors
- 1x UART circuit (with 12V/6A power supply)
- 1x aluminum base
- 1x plastic platform
- 12x M3 ball joints
- 6x threaded rods
- 6x servo brackets
- 6x servo horns
- lots of tiny screws

The servo brackets and horns came with the motors we bought, but we also
modeled them in case we needed replacement parts. Simpler acrylic pieces can
be used to hold whatever object you need on top of the platform, but since we
wanted to hold cellphones in an upright position, we ended up designing a
custom platform. STLs are attached here. The servo horns can also be cut from
1/8” acrylic or wood using a laser cutter.

Step 2: Testing

We tested our AX-12A library and motor driver circuit using one of the motors
attached to our platform. We quickly realized that the servos on the platform
can’t be driven independently of each other; in most cases, all six motors have
to move simultaneously in order to achieve a desired position or pose.

This is a video showing how all the motors are affected when we drive only one
motor up and down.

Add TipAsk QuestionCommentDownload

56
Step 3: Math: Papers

Before we started sending signals to all the servos, we figured it would be

worthwhile to get familiar with the mathematics of the Stewart Platform.

Unlike articulated robotic arms, the Stewart Platform’s inverse kinematics are
simpler than its forward kinematics. What this means is that it’s easier to
calculate the leg lengths and motor parameters given a desired position for the
platform, than to calculate where the platform is located for a given set of motor
parameters. This is fine; we really want the inverse kinematics anyway, and that
way we avoid solving a system of 18 non-linear equations with 40 possible
solutions.

Some of the papers that we found were not specific to servo-based Stewart
Platforms; they simply described the math based on desired leg lengths, and
sometimes assumed that linear actuators would be used. This is the case for
this MICS journal paper, which describes a very specific Stewart Platform and
focuses on its forward kinematics.

57
We also found two papers that were more specific for servos. This paper by
Filip Szufnarowski describes the inverse kinematic problem very nicely, but it
was this document by an unknown author from the Wokingham U3A Math
Group that had the most detail and cleanest notation. For example, this image
that labels all the relevant points of a Stewart Platform with their corresponding
coordinate system.

Add TipAsk QuestionCommentDownload

Step 4: Math: Joint Equations

58
The inverse kinematics problem of a Stewart Platform can be broken up into
two stages:

(1) Given a desired position and orientation for the platform, how far is each
joint on the platform from its corresponding base joint,

and

(2) What servo angles, if any, put each platform joint in the positions calculated
in the previous step.

The first problem is easy to solve; once you have the appropriate points and
coordinate systems defined like in the above image, the distances between base
joints and platforms joints can be calculated with simple matrix operations for
rotation and translation.

Namely, the length of each leg is calculated using the li equation in the image
above. T is the translation vector between base coordinate system and platform
coordinate system (where you want the platform to move), bi and pi are the
locations of the joints in base and platform coordinate systems, respectively,

59
and PRB is a rotation matrix describing how you want the platform to rotate.
PRB is also described in detail in one of the images above.

Add TipAsk QuestionCommentDownload

Step 5: Math: Length Equations

60
The second part of the inverse kinematics problem is a little bit trickier. For
each servo, given a platform joint position P, relative to the base joint position
B, and fixed lengths for the servo horn a and support leg s, what is the servo
arm angle that satisfies the distance l calculated in the previous step.

Since l increases as you vary the servo arm angle from -90° to +90° (relative to
base plane), one way to solve for this angle is to do a binary search over the
angle values, and find the one that more closely satisfies all the distance
constraints. This is done on the code for this Stewart Platform.

But, the Wokingham U3A Math Group document actually steps through the
derivation of a closed-form expression for this angle, using some pretty sweet
geometry, algebra and trigonometry tricks.

61
The form of the equation for the angle is shown in one of the images above
(read the paper for the full derivation and definition of the variables).

We actually found a small bug in this part of the U3A document: instead of
using the p values for the platform joint positions in the platform coordinate
system, you have to use the q values, which are relative to the base coordinate
system.

Add TipAsk QuestionCommentDownload

Step 6: Math: Testing

We wrote a simple Stewart Platform simulator for our platform, to see the range
of movements that it will be able to achieve, and to double-check the math. The
code is in Processing, and is on github.

tep 7: Controlling the Stewart Platform

62
Once the math was verified, it wasn’t hard to extend the code to get real-time
communication going between the simulator and the platform. We used OSC to
send motor angle packets to the RaspberryPi controlling the platform, and a bit
of code to transform these angles into AX12-A motor commands. The code for
controlling the motors using the simulator is on github.

CHAPTER 4

BLOCK DIAGRAM

SWITCH

ARDIUNO SERVO MOTOR MOTION

CONTROLLER CONTROL

63
64
WORKING PRINCIPLE

In this project we are using to servo motor for motion control system. The
project is a productivity increasing feature in a factory. During the
manufacturing of this mechanism we had found many of intelligence that can be
given to it. We provide the basic functions like three motion. And the main
function, up down and rotation.

65
ADVANTAGES

1. Low cost automation project

2. Less power consumption

3. Operating principle is very easy

4. Installation is very easy

5. Simple in construction

66
CONCLUSION

Thus the STEWART PLATFORM project was successfully completed. In this

technology further the use of environmental technologies while delivering
comforts and driving pleasure. It is a very good substitute and replacement for a
conventional alert system.

67
REFERENCES

1. Chung M., Malone E., Tolley M. T., Chepaitis A. J., Lipson H., (2008)
“Object Augmentation for the Visually Impaired Using RP”, Proceedings of the
19th Solid Freeform Fabrication Symposium, Austin TX, Aug 2008

2. Dafoui, El-Mouloudi, (1998) “Analysis and Design of a Six-DOF Parallel

Manipulator, Modeling, Singular Configurations, and Workspace”, IEEE
Transactions on Robotics and Automation, Volume 14, No. 1

3. Grenouilleau J. C., O. Housseini, and F. Pérès., (2000) “In-Situ Rapid Spares

Manufacturing and Its Application to Human Space Missions”, Proceeding of
Space 2000, p.42-48.

4. Jongwon Kim, Chongwoo Park, Sun Joong Ryu, Jinwook Kim, Jae Chul
Hwang, Changbeom Park, Iurascu, C.C., “Design and analysis of a redundantly
actuated parallel mechanism for rapid machining” , IEEE Transactions on
Robotics and Automation, August 2001, Volume 17, Issue 4, p. 423-434.

5. Kang, Ji-Yoon, (1996) “Robust Tracking Control of Stewart Platform”,

Proceedings of the 35th Conference on Decision and Control, Kobe, Japan,
December 1996

6. Lazard, Daniel, “The (true) Stewart Platform Has 12 Configurations”, IEEE

1994

7. Lipson, H., (2006) "A Relaxation Method for Simulating the Kinematics of
Compound Nonlinear Mechanisms", ASME Journal of Mechanical Design,
Volume 128, Issue 4

8. Malone E., Lipson H., (2007) “Fab@Home: The Personal Desktop Fabricator
Kit”, Rapid Prototyping Journal, Vol. 13, No. 4, pp.245-255

68
9. Masuda, T., “Specific Kinematic Changes in a Linear-Actuated Parallel
Mechanism According to Differences in Actuator Arrangement”, Proceedings
of the 2000 IEEE/RSJ Internal Conference on Intelligent Robots and Systems

10. Merlet, J. –P, “Solving the Forward Kinematics of a Gough-Type Parallel

Manipulator with Interval Analysis”, The International Journal of Robotics
Research 2004, Volume 23, No. 3

11. Petko, M., Karpiel, G., “Mechatronic Design of a Parallel Manipulator for
Milling”, 2005 Proceedings IEEE/ASME International Conference on
Advanced Intelligent Mechatronics., p.759-764

12. Weiss, L. and Prinz, F., 1998, "Novel Applications and Implementations of
Shape Deposition Manufacturing,” Naval Research Reviews 1, Office of Naval
Research

69