Project Number: FJL- MQP - MARM

MODULAR ROBOTIC ARM

4/26/2015

Project Team/Authors
Derek Calzada-Mariaca [email protected]
Monica Preston [email protected]
Yihao Zhou [email protected]

Advisor
Prof. Fred Looft [email protected]
Electrical and Computer Engineering
Systems Engineering

This report represents the work of WPI undergraduate students submitted to the faculty as evidence of
completion of a degree requirement. WPI routinely publishes these reports on its website without
editorial or peer review. For more information about the projects program at WPI, please see
http://www.wpi.edu/academics/ugradstudies/project-learning.html

i|Page
Table of Contents
Table of Figures ......................................................................................................................... iv
Acknowledgements .................................................................................................................... vi
Abstract..................................................................................................................................... vii
1. Introduction ......................................................................................................................... 1
1.1 Problem statement ........................................................................................................... 1
1.2 Goal Statement................................................................................................................. 1
1.3 Objectives......................................................................................................................... 1
1.5 Stakeholder Analysis ........................................................................................................ 2
1.6 System Needs and Requirements .................................................................................... 2
1.7 Summary .......................................................................................................................... 4
2. Background......................................................................................................................... 5
2.1 Introduction....................................................................................................................... 5
2.2 Gap Analysis .................................................................................................................... 5
2.3 Material Study .................................................................................................................13
2.4 Components Study ..........................................................................................................13
Motors ................................................................................................................................13
Power Transmission ..........................................................................................................18
3. Methods .............................................................................................................................21
3.1 Introduction......................................................................................................................21
3.2 Methods ..........................................................................................................................21
3.3 Summary .........................................................................................................................22
4. System Design ...................................................................................................................23
4.1 Mechanical Design ..........................................................................................................23
Explanation of mechanical design ......................................................................................23
FEA Simulation ..................................................................................................................25
Machining ..........................................................................................................................27
4.2 Electrical Design ..............................................................................................................30
Component Selection .........................................................................................................31
Schematic Design ..............................................................................................................39
ii | P a g e
 PCB Design .......................................................................................................................39
4.3 Programming Design .......................................................................................................40
Program Flow Chart ...........................................................................................................41
4.4 Summary .........................................................................................................................42
5. Results ...............................................................................................................................43
5.1 Introduction......................................................................................................................43
5.2 Results ............................................................................................................................43
Mechanical Results ............................................................................................................43
Electrical and Programming Results ..................................................................................44
Integrated System Testing Results.....................................................................................46
Thermal Analysis ...............................................................................................................48
5.3 Summary .........................................................................................................................51
6. Summary and conclusions .................................................................................................52
6.1 Introduction......................................................................................................................52
6.2 Summary of System Design and Results .........................................................................52
6.3 Overall assessment .........................................................................................................52
Summary of issues ............................................................................................................52
Recommended future work ................................................................................................52
6.4 Conclusions .....................................................................................................................53
7. Bibliography .......................................................................................................................54

iii | P a g e
Table of Figures
Figure 1: Low cost compliant robotic manipulator [1] .................................................................. 6
Figure 2: Low Cost Robotic Manipulator Joint [2] ....................................................................... 7
Figure 3: SCHUNK Powerball Lightweight ARM LWA 4P [3] ...................................................... 8
Figure 4: igus Robolink [4] ......................................................................................................... 9
Figure 5: Various joints and possible configurations of the igus Robolink [4] .............................10
Figure 6: Kinova Jaco Robotic Manipulator [5] ..........................................................................11
Figure 7: Types of Motors [6] [7] ...............................................................................................14
Figure 8: Illustration of a brushed DC motor's components [28] ................................................15
Figure 9: Illustration of a brushless DC motor's components [17] ..............................................16
Figure 10: Illustration of a Stepper motor's components [18] .....................................................16
Figure 11: Harmonic Drive Gear [12].........................................................................................18
Figure 12: Diagram showing setup of a series elastic actuator ..................................................19
Figure 13: Series elastic actuator used in a WPI project [16].....................................................20
Figure 14: The Systems Engineering Approach [29] .................................................................21
Figure 15: Design Charts ..........................................................................................................22
Figure 16: Internal gear configuration........................................................................................23
Figure 17: Picture of both types of linkage adapters ..................................................................24
Figure 18: First version of the joint design .................................................................................25
Figure 19: Deformation Result ..................................................................................................26
Figure 20: Von Mises Stresses .................................................................................................26
Figure 21: 3D printed prototype .................................................................................................27
Figure 22: Rounded edges can be seen in all the inside corners of the base ............................27
Figure 23: Screenshot of ESPRIT .............................................................................................28
Figure 24: Machined base piece next to stock aluminum block .................................................28
Figure 25: Assembled aluminum components ...........................................................................29
Figure 26: General electrical system diagram ...........................................................................30
Figure 27: Free body diagram of a 3-link planar robotic arm .....................................................31
Figure 28: Static torque calculation equation .............................................................................31
Figure 29: Torque-speed curve of a DC motor [20] ...................................................................32
Figure 30: ESC wiring diagram [21]...........................................................................................33
Figure 31: Turnigy TrackStar 25A ESC [33] ..............................................................................34
Figure 32: Encoder Electrical Connections [27] .........................................................................36
Figure 33 Diagram demonstrating how multiple joints could be wired in parallel [26] ................37
Figure 34: Functional diagram of a joint's internal electrical components ..................................38
Figure 35: Arduino pinout diagram ............................................................................................39
Figure 36: PCB diagram............................................................................................................40
Figure 37: Main Controller Flow Chart .......................................................................................41

iv | P a g e
Figure 38: Sub Controller Flow Chart ........................................................................................42
Figure 39: The mechanical and electrical components of the modular robotic joint ...................43
Figure 40: The assembled modular robotic joint ........................................................................44
Figure 41: Comparison of design specifications to the final results............................................46
Figure 42: Comparison of design specifications to the final results [con't] .................................47
Figure 43: Simplified joint for thermal analysis ..........................................................................49
Figure 44: ANSYS Workbench thermal simulation ....................................................................50
Figure 45: ANSYS Workbench temperature vs. time graph .......................................................50

v|Page
Acknowledgements
We would like to thank Fred Looft for his guidance through the system engineering
approach to this project. We would also like to extend our gratitude to Mik Tan for his help
machining, Joe St. Germain for 3D printing, William Appleyard for help machining and solving
mechanical issues and Robert Boisse for ordering and advising.

vi | P a g e
Abstract
The following paper describes the process and results undertaken to create a modular
robotic arm system. The intent of the project was to create a low cost modular robotic arms
system with features seen in more expensive systems as such a product does not exist on the
market today. By following a systems engineering approach, our team was able to develop a
modular robotic joint in an attempt to fill this market gap.

vii | P a g e
1. Introduction
1.1 Problem statement
A modular robotic arm, which is both advanced and inexpensive, is not currently
available on the market today. A consumer’s choices consist of inexpensive but flimsy toy arms
or commercial grade robotic systems with prices far above the affordability of the average
hobbyist.

1.2 Goal Statement

The goal of our project was to design and build a controllable modular joint that could be
used to assemble a robotic arm. The design and component selection for the joints would allow
a relatively inexpensive arm to be assembled while still maintaining precision and accuracy in its
movements. The objective was to fill the gap between the inexpensive but fragile toy arms and
the strong, precise, but expensive industrial grade arms. The target users of the arm are
expected to be hobbyists and students.

1.3 Objectives
In order to measure capstone project progress, certain objectives are needed. The
objectives help guide the group as well as define the project early on. For the creation of a
modular robotic arm, the following objectives were determined:

● Study robotic arm products currently on the market to determine gaps in features,
pricing, and applications in order to determine the best market design goals (e.g. gap
analysis).
● Specify Measures of Performance (MOPs), Measures of Effectiveness (MOEs) and Key
Performance Parameters (KPPs) that will drive the design and design choices.
● Design prototype joints that are inexpensive but robust using proper engineering
techniques.
● Perform appropriate trade studies to select system design components based on
minimizing risks to achieve KPPs, MOEs, etc.
● Create a fully functioning joint prototype.
● Write software to demonstrate control the position and movement of the modular joint
system and to be used for testing.
● Test the prototypes in various configurations and conditions to ensure that the design
specifications are met.

1|Page
1.5 Stakeholder Analysis
As part of a systems engineering approach, it was important to identify the stakeholders
of the project. Each stakeholder or group of stakeholders was identified by an ID, a defined role,
and their priority with regard to completing the project. The following stakeholders were
determined:

ID Stakeholder Roll Priority

D. Calzada, M. Preston, Y. Zhou - Engineering team
SH 01 Team Members 1
designing and creating a modular robotic arm
F. Looft - Technology and overall advisor, grades,
SH 02 Project Advisor 1
funding, equipment, and facilities
End users of product - Determine market and need
SH 03 Customers 2
for features
SH 04 WPI Co-owner, potential patent filing 2
Various companies supplying parts (motors, chips,
SH 05 Suppliers 2
etc.) used in the robotic joints
Manufacturers of robotic joints if product is
SH 06 Manufacturer 3
commercialized
Defines safety procedures used in design and
SH 07 OSHA 3
manufacturing of joints
Table 1: Stakeholders

1.6 System Needs and Requirements

Certain system needs were decided upon for the arm in order to further define the
capabilities of the end product. Needs were determined based on preliminary background
research along with the input of our project advisor and other stakeholders. Each need was
defined by an ID, a description, a reasoning behind the need, a link to the stakeholders who
would be most concerned with the specific need, and a priority. Further, as appropriate a
preliminary MOE/KPP was assigned to selected needs based on stakeholder input.

2|Page
 ID Need Description Reasoning Stakeholders Priority
All joints should be able Modularity allows a robot with an
SH 01, SH 02, SH
Modular to be used in arbitrary number of degrees of
N 01 03, 1
joints combination with any freedom to be constructed based
SH 04
other joint(s) upon user needs
A joint should be able Without being able to manipulate a
to lift a 1 kg load held payload, a robotic arm would be
SH 01, SH 02, SH
N 02 Payload at one meter nearly useless. Defining the payload 1
03
horizontally from the also defines what applications the
axis of rotation arm could be used for
While under no load,
This appears to be a reasonable
Rotational each rotational joint SH 01, SH 02, SH
N 03 speed which would allow tasks to be 2
velocity should be able to rotate 03
completed quickly enough
at at least 45 deg/s
A joint should be able
to rotate at a minimum
Rotational
of 45 deg/s while The robot needs to operate under SH 01, SH 02, SH
N 04 velocity 2
holding a 1 kg load one load in a timely fashion 03
under load
meter horizontally from
the axis of rotation
The joint should not
If a joint is too wide, it could
Motion interfere with the range SH 01, SH 02, SH
N 05 potentially interfere with the range 2
constraints of motion of other 03
of motion of the other joints
joints on the arm
Each fully extended
The longer each joint is, the longer
Length of joint should be SH 01, SH 02, SH
N 06 an arm will have to be to add joints 2
joint approximately 15 - 20 03
and therefore degrees of freedom
cm long
The wires on the arm
should present no
SH 01, SH 02, SH
safety hazard.
N 07 Wire Safety To meet safety standards 03, 2
Including, areas where
SH 07
fingers could be
crushed
The price of materials This is to keep the cost down for the
SH 01, SH 02, SH
N 08 Cost for each joint should be end users as well as to fit in the 1
03, SH 05, SH 06
no more than $150. market gap.
Table 2: System needs

3|Page
 ID Need Description Reasoning Stakeholders Priority
Any rotation to a
software defined Allows arm to move to predictable
SH 01, SH 02, SH
N 09 Accuracy angular position should positions through software 1
03
be within 1% of the full commands
rotation of the joint
Any three consecutive
rotations to the same
software defined Allows arm to repeat tasks in a SH 01, SH 02, SH
N 10 Precision 1
angular position should predictable manner 03
be within 1% of the full
rotation of the joint
Motors towards the base of the arm
will need to create enough torque to
Maximum Each joint should not SH 01, SH 02, SH
N 11 lift the rest of the arm. The lower the 1
weight weigh more than 0.5 kg 03
weight the lower the necessary
torque is.
A pitch joint should be The range of motion defines the
able to rotate at least workspace as well as tasks that an
Range of SH 01, SH 02, SH
N 12 120 degrees. A roll joint arm may be able to perform with 2
motion 03
will be able to rotate at larger ranges of motion able to have
least 180 degrees larger workspaces
All joints need to run off of the same
The joints should
voltage which in turn should be fairly SH 01, SH 02, SH
N 13 Power supply operate from a 12 VDC 2
a common voltage that sensors, 03
power source
motors and microcontrollers can use
Each joint should
contain a
microcontroller and
Since it is a modular system, having
On board sensors needed to SH 01, SH 02, SH
N 14 each joint be at least semi-intelligent 2
electronics accept positional input 03, SH 05
allows for easier setup and use
and move to the
desired position within
the defined accuracy
Table 3: System needs [con’t]

1.7 Summary
Once all of the design needs were outlined, it was time to research current systems and
available components and systems. This would allow us to draw inspiration from existing
products as well as more accurately define the market gap in which our modular robotic was
attempting to fill.

4|Page
2. Background
2.1 Introduction
This section is a compilation of background research performed for this project.
Background research includes a Gap Analysis of products already on the market, a study of
possible materials to be used in manufacturing, and a study of various types of components
(motors, power transmission, etc.) that were considered for use.

2.2 Gap Analysis

Before designing the modular robotic manipulator, the team first conducted a gap
analysis evaluating existing modular robotic manipulators. The following is a detailed analysis of
various products that have been used academically and commercially.

Low Cost Compliant Robotic Manipulator

The robotic manipulator shown in Figure 1 was designed and manufactured by Morgan
Quigley, Alan Asbeck, and Andrew Ng with the Department of Computer Science at Stanford
University [1]. The manipulator has seven degrees of freedom with a human-scale workspace.
The maximum payload of the manipulator is 2 kg.
The key features for this manipulator are its zero backlash performance as well as being
easily back drivable, which is one of the important features for human safety. In order for the
arm to have the zero backlash, a stepper motor together with a timing belt and a cable driver
are used. In order to allow the robotic manipulator to safely interact with humans, it uses a
series elastic design and reduces the flying mass to less than 4 kg by keeping the motors close
to the base.
The robot used a stepper motor because it provides a large torque at low speed.
Stepper motors also act as an electromagnetic clutch; therefore in the event of power loss, it
can remain stationary while conventional motor may continue rotating. A disadvantage of using
a stepper motor is if the motor slips, the arm position is unknown. Another reason of using
stepper motor is a high performance DC brushed motors cost much more than the stepper
motor with the same performance level. The robot manipulator uses joint encoders to achieve
closed-loop PID control.
The force sensing was accomplished by measuring the displacement of the series
elastic component, which will be discussed later. The overall structure was made of plywood in
order to minimize weight and a laser cutting machine was used to manufacture all the
components. The group used OROCOS-KDL kinematics library in C++ to control the motion of
the arm and integrated the software with ROS to graphically simulate the arm on a computer.
The obvious disadvantage of this arm is that using series elastic actuators with cables
makes the arm difficult to build. To assemble such arm requires expert skill, which is not our
intended customer base. The structure material was plywood and all the motors are exposed
without protection this make the arm brittle and a potential safety risk to users.

5|Page
 Figure 1: Low cost compliant robotic manipulator [1]

Low Cost Robotic Manipulator Joint

A similar project to the one that we are tasked with was worked on in 2009 by capstone
project students [2]. Their goal was to design and build a modular robotic manipulator joints that
were low cost and highly accurate. There were two different types of joints, an elbow joint and a
rotator joint. In the design phase, they had three iterations for both joints. In their design, each
joint is able to receive data and electrical power, generate motion, and transfer data to the next
joint. As shown in Figure 2, one of the designs for the elbow joint used a worm gear drive
system, which prevents backlash while providing a high static torque. The disadvantage of a
worm gear is if the robot arm were to experience high external force, it may damage the gears.
This version was never implemented or tested by the 2009 team.

6|Page
 Figure 2: Low Cost Robotic Manipulator Joint [2]

SCHUNK Powerball Lightweight ARM LWA 4P

The industrial robotic arm joint show in Figure 3 was design by SCHUNK Inc. A
multinational manufacturing company based in Germany [3]. The Powerball controls the rotation
of two axes, oriented 90 degrees from each other. This allows the single joint to have 2 degrees
of freedom and to perform both pitch and rotational motion. The combination of three
Powerballs results in 6 degrees of freedom allowing the arm to reach any point in a workspace.
The Powerball has universal CANopen communication interfaces and cable technology
that was designed by SCHUNK for data transfer, power supply, and as a way to easily integrate
with other modules. The joint module uses brushless DC motors with permanent magnet
brakes. Compared to other industrial arms, the Powerball is more compact and has better
mobility and quick-change system. The battery allows for an 8-hour operation time without
charging. However, the high cost of the arm limits the customers to large factories.

7|Page
 Figure 3: SCHUNK Powerball Lightweight ARM LWA 4P [3]

igus Robolink
The igus Robolink shown in Figure 4 is a five degrees of freedom serial manipulator
driven by cables. It was designed and manufactured by igus, a German based worldwide
manufacturer and distributor of different types of plastic products [4]. The manipulator joint is
made of fine polyamide and carbon fiber. The cables, which control the joints, are made of
Dyneema, a high-strength synthetic fiber that has a tensile strength of 4,000N. The links
between joints are made of aluminum tube. This results in a relatively light weight arm. In order
to largely reduce the moving mass, the actuators and control module are separated from the
body of the manipulator. Because motors are not directly mounted on the manipulator, the
weight of the motor does not create a moment force on the arm. Stepper motors are usually the
first choice with their system. The Robolink manipulator has four different joints, a rotating joint,
a pivoting joint, a base joint, and a 2-axis joint.

8|Page
 Figure 4: igus Robolink [4]

Users can assemble this arm with different combinations as seen in Figure 5. A few
advantages of the Robolink arm are that they are lightweight, compact, flexible, modular,
compliant, and waterproof. The Robolink can be controlled by any control solution, such as
PLC, SPC, Matlab, ROS etc. The cost of the Robolink is about $370 - 750 per joint without the
motor, and one six degrees of freedom arm with complete actuating system costs about $6000.
This price point is low-cost when compared to the Powerball but is likely out of the budget of a
hobbyist. Another disadvantage for the Robolink is that although each joint is modular and
lightweight, the heavy motor box still prevents the arm from being portable. Besides this,
assembly takes time as each cable must be run from the motor, throughout the arm until it
reaches its respective joint.

9|Page
 Figure 5: Various joints and possible configurations of the igus Robolink [4]

Kinova Jaco Robotic Manipulator

The Jaco robotic manipulator was produced by Kinova, a Canadian company engaged
in the design and manufacturing of innovative personal robotic products [5]. The Jaco robotic
arm is the size of human arm. It was designed for rehabilitation and can be mounted on any
assistive equipment such as a wheelchair. The manipulator has six degrees of freedom and can
reach the distance of 0.9m. The body is made out of carbon fiber, which results in the arm
weighing 5.7kg. A DC brushless motor combined with a 1:100 gear ratio Harmonic Drive allows
for a maximum payload of 1.5kg. The arm has a powerful control system with a 1 Mb/s
CANBUS to provide fast data transfer between each joint. The arm can be controlled using any
existing controlling software. Kinova also provides control software as well as an API. The retail
price of Jaco manipulator is about $48,000 making it much too expensive for a student or
hobbyist.

10 | P a g e
Figure 6: Kinova Jaco Robotic Manipulator [5]

11 | P a g e
 Table 4 below compares these arms.

Low-cost robotic SCHUNK Powerball

Low-cost Compliant
manipulator joint Lightweight Arm igus Robolink
Robotic Manipulator
(MQP) LWA 4P
1-DOF Rotating
Joint, 2-DOF
Degree of Freedom 7 DOF 1 DOF per joint 2 DOF per joint
Swiveling and
rotating joint
1000g for rotator- pivoting: 1.2kg,
Maximum Payload 2 kg (4.4 lb) joint, 500g for 6 kg rotation: 0.5kg,
elevator-joint 6-DOF link: 0.5kg
Link 1:
285.75mm
Length 1.0 m to the grip radius > 700 Link 2:
Size
wrist mm 450.85mm
Link 3:
450.85mm
Repeatability
<= 3mm unknown 0.06 mm 1-2mm
Accuracy
Low-cost stepper ML-50 50:1 Brushless DC 5 EC 45 flat
motors with timing belt Permanent Magnet Motors with Ø42.8 mm,
Drive and cable drives and Geared Motor, permanent magnet brushless, 50
the series-elastic Beetle B231 Gear brakes, harmonic Watt, with Hall
components motor 231:1 drive gear sensors.
Dead Weight 11.4 kg (25 lb) 12.5 kg 345g
72°/s under
Joint Speed 1.5 m/s 5-10 cycles/min
nominal load
24V DC, 3A avg.,
Power Supply 12V DC 12V DC 24V DC / 10 A
12A max
Approximately
$291.68 for elevator- $6000 for 6-DOF
Cost $4,135 joint, $203.04 for arm, $370 - 750
rotator-joint per joint without
motor
Universal CANopen
Communication RS485 communication
interfaces
Pivot: ±90°,
Range of motion ±170°
Rotation: ±170°
Compatible with
ROS with OROCOS-KDL ROS node or KEBA
Control System any control
Library CP 242/A
solution
Plastic &
Material Plywood Plastic Metal
Aluminum
Table 4: Comparison of robotic arms

12 | P a g e
2.3 Material Study
When designing and manufacturing our robotic joint, there were a wide range of metals,
plastics, and composites from which the joint could be produced. Each of these materials has its
own advantages as well as disadvantages. In order to determine which would be suitable for
this application, the various options were studied. In Table 5 below, popular mechanical design
materials are compared.

Young’s Modulus Tensile Yield Ultimate Tensile

(GPa) Strength (MPa) Strength (MPa)

Aluminum 6061 276 310

Steel (SAE 1040 hot

207 290 524
rolled)

Brass 110 250

PVC Pipe 2.3 51

PLA Plastic 3.5 50

Table 5: Material Properties

In addition to considering the physical properties of each material, their ease of

manufacture is another important design consideration. PLA plastic, used commonly in 3D
printers, is very easy to work with and can be formed into complex designs quickly. The
downside being a PLA part is not as strong as an equivalent mental piece. For this reason, a 3D
printed part would be ideal for a prototype or non-loadbearing part. Even within metals, the
manufacturing advantage is something to be considered. As seen in Table 5, steel has a much
higher tensile strength than Aluminum but when machining, aluminum is a much easier material
to work with.

2.4 Components Study

Motors
Motors are the most important component in the robotic arm system. There are two
classes of motors: alternating current (AC) motors and direct current (DC) motors. The power of
motors can range from 1/100 hp to 100,000 hp and the rotation speed can range from less than
0.001 rpm to 100,000 rpm [6]. The physical size of motors also varies in a large range. Because
our robotic arm requires modularity, meaning that the system is powered by external battery, we
limited the selection to DC motors. The diagram below lists most of the existing DC motors.

13 | P a g e
 Figure 7: Types of Motors [6] [7]

There are three main types of DC motors: stepper motors, brushed DC motors and
brushless DC motors. The brushed DC motor is an internally commutated electric motor that
has been used for over 100 years [8].
The four essential parts of a brushed motor are the field winding, armature winding,
brushes, and commutator as seen in Figure 8.

14 | P a g e
 Figure 8: Illustration of a brushed DC motor's components [28]

The field winding is used to generate the magnetic field if a permanent magnet is not
used. Once current is applied, a magnetic field is generated around the armature which pushes
against the magnetic field generated by the field winding. The magnetic force then drives the
armature to rotate. The brushes of the motor are made with graphite material, making contact
with the commutator and transferring the current from commutator to armature. When the
armature winding becomes horizontally aligned with the field winding, the magnetic field
generates zero force. At this point, the commutator reverses the direction of current, reversing
the magnetic field [8].
The brushless motor was first used in 1962. It overcomes many disadvantages the
brushed motor has. A brushless motor replaced brush and commutator with an electronic
controller to continually switch the phase to the winding to keep the motor rotating. Instead of
rotating the armature, a brushless motor rotates the permanent magnets around a fixed
armature [9].

15 | P a g e
 Figure 9: Illustration of a brushless DC motor's components [17]

A typical brushless motor uses Hall sensors (shown as a grey square with a semicircle
cut-out in Figure 9) to detect the change of the magnetic pole and switch the coil (seen as
yellow lines in Figure 9) on and off accordingly. It works similar to the brush and commutator on
brushed motor. In a brushless DC motor, the permanent magnet, which alternate in polarity
(shown as the blue and red ring), rotate around the static coil core as the coil’s polarity is quickly
alternated between two polarities.
A stepper motor is a special type of brushless DC motor that divides a full rotation into
many steps (shown in Figure 10). Each step has equal angle and the motor can be commanded
to move and stay at one of these steps [10].

Figure 10: Illustration of a Stepper motor's components [18]

Compared to regular brushless DC motor which usually works together with a feedback
encoder, a stepper motor has better motion control because it is operated in steps. As long as
the motor does not skip steps, accurate position control can be achieved in an open loop
system. However, if the motor skips steps, the motor position needs to be recalibrated. Adding
an encoder to the shaft can solve this problem. A stepper motor also has a large holding torque

16 | P a g e
because the winding current flows through the motor even when the motor stops moving.
However, it wastes a lot of energy by holding the torque at the position.
In order to making a proper decision in terms of the motor type the team will be used for
the project, the comparison table has been made. The table below lists all the advantages and
disadvantages of brushed motor, brushless motor and stepper motor.

Brushed Motor Brushless Motor Stepper Motor

● Low cost ● Electronic controls the ● Electronic controls the

● Simple control motor motor
● No controller is required for ● No sparking and less noise ● No sparking and less noise
fixed speeds ● No need to replace ● No need to replace brushes
Pros ● Can operate in extreme brushes periodically periodically
environment ● Easy to cool down ● Easy to cool down
● High efficiency ● Higher speed range
● Higher speed range ● Accurate motion control
● Maximize holding torque

● Needs to replace brush ● Higher cost ● Higher cost

periodically ● Complex control ● Requires an
● The brush causes friction ● Requires an microcontroller to run the
and reduces the torque microcontroller to run the motor
Cons ● Poor heat dissipation motor ● Energy inefficient
● Higher rotor inertia ● Needs to recalibrate if
● Lower speed range motor skip steps
● Brush generates noise and
sparking
Table 6: Comparison of DC electric motor types [11]

17 | P a g e
Power Transmission

Harmonic Drive Gear

A Harmonic Drive Gear, also called a strain wave gear, is a type of mechanical gear
system based on elastic dynamics. According to the definition from Oxford Dictionary, harmonic
drive is a reversible reduction drive providing a large reduction ratio with effectively zero
backlash, high torque capability, and high efficiency [13]. The drive has concentric input and
output shafts and is of considerably lower mass and volume than comparable drives.

Figure 11: Harmonic Drive Gear [12]

As shown in Figure 11 above, the mechanical system consists of three basic

components: wave generator, which is an elliptical disk contains outer ball bearings; flexspline,
which made of an elastic material (usually a thin walled steel cylindrical cup); and the circular
spline, which is a rigid circular ring with teeth on the inside edge. The harmonic drive works by
directly connecting the motor shaft to the wave generator. As the motor drives the wave
generator rotating, the flexspline teeth, which are meshed with those of the circular spline,
change [12].
The flexspline has fewer teeth than the circular spline has; therefore for every full
rotation of the wave generator, the flexspline rotates only a little, which reduces the output
speed. The reduction ratio can be calculated as formula shown below:

𝑅𝑒𝑑𝑢𝑐𝑡𝑖𝑜𝑛 𝑟𝑎𝑡𝑖𝑜 = (𝑓𝑙𝑒𝑥 𝑠𝑝𝑙𝑖𝑛𝑒 𝑡𝑒𝑒𝑡ℎ − 𝑐𝑖𝑟𝑐𝑢𝑙𝑎𝑟 𝑠𝑝𝑙𝑖𝑛𝑒 𝑡𝑒𝑒𝑡ℎ) / 𝑓𝑙𝑒𝑥 𝑠𝑝𝑙𝑖𝑛𝑒 𝑡𝑒𝑒𝑡ℎ

18 | P a g e
 A harmonic drive has many advantages comparing to other gearing mechanism.
Including: simple structure (3 components), compactness, light weight, zero backlash, achieving
high gear ratios with less space, high positional accuracy, high repeatability, high torque to
weight ratio, coaxial input and output shafts, and constant performance without wearing [12].
Due to the many advantages harmonic drive systems have, they have been largely used in
areas including: industrial machine, robotics, aerospace, medical equipment, and radar tracking
systems etc. Harmonic drives have many different types in accordance with different
requirements. The criteria of selecting the proper drive type includes input/output torque ratio,
lubrication type, output rpm, size and weight, maximum backlash, and maximum transmission
error [14].

Series Elastic Actuator

Series elastic actuators are an unconventional design of robotic arms and legs. The idea
is to introduce the elastic component between motor and load, mimicking the structure of human
muscle. The traditional design between motor and load was to use a gear train to reduce the
speed. However, using gears causes friction, backlash, torque ripple, and noise. To solve some
of the problems listed above, a compliant component is added between gear train and load, as
shown in Figure 12 below.

Figure 12: Diagram showing setup of a series elastic actuator

19 | P a g e
 There are many advantages of using a series elastic actuator as well as disadvantages
as shown in Table 7.

Advantages Disadvantages

Human safety and less chance to damage the environment Less precision

Accurate and stable force control Takes more space

Shock tolerance Not as stiff

Lower reflected inertia Complex design

Energy storage

Table 7: Advantages vs disadvantages of a series elastic actuator [15]

One remarkable advantage of a series elastic actuator is it allows the robot arm to have
more accurate and stable force sensing than any other force sensing method such as a strain
gauge or a current sensor. By calculating the difference of displacement between motor and
load, multiplied by the elastic coefficient of the compliant component, we know how much force
is exerted to the load. Figure 13 below is one example of applying series elastic actuator to the
robot. This is the robotic leg that was made by a WPI graduate student as his research project.
Series elastic actuators are used for all the joints the robot has.

Figure 13: Series elastic actuator used in a WPI project [16]

20 | P a g e
3. Methods
3.1 Introduction
In order to ensure that the task specifications devised in earlier portions of the projects
were met, certain methodologies had to be followed. Those selected can be seen below
alongside with reasoning behind selecting each.

3.2 Methods
From the beginning of the project, a systems engineering approach was used.
Background research was conducted to understand the market gap and create a project
proposal. A market gap was identified and certain design specifications and goals were chosen.
This technique was decided upon to run the project as it would be ran in an engineering
company.
Figure 14 below is the “V” of systems engineering [29]. On the left side of the V the
project is decomposed to know the finer details and make a concept design. The system
requirements for our project are our design specifications. In the initial design section our high
level design is outlined.

Figure 14: The Systems Engineering Approach [29]

After research was completed, various designs were considered. This was in an attempt
to find a broad number of solutions to the problem outlined in the project. Each design was
ranked via various weighted criterion such as cost, ease of manufacturing, precision, etc. There

21 | P a g e
were multiple charts made to understand all options. Each of these focused on a different
design idea including drive trains and degrees of freedom, Figure 15. The highest ranking
design aspects were then selected and combined as the base design to further develop.

Figure 15: Design Charts

Once a general design was selected, modeling of the parts was started. SolidWorks was
used as the primary design and simulation program for the physical models of the joints. Parts
were created and assembled in order to test the viability of each design. Aspects such as
motion and stress can be simulated without needing to create a physical prototype. This saved
both time and money adding to the efficiency of the design process.
Concurrent with the mechanical design, the electrical components were also selected
and designed. Because of the desired level of control and intelligence within each joint, wiring
diagrams and custom circuits had to be mapped out digitally before being prototyped to ensure
that they would work as desired.
After completion of a test circuit, programming could be started. Using a software
engineering approach, various features were first selected. From there, the simplest of the
features were instituted and tested before progressing to a more complex, full program
Once the digital version of the prototype is tested, a physical one was built. The
prototype was made of readily available materials. After the creation each of the prototypes,
physical testing was done to ensure that the design specifications were met.

3.3 Summary
After the methods of completing the project were decided upon, the process of
implementing these systems began. This began with creating a design based off our system
needs using the systems engineering approach.

22 | P a g e
4. System Design
4.1 Mechanical Design

Explanation of mechanical design

The first section of the joint that was designed was the gear system. It was believed that this
would be the most important design factor in determining the physical size of the joint and would
therefore determine how the rest of the physical system was configured. After the most compact
possible gearing solution was determined, the body was designed to fit as tightly as possible
around it. A circle was drawn with its center in the axis of rotation of the output whose radius
extended to encompass the gears. The result was a 4.25” diameter circle which would form the
base of the cylinder-like shape of the joint. Because of the geometry of the worm and worm
gear, the distribution of the gears within the body of the joint ended up being lopsided as seen in
Figure 16. This allowed for wiring and electronics to be placed within the empty side of the body.

Worm
Motor

Support Pillar
Joint Base

Figure 16: Internal gear configuration

The final gear ratio achieved in our design was 160:1 with 160 revolutions of the motor
shaft resulting in one full revolution of the output shaft though this was in reality limited to 270°.
In order to achieve the 160:1 gear ratio, three gear reductions were used, two being with spur
gears and the last with a worm and worm gear. A worm gear was used because they allow high
torque to be generated as well as prevent back driving of the output shaft. This meant that our
joint would be both strong enough to meet our design specifications as well as that power would
not have to be provided for the joint to retain its static position. The tradeoff with this type of

23 | P a g e
gear is that its efficiency can range from 90% down to 50%. The spur gears used consisted of
two 12 tooth and two 25 tooth gears. This meant two 2:1 gear reductions. The benefit of spur
gears is that they have a relatively high efficiency of 94-98%. In our case, the disadvantage was
that in order to achieve a high enough gear reduction, the output gear would be four times larger
than the input gear for the 4:1 reduction needed. Since the gear system would have the greatest
influence on overall size of the joint, two 2:1 reductions were used in order to be more compact.
Initially this project was envisioned as having two different types of joint, one for rotation
along the arm axis and another for rotation offset 90° from the arm axis. This idea was
eventually dismissed in favor of using one joint to achieve both types of rotation. As seen in
Figure 17, by using a joint which was at its core is two cylinders stacked on top of each other
which could rotate in line with each other, along with linkage adapters, a joint could be created
which would fulfil the roles of both aforementioned joints.

Arm Adapter

Arm
Adapter

Electronic Speed
Controller
Figure 17: Picture of both types of linkage adapters

By the end of the design phase, there were two different versions of the robotic joint
designed. The first one was nearly completed before the second one was started. This second
version was created due to several problems in the original design, key amongst them were
manufacturability and strength. Though this was a new design, many ideas used in the original
were used.

24 | P a g e
 Figure 18: First version of the joint design

Between the two designs, the same gear arrangement was used although in the new
version, the metal plates which hold the bearings are no longer as critical to the strength of the
body. This was due to the introduction of three support pillars. In the preliminary (or first) design,
most of the shear stress on the shaft would have been transmitted through a plate on the side of
the joint with the rest being supported by the plates holding the bearings of the gear system.
These new pillars added rigidity to the body of the joint while also distributing any stress more
evenly.
During initial brainstorming, the wires which supplied power and serial communication
would have entered through plugs on the bottom of the joint and exited through the top. It was
later realized that this may cause wires to become tangled around internally as they would have
to rotate between the two halves. For this reason, the final version has all wiring and electrical
components in the bottom half of the joint. The circuit board was designed such that the top of
the joint could rotate around the bottom without interfering or snagging on any wires.
In order to save weight, the structural components of the joint were machined out of
aluminum while the outer shell was 3D printed. Since PLA plastic is less dense than aluminum,
weight was saved while still being able to contain all the internal components of the joint within a
shell.

FEA Simulation
The CAD model was tested using the simulation package in SolidWorks. Screws were
added to the analysis to achieve the most accurate result. A force of 1N was applied to identify
the critical points in the model. There are many orientations that the joint could be placed in the
arm. The test run had the bottom fixed and a rotational torque about the center shaft. This would
be if the joint was orientated as a rotary joint. Figure 19 displays the results of the deflection
test. The largest deflection was on an arm of the top cross with a value of 3x10-4 in.

25 | P a g e
 Figure 19: Deformation Result

In Figure 20 Von Mises stresses are displayed on the part with the same input conditions
as the deflection analysis. As expected the highest stress of 3x106 N/m2 is at the surface of
force application.

Figure 20: Von Mises Stresses

26 | P a g e
Machining
First, a 3D printed model was printed and assembled to verify fits of gears and electronic
components. In Figure 21 the model is show without the bearing plates to allow the inner
features to be seen.

Figure 21: 3D printed prototype

Figure 22: Rounded edges can be seen in all the inside corners of the base

Rounds were added to the inner features of the parts to allow for manufacturability as
seein in Figure 22. Below is a diagram of the manufacturing process. From the SolidWorks
model, G code is generated using Esprit, a CAM software. To prep the machine, tools are
assembled, the tools and parts are probed, and the stock is fixtured. The Haas MiniMill, VM-2,
and Drill Mill Center were used to create all the parts.
Figure 23 shows a screenshot of Esprit to make the vertical support. To use Esprit first
chains are made that reference the main geometry. These chains are in the program manager
on the left and are shown in blue in the graphics screen. Milling operations reference these
chains for the tool pat. To make the vertical support the contouring of the part was done with a
⅜” endmill. A ½” ball end mill was used to make the curved edges, reference 1 in the graphic
below, to prevent the need for an additional fixturing.

27 | P a g e
 1

Figure 23: Screenshot of ESPRIT

The VM-2 runs at twice the RPM as the MiniMill; therefore, it is able to facilitate faster
machining. The probe for the VM-2 is also in better condition than the MiniMills. For these
reasons the VM-2 was used to make complicated parts and fine tune bearing holes for press
fits. In Figure 24, the stock material is shown on the right and the base is shown on the left. To
reduce the spindle load the step over was set to 30% of the max for the endmill. A sacrificial
piece of metal was used as a fixturing point to make the base, top cross, and the top. The
sacrificial metal was required to be rotated 180 degrees while making the base; thereby
effecting the relation between the holes and the remaining cuts by 3 mm. The holes are used to
attach the link collar. Only the position of the arm in relation to the joint was affected.

Figure 24: Machined base piece next to stock aluminum block

To assure that no preloading would result from assembly the top screw holes on the
vertical supports were not drilled and tapped until all other components were assembled. The
complex geometry of the parts resulted in burs from machining not being able to be removed
using a de-burring knife. Sand blasting the components removed all burs, resulting in a safe to
handle assembly.

28 | P a g e
Figure 25: Assembled aluminum components

29 | P a g e
4.2 Electrical Design
Before designing the electrical system for the modular joint, several requirements had
been decided upon based on the initial requirements for the project. The electrical requirements
are listed below:
 The motor shall be able to generate enough power to rotate the joint at the minimum
speed of 45 degree per second through gear reduction and be able to lift a mass of 1kg
at 1 meter from the axis of rotation.
 The motor driver shall provide enough current to power the motor.
 The joints shall operate from a 12V DC power source.
 The encoder shall be able to detect the joint movement with a precision of 1 degree.
 The limit switch shall set the range of motion to at least 180 degrees.
 The microcontroller shall be able to read in the sensor data, receive the data from the controller,
calculate the PID algorithm, and output the PWM signal.
 The control circuits shall be integrated within the joint.

The electrical diagram that is shown in Figure 26 below was initially designed to
demonstrate the electrical system of joint. From the diagram we can see all the joints are
connected with one serial I/O line and one power line. The serial I/O line provides data
communication between base and joints. Each joint has an individual microcontroller that
connects to an encoder and motor controller. The encoder is physically mounted on the
rotational shaft to provide positional feedback. The microcontroller then compares the position
feedback with desired position and calculates the speed of motor using a PID algorithm.

Figure 26: General electrical system diagram

30 | P a g e
Component Selection

Motor Selection
In order to select the motor that best fits the requirement, the team calculated the
required motor power based on the task specifications. Assuming we have a three linked robotic
arm with two pitch joint J1 and J2 as shown in picture below.

Figure 27: Free body diagram of a 3-link planar robotic arm

The equation to calculate the static torque for two joints is shown below:

Figure 28: Static torque calculation equation

Based on our requirement, the robot should be able to lift up 1 kg weight at the distance
of 1 meter with the speed of 45 degree/s, which is 7.5 rpm. The maximum weight of each joint
should be no heavier than 0.5 kg. The PVC pipe we are going to be using weighs about 0.42 kg
per meter and each link will be about 0.5m long. With the following values:

𝐿1 = 𝐿2 = 0.5 𝑚
𝑚1 = 𝑚2 = 0.21 𝑘𝑔
𝑊𝑙 = 1 𝑘𝑔
𝑊𝑚 = 0.5 𝑘𝑔

We then calculated:

31 | P a g e
 𝑇𝑠1 = 14.31 𝑁𝑚
𝑇𝑠2 = 5.41 𝑁𝑚

Since the joints are the same for robotic arm, we choose Ts1 as the maximum torque
arm needs to provide. As shown in the torque-speed graph of motor below, in order to run a
motor in its maximum efficiency, the speed should be running around 70%-90% of the no load
speed and the torque should be around 10%-30% of the stall torque [19].

Figure 29: Torque-speed curve of a DC motor [20]

Thus, we can estimated the stall torque to be around 71.55Nm and the speed to be
around 9.4 rpm.
We then calculate the maximum power needed to lift the weight. The maximum power
can be calculated by multiplying half of stall torque and half of the no load speed, as shown in
equation below:

𝑇𝑠𝑡𝑎𝑙𝑙 𝑆𝑚𝑎𝑥𝑟𝑝𝑚
𝑃𝑚𝑎𝑥 = ∗ / 9.5488
2 2
71.55 𝑁𝑚 9.4 𝑟𝑝𝑚
= 2
∗ 2
9.5488 = 17.6 𝑊

We calculated the maximum power is around 17.6W. Because the motor loses power
during operation, we need to give a safety factor of 1.5 to assure the right amount of power
needed; therefore, the motors that fill our requirement need to have power at least 26.4W.

32 | P a g e
Motor Driver Selection
To drive the brushless motor, the team decided to use electronic speed control, or ESC.
It is an electrical circuit used to control the direction and speed of brushless motor. ESC
controllers are often used by model hobbyists in radio controller model planes and cars.

Figure 30: ESC wiring diagram [21]

As shown in Figure 30 above, the ESC is powered by external battery. The battery type
can be Li-ion or NiMH. It has three wires connected to receiver which are ground, power, and
signal. In our application, the three wires will be directly connected to microcontroller. Most
ESCs have built-in battery eliminator circuit (BEC) that can provide 5V power through a power
line. It is an electronic circuit used to provide electrical power to other circuit without the need for
a battery. 5V is used to power the microcontroller. The microcontroller then sends a PWM signal
through signal line to control the direction and speed of motor. On the right side, the ESC
provides three-phase power to drive the brushless motor.
In order to provide enough current to drive the motor without damaging the ESC, the
appropriate ESC must withstand continuous current that is larger than the motor’s maximum
current, which is 12A. Initially, the team chose an Exceed RC Proton 12A brushless ESC which
is able to withstand continuous current up to 18A. The input voltage can range from 5.6V to
16.8V. However, when testing the ESC we realized that Exceed RC Proton 12A brushless ESC
did not support reverse motor direction because this controller was designed to be used on RC
planes to rotate the propeller, which only rotate in one direction. Through research on the
Internet, we found out an ESC used on RC cars has the mode to allow the motor to rotate in
both directions.

33 | P a g e
 Figure 31: Turnigy TrackStar 25A ESC [33]

The team soon found that the Turnigy TrackStar 25A brushless car ESC (shown in
Figure 31) can provide continuous current up to 25A and is able to reverse the motor direction.
This ESC has BEC that provides 5V/1A output to microcontroller.

34 | P a g e
Encoder Selection
In a close-loop system, an encoder plays a significant role by providing feedback data
into the system. There are many existing encoders on the market and they are categorized in
two types, incremental encoders and absolute encoders. Incremental encoders provide a series
of periodic signals when motion occurs. The number of cycles corresponds to the increments of
motion. Many incremental encoders are two-channel, meaning that there are two output signals
and the two signals are offset by one cycle. Two-channel outputs allows encoder to detect the
direct of motor. Absolute encoders provide unique output for every position in the range of
motion. Therefore the output corresponds to the position. Initial research determined four
encoders of either type that would work, as shown in the table below.

Name Picture Company Price Type Resolution Shaft

Diameter

E4T OEM US Digital $39 Optical 300 6.35 mm

Miniature Miniature cycle/rev
Optical Kit Encoder
Encoder

ENC300 CPR Solutions $35 Quadrature 300 2 mm

Easy Roller Cubed Encoder cycle/rev
300 CPR
Quadrature
Encoder

MAE3 US Digital $69 Magnetic 4096 2 - 6 mm

Absolute Absolute position/rev
Magnetic Kit Encoder
Encoder

AEAT-6010- Digikey $25.68 Magnetic 1024 6 mm

A06 Absolute position/rev
Encoder

Table 8: Encoder Selection [22][23][24][25]

The first two encoders are both 2-channel quadrature encoders and have 300 cycles per
revolution. The last two encoders are both magnetic absolute encoders. The First one has 4096
position per revolution, which is four times of the resolution than the second. The first also has
the price that is almost three times higher than the second one. In the design requirements, the
tolerance of angular position should be within 1% of the full rotation of the joint, which is 3.6
degrees. The 1024 position per revolution can measure 0.35-degree position, which is much

35 | P a g e
more accurate than the requirement states. The team decided to use AEAT-6010-A06 magnetic
absolute encoder.

Figure 32: Encoder Electrical Connections [27]

As shown in figure above, AEAT-6010-A06 magnetic absolute encoder has five

connections, which are 5V supply voltage (VDD), chip selection (CSn), supply ground (Vss),
serial clock (CLK), and serial data (DO).

Communication Protocol
Among all different types of serial communications, the Inter-Integrated Circuit (I2C)
communication is the most efficient and uses the least physical communication ports. I2C allows
multiple slave devices to communicate with one or multiple master devices. It was designed for
a short distance communication. I2C uses two wires (SCL and SDA) to communicate, SCL
synchronizes the clock and SDA transmits data. The protocol supports up to 1008 devices to
interconnect and communicate at a rate up to 400kHz [26]. Figure 33 demonstrates how
multiple joints could wired in parallel using I2C communication.

36 | P a g e
 Figure 33 Diagram demonstrating how multiple joints could be wired in parallel [26]

Microcontroller Selection
To select the microcontroller that best fit requirements as well as being low-cost, the
team created a detailed functional diagram of a joint’s internal electrical components, which can
be seen in Functional Diagram section. As discussed in the previous section, an ESC has a
three-pin connection to a RC receiver, which in our case was replaced by microcontroller. The
three pins were 5V power output, ground, and PWM signal input, which can be connected to a
digital IO port on the microcontroller. The absolute encoder we selected has five-pin connector
shown in Figure 34 below. Chip Select, Serial Clock, and Serial Data were connected to IO
ports on microcontroller. The limit switch required one digital IO port and ground. To
communicate with main board, we used I2C communication which requires two ports. Through
researching, we found Arduino Pro Mini 328 5V/16MHz that fit all the minimum requirement and
has a relatively small 18x33mm size.

37 | P a g e
Figure 34: Functional diagram of a joint's internal electrical components

38 | P a g e
Schematic Design
After selecting all the electrical components, we designed a schematic sheet for the joint
circuit as shown in Figure 35. The schematic was designed based on the specification of pins
on the Arduino Pro Mini. To eliminate the signal noise of the limit switch, we added a signal
debouncing circuit between limit switch and digital IO port on Arduino Pro Mini. To ensure the
constant 5V power input, we connected a 0.1μf capacitor between ESC power line and Vcc on
Arduino Pro Mini.

Figure 35: Arduino pinout diagram

PCB Design
After completing the schematic design, we designed a PCB board using ExpressPCB
software based on the schematic sheet. As shown in Figure 36, the overall board size is 35mm
by 35mm. There is a circular hole in the middle of the board to allow shaft to pass through. The
board is mounted on the top cross of the joint. The black lines are the wires that connect all the
pins. The thin lines are the signal wire. The thicker lines are the power and ground lines
because they have more current going through than signal lines.

39 | P a g e
 Figure 36: PCB diagram

4.3 Programming Design

The programming design of the modular robotic arm system can be separated into three
modules, which are the program on the joint to control the rotation, the program on the main
board to distribute the position data to the joints, and the program on the computer to interact
with users. The general requirement for the program are:
 The user shall be able to plug in the main board into computer and control the robot arm
using any interactive user interface
 The robot arm shall be able to automatically rotating and moving to target position once
receiving the command from computer
The detailed programming requirements were:
 There is a simulation of the arm in the computer for the user to interact with
 Software needs to use inverse kinematic tool to convert the target position to the
rotational angle for each joint
 The master board takes the rotation data of all the joints and distributes to each joint
through I2C and waits for response
 When each joint receives the data, it starts PID controller to control the movement
 Once the joint reaches the target angle, the joint will send back to main board the actual
angle and success message. If the joint fails to reach the target angle, it will send the
actual angle and a failure message
 Once the main board receives the failure message, it will terminate the action and send
error message to computer.
 The computer will display the error message to users.

40 | P a g e
Program Flow Chart
Figure 37 shows the flow chart of main controller unit. As shown in the graph, the
initialization function is first called. In the initialization function, the system first checks the status
of all the joints. Then it reads in the actual position for each joint and set default target position
same as actual position. Finally, the system starts timer interrupt. In the interrupt service routine,
the system first requests the actual position from each joint. If the system receives the actual
position data from joint, it updates the local actual position data as well as sending actual
position data to PC. The system then request target position data from PC. If the system
receives the target position from PC, it updates the local target position data as well as sending
the data to the joint.

Figure 37: Main Controller Flow Chart

Figure 38 shows the flow chart of the joint controller unit. Once the program starts, it first
checks if the limit switch has been pressed or not. If the limit switch is not pressed, the system
starts to read in the encoder value and set the value as the input for the PID algorithm. Then the
system computes the output value using the PID algorithm. The output value is the speed to
drive the motor. The joint controller program has two I2C interrupts that are called when the
main controller requests for the actual position data or when the joint controller receives data

41 | P a g e
from the main controller. When sending data to the main controller, the program parses in
encoder value as a 2 byte package and then sends to the main controller. When receiving the 2
bytes data from main controller, the joint controller parse in the data as the set point for PID
algorithm.

Figure 38: Sub Controller Flow Chart

4.4 Summary
Following a systems engineering approach the team designed a joint that was analyzed
using simulation packages in SolidWorks. The team also designed the customized electrical
system for the joint and wrote programs to control multiple joints.

42 | P a g e
5. Results
5.1 Introduction
As each subcomponent of the joint was completed, various levels of testing were run to
ensure that the final joint would behave as per our task specifications. The electrical and
software subsystems were the first to be completed followed by the mechanical components.

5.2 Results

Mechanical Results
In the end, we were able to machine or purchase all of the necessary mechanical
components. These can be seen in Figure 39.

Motor &
Top
Gearbox

Adapters

Microcontroller,
Bottom
encoder, etc.

Figure 39: The mechanical and electrical components of the modular robotic joint

The most significant obstacle in implementing the mechanical systems was dealing with
the tolerances needed for the gearbox to work properly. In an earlier 3D printed version of the
joint, the tolerances produced by the 3D printer were not sufficient to test the design. The shafts
ended up misaligned to the point where the gears could not rotate freely even without a load. At
the time, the proposed solution was to create the parts out of aluminum for the final version
which would have tighter tolerances and would be able to operate correctly.
During the manufacturing process of milling the various parts out of aluminum, it was
discovered that the tolerances from WPI’s CNC machines were not much better than those of
the 3D printer. We speculate that the reason for this is that the machines have lost their
precision due to improper use from students learning to machine on them.
In order to get the gearbox to operate correctly, several modifications were made. The
hole on the plate which held the brushless DC motor had to be enlarged to fit the shaft and the

43 | P a g e
mounting holes needed to be repositioned. On the other bearing plate, two of the bearing holes
were too close together forcing two gears into each other and preventing rotation. This was
solved by replacing one of the bearings with a bushing which was then drilled into and enlarged
to allow the gear shafts to be further from each other.

Figure 40: The assembled modular robotic joint

When finally assembled as shown in Figure 40, the output shaft rotated with an input
rotation from the motor as designed but was not without much friction between the gears. This
led to the gear train not rotating as smoothly as desired, even with ball bearings, which could
potentially reduce the output torque to the output shaft.

Electrical and Programming Results

After the designing the customized circuit board, we ordered a PCB from a circuit
printing service, ExpressPCB. Once the printed circuit board was delivered, we soldered all the
components onto the circuit board and tested the functionality of each component separately.
We first tested the debouncing circuit for the limit switch by programming a simple
Arduino program that would print to the serial monitor “0” when a switch was not pressed and a
“1” when switch was pressed. Then we tested the PWM port. We used a power generator to
provide a 12V output power to the ESC controller. Because the microcontroller was powered by
a computer via USB, we did not need the ESC to provide 5V BEC power. Therefore, when
connecting the ESC with the circuit board, we only connected the PWM signal and ground
wires, leaving the power wire disconnected. We then programmed a simple Arduino code that
varied the PWM output value between 1000 and 2000. The motor was able to spin at different

44 | P a g e
speeds and directions. The motor spun forward at maximum speed when setting value to 2000.
The motor spun backward at maximum speed when setting the value to 1000. The motor
stopped when the value was set to 1500. We also tested the encoder by writing code that
triggered a clock signal every 50 milliseconds and printed the value that was read from encoder
to serial monitor. As the result, the encoder was able to send the values from 0 to 1023 when
rotating the shaft from 0 to 360 degrees. Finally, we tested the I2C communication between the
joint circuit board and main board. We used an Arduino Uno as the main board and connected it
to our customized circuit board using two jumper wires. The Arduino website provided a library
specifically for I2C communication, which we used for our serial communication between the
two boards. We were able to successfully send string data from the main board that was then
received by the joint board as well as send string data from joint board and have it be received
by the main board.

45 | P a g e
Integrated System Testing Results
ID Need Description Final Result Explanation
Through adapters, each joint was able
All joints should be able Each joint contained all to connect mechanically to any other in
Modular to be used in necessary parts to operate series. Due to the serial connections,
N 01
joints combination with any alone yet was able to each joint could receive its own
other joint(s) interface with others designated commands and operate
independently
The adapters which would allow the
A joint should be able This was not tested but is
joint to connect to an arm-piece and lift
to lift a 1 kg load held potentially not possible
a weight were 3D printed due to time
N 02 Payload at one meter due to poor meshing and
constraints and are unable to hold the
horizontally from the internal friction of the
necessary torque unlike a piece
axis of rotation gearbox
machined from aluminum
The motor is able to reach the required
While under no load, This has been tested
velocity. However, due to the friction
Rotational each rotational joint running motor under full
N 03 between gears, the motor needs to be
velocity should be able to rotate speed. The joint is to
running under full speed to be able to
at at least 45 deg/s rotate 60 deg/s.
rotate smoothly
A joint should be able The adapters which would allow the
This was not tested but is
to rotate at a minimum joint to connect to an arm-piece and lift
Rotational potentially not possible
of 45 deg/s while a weight were 3D printed due to time
N 04 velocity due to poor meshing and
holding a 1 kg load one constraints and are unable to hold the
under load internal friction of the
meter horizontally from necessary torque unlike a piece
gearbox
the axis of rotation machined from aluminum
Since only one joint was fully
The joint should not
manufactured, multiple joints could not
Motion interfere with the range This was designed but not
N 05 be tested together though through
constraints of motion of other tested
simulating Solidworks assemblies, this
joints on the arm
task spec is believed to have been met
The final joint measured
Each fully extended Through the use of a relatively small
9.29 cm in height and 10.8
Length of joint should be brushless DC motor and a custom
N 06 cm in diameter at its
joint approximately 15 - 20 gearbox, we were able to stay below
widest point without the
cm long our desired demensions
attached adapters
The wires on the arm Each joint contained all its necessary
should present no wiring within its shell. The wires to
This was met individually
safety hazard. connect each joint to each other
N 07 Wire Safety by each joint but not when
Including, areas where however, need to be run along the
assembled into an arm
fingers could be outside of the arm pieces instead of
crushed internally
Much of the aluminum stock was
Though difficult to give a salvaged from other projects and all of
final cost, it is believed the machining was done within WPI’s
The price of materials
that each joint would cost machine shop meaning that the only
N 08 Cost for each joint should be
approximately $400 cost involved was time. Estimates were
no more than $150
without a profit margin if gathered from local machine shops
mass produced though which gives us an approximate
total cost
Figure 41: Comparison of design specifications to the final results

46 | P a g e
 ID Need Description Final Results Explanation
The encoder used was 10-bit meaning
Any rotation to a
This is achievable through that it was able to measure accurately
software defined
the use of the built-in to within approximately 0.352 degrees
N 09 Accuracy angular position should
absolute encoder along of rotation. This coupled with proper
be within 1% of the full
with proper PID setup PID values would allow a user to attain
rotation of the joint
this accuracy
Any three consecutive The encoder used was 10-bit meaning
rotations to the same This is achievable through that it was able to measure accurately
software defined the use of the built-in to within approximately 0.352 degrees
N 10 Precision
angular position should absolute encoder along of rotation. This coupled with proper
be within 1% of the full with proper PID setup PID values would allow a user to attain
rotation of the joint this precision
The structural strength of each joint
was much more than sufficient for the
The final joint weighed potential loads on the joint meaning
Maximum Each joint should not
N 11 0.584 kg without the much excess material remained, which
weight weigh more than 0.5 kg
adapters or outer shell added extra weight. Removing this
extra weight would have proven a
difficult machining task though
Since each joint could fulfil the role of
A pitch joint should be both a pitch joint and a roll joint
able to rotate at least through the use of adapters, all joints
Range of Each joint could rotate 270
N 12 120 degrees. A roll joint retained the same freedom of rotation.
motion degrees
will be able to rotate at The design and mechanical dimensions
least 180 degrees of the adapters allowed for at least 270
degrees of rotation
The voltage and power requirements of
The joints should
Each joint operated off of the electrical components were
N 13 Power supply operate from a 12 VDC
a 12 VDC power supply selected such that each joint would
power source
operate off of 12 VDC
Each joint should
contain a
The only input requirements to each
microcontroller and
Each joint contained the joint are a power source and a serial
On board sensors needed to
N 14 necessary electronics and command. Each joint is then able to
electronics accept positional input
sensors to achieve this handle motion on its own based on
and move to the
these
desired position within
the defined accuracy
Figure 42: Comparison of design specifications to the final results [con't]

47 | P a g e
Thermal Analysis
When designing an electromechanical system, analyzing the thermal properties of the
system is important in ensuring the device works properly. This is because electrical
components have operating temperatures which, when exceeded could cause failure of the
circuit while structural materials can be weakened and melt at higher temperatures.
For our modular robotic arms joint, excessive heat could be a potential issue for several
components. Key amongst these was the Arduino Pro Mini whose Atmel microcontroller has a
rated temperature range of -40℃ to 85℃ [30]. The PLA (Polylactic acid) plastic used for the
outer shell of is also of concern with regards to its temperature. At 60-65℃ PLA undergoes
glass transition meaning it becomes softer and rubber-like [31]. At 173–178℃, PLA will begin to
melt. Because of this, our ideal maximum temperature achieved inside of the joint would be
below 60℃ while temperatures up to 85℃ could be tolerated as the PLA plastic does not
comprise any of the structural elements within the system.
The components that could create the most heat within the joint are the motor controller
and motor with the Arduino generating a slight but likely negligible amount of heat. By
comparison, the Exceed RC Rocket brushless motor used in our system consumes up to 90
Watts of power when under load [32]. Its maximum efficiency is listed at 75% meaning that
while consuming maximum power, under ideal efficiencies, 22.5 Watts is lost due to heat and
radiation.
In order to power the brushless motor, a Turnigy TrackStar 25A brushless motor ESC
was used. This motor controller can provide continuous current up to 25A and bursts of up to
90A [33]. The motor controller comes with a built-in heatsink meaning that heat generation and
dissipation is a legitimate design concern.
To assure that performance issues would not arise due to heat building up inside the
joint, ANSYS Workbench was used to perform a thermal study on our joint. The first step in this
analysis was to create a simplified Solidworks model of the joint in order to make thermal
simulation easier. The critical components were kept such as the motor and motor controller but
parts such as the outer shells were combined into a singular piece and some of the details of
the base and top were removed. This can be seen in Figure 43.

48 | P a g e
 Figure 43: Simplified joint for thermal analysis

This model was then exported from Solidworks and imported into Workbench. Within
Workbench the various material properties were set such as material type. For the outer shell,
polyethylene was used instead of PLA plastic. Since Workbench did not contain PLA within its
library of materials, polyethylene was a good substitution and a material that could potentially be
used in a more mass produced version of the joint since 3D printing is not ideal for mass volume
production.
Convection, internal heat generation, and initial temperature conditions were all set for
the various components within the assembly. In the simulation seen in Figure 44, the ambient
temperature was set to 22°C, slightly below room temperature. The base, shell, and top parts
each began at 26°C to simulate having already been heated above the ambient temperature.
The motor and motor controller were both set to 50°C in order to simulate having been heated
by operation. These are temperatures which could be realistically seen from these components.

49 | P a g e
 Figure 44: ANSYS Workbench thermal simulation

Figure 45: ANSYS Workbench temperature vs. time graph

Based on the part specifications, the heat given off by the motor would be 22.5 Watts
under ideal conditions. This heat generation was also defined within Workbench. Convection
and conduction coefficients based upon individual material properties were also taken into
account.
The results seen from this simulation show that the hottest components in the system
reach a steady state temperature of about 60°C as shown in Figure 44. Though this result is
under ideal heat dissipation and motor heat generation conditions, this steady state takes
approximately 250 seconds to achieve meaning that the motor would have to operate for that
long under maximum load to reach these temperatures. This is unrealistic as the motor can only
rotate for a limited amount of time before reaching the rotational limits imposed by the geometry
and limit switches of the joint. The joint would also need to be supporting a kilogram at a meter
from the axis of rotation horizontally for the current draw to rise high enough to generate the
22.5 Watts of heat. Because of the simulation as well as these reasons, heat dissipation within
the joint does not appear to be a concern. Though only an approximation, the maximum
temperature simulated is 25°C below the maximum limit of 85°C for the electrical components
and takes too much time to reach to be an issue.

50 | P a g e
5.3 Summary
Most of the design specifications defined at the beginning of the systems engineering
process were met by the final iteration of the joint. Most of the issues arose from the precision of
the machining equipment used to manufacture the joint. The joint was successful however in
proving that the original objective could be realistically met through future iterations of this
project.

51 | P a g e
6. Summary and conclusions
6.1 Introduction
The final implementation of the modular robotic arm joint proved mostly successful
though not entirely as imagined. The reasoning, issues encountered and recommended future
work can all be seen in the sections below.

6.2 Summary of System Design and Results

The joint created by our group consisted of a single degree of freedom modular arm
system. With adapters, a user would be able to connect multiple joints together by using SCH80
PVC pipe which would be easy to cut to a custom length. Each joint then only requires power
and an I2C serial message to operate. In this serial message, a user can set the speed, position
or PID values of the joint in order to control it.

6.3 Overall assessment

Summary of issues
The parts machined were small and complex. It required fixturing surfaces as small as
2
1cm . The machine exerts large forces while milling away the material. These forces pushed the
part out of the fixturing multiple times. To make the top part the Drill Mill Center (DMC) had to be
used. The program was too large to fit on the MiniMills and the VM-2 was equipped for a five-
axis lab for a class. The DMC does not have probing abilities; therefore, the tools were probed
on the MiniMills and the numbers were transferred to the DMC. While making the top part the
tool did not travel in the z+ direction enough to clear the part, resulting in a chipped part and a
broken endmill.

Recommended future work

The two major improvements which could be made to this project are the machining
process and the final weight of each joint. Besides these, the joint was able to meet its design
specifications.
The tolerances of the final machined pieces proved to be inadequate to produce the
gearbox as desired from the Solidworks model. Potential solutions to this include using better
CNC machines, creating a design in which tolerances are less important or using other
techniques besides milling parts from aluminum. The CNC machines available to our group
have been used countless times by students learning to machine parts and have therefore lost
their precision and accuracy over time from normal wear and tear. It is likely that our parts would
have been better in terms of tolerances and accuracy had a newer machine been user or had a
machinist with much more experience milled the parts. Due to the nature of our gearbox design,
tolerances between and within parts was very important. If another design had been used in
which tolerances were less important or had a premade gearbox been used, the operation of
the final joint would likely have been improved measurably. Neither of these options seemed
possible or viable for various reasons though during our design process. A final option would be

52 | P a g e
to consider a different manufacturing process besides CSC milling. A 3D printed prototype was
attempted but the same tolerance issues arose. It may be possible to use a higher quality 3D
printer, potentially one which used better materials than PLA plastic, to create better versions of
the required parts.
The weight of the final joint also proved to be an issue. Since our design specifications
called for the joint to be able to achieve a certain desired torque, it is likely that the joint would
struggle to move other joints connected further in series as well as a payload since a majority of
power would be used to lift the weight of the joint. The current joint contains excess material
which could be removed in future implementations while still meeting the required structural
strength requirements. Another critical source of excess weight are the gears used in the gear
train. The worm gear was made from brass which is much more dense than aluminum which
could be a potential replacement material. The size of each gear could likely have been reduced
as well in terms of both width and diameter. This would also have the added benefit of reducing
the overall size of each joint.

6.4 Conclusions
While not every task specification defined at the beginning of this project was met, the
final modular robotic arm joint proved that through future versions and iterations of this project, a
modular arm system could be developed which could fill the market gap intended by this project.
Through the use of better manufacturing techniques and lighter components, the results of this
project could be improved with little time and effort. Our team was also able to use a systems
engineering approach to simulate a real engineering design process and create a functioning
robotic system. Because of these reasons, our team believes that we were successful in
completing this project

53 | P a g e
7. Bibliography
[1] M. Quigley, A. Asbeck, and A. Y. Ng, “A low-cost compliant 7-DOF robotic manipulator,” in
IEEE International Conference on Robotics and Automation, 2011, pp. 6051 – 6058.
[2] J. Baldiga, C. McCarthy, S. Fitzell, T. Watson, “Design of Low Cost Modular Robotic
Manipulator Joints,” Worcester Poly. Inst. MA. 2009.
[3] Schunk. Powerball Lightweight Arm LWA 4P [Online]. Available: http://mobile.schunk-
microsite.com/en/produkte/produkte/powerball-lightweight-arm-lwa-4p.html
[4] igus inc. Robolink: Articulated joint modules for robots [Online]. Available:
http://www.igus.com/wpck/6658/robolink.
[5] JACO 2 | Robotics [Online]. Available: http://kinovarobotics.com/products/jaco-robotics.
[6] Magnet Man. “Experiments with motors.” [Online]. Available:
http://www.coolmagnetman.com/magmotor.htm.
[7] Electical4U. “Types of DC Motor Separately Excited Shunt Series Compound DC Motor.”
[Online]. Available: http://www.electrical4u.com/types-of-dc-motor-separately-excited-shunt-
series-compound-dc-motor/.
[8] M. Brain. (2006, Dec. 15). “How does a brushless electric motor work?” [Online]. Available:
http://electronics.howstuffworks.com/brushless-motor.htm.
[9] “Brushless DC electric motor,” in Wikipedia, the Free Encyclopedia [Online]. 2014. Available:
http://en.wikipedia.org/wiki/Brushed_DC_electric_motor.
[10] “Stepper motor,” in Wikipedia, the Free Encyclopedia [Online]. 2014. Available:
http://en.wikipedia.org/wiki/Stepper_motor.
[11] Dynetic Systems. “Brushless vs Brushed Motors,” [Online]. Available:
http://www.dynetic.com/brushless%20vs%20brushed.htm.
[12] “Harmonic Drive,” in Wikipedia, the Free Encyclopedia [Online]. 2014. Available:
http://en.wikipedia.org/wiki/Harmonic_drive.
[13] D. Ince, “Harmonic Drive,” in A Dictionary of the Internet. Oxford University Press [Online].
2013. Available: http://www.oxfordreference.com.
[14] Z. Susman. ECE 5320. Class Lecture, Topic: “Harmonic Drive.” Dept. of Electrical and
Computer Engineering, Utah State University, Logan, UT, Mar. 11, 2010.
[15] G. A. Pratt and M. Williamson, “Series Elastic Actuators.”. Intelligent Robots and Systems
95.'Human Robot Interaction and Cooperative Robots', Proceedings. 1995 IEEE/RSJ
International Conference on. Vol. 1. IEEE, 1995.
[16] V. Chernyak, “The Design and Realization of a Sensitive Walking Platform” Worcester Poly.
Inst. MA, URN etd-042412-160012. 2012
[17] G. Lazaridis, “How Brushless Motors Work (BLDC Motors),” [Online]. Available:
http://pcbheaven.com/wikipages/How_Brushless_Motors_Work/.
[18] L. George, “Stepper Motor or Step Motor,” [Online]. Available:
https://electrosome.com/stepper-motor/.

54 | P a g e
[19] E. Cenciper, “How to Select a Motor? (Part 2) – Understand Motor Specification,” [Online].
Available: http://www.robot-and-machines-design.com/en/Articles/Mechanics/Tips-Guides/557-
Motor-Selection-How-To-Select-A-Motor-Understand-Motor-Specification.html.
[20] J. Pearman, “Motor Torque-speed Curves,” [Online]. Available:
http://www.vexforum.com/showthread.php?t=70588.
[21] DraganFly, “Manual of Thunder Power Sensor less Brushless Motor Speed Controller
ESC,” [Online]. Available: http://www.rctoys.com/pr/pr-html/thunder-power-brushless-speed-
controller-esc-manual.htm.
[22] US Digital, “E4T OEM Miniature Optical Kit Encoder,” [Online]. Available:
http://www.usdigital.com/products/encoders/incremental/rotary/kit/E4T?specification=phase-
relationship.
[23] RobotShop, “ENC300CPR Easy Roller 300 CPR Quadrature Encoder.” [Online]. Available:
http://www.robotshop.com/en/solutions-cubed-enc300cpr-easy-roller.html.
[24]US Digital, “MAE3 Absolute Magnetic Kit Encoder,” [Online]. Available:
http://www.usdigital.com/products/encoders/absolute/rotary/kit/mae3.
[25] Digi-Key, “AEAT-6010-A06,” [Online]. Available: http://www.digikey.com/product-
detail/en/AEAT-6010-A06/516-2635-ND/2210457.
[26] Sparkfun, “I2C,” [Online]. Available: https://learn.sparkfun.com/tutorials/i2c.
[27] Avago, “AEAT-6010/6012 Magnetic Encoder datasheet,” 2011.
[28] Experimentalev, “Motor Brush Replacement – A Step by Step How To,” [Online] Available:
https://experimentalev.wordpress.com/2011/03/22/motor-brush-replacment-how-to/
[29] GrabCAD, “Systems engineers, the multi-domain 3D revolution is coming for you,” [Online]
Available: http://blog.grabcad.com/blog/2015/03/18/multi-domain-3d-revolution/
[30] Wikipedia, “Polylactic acid,” [Online] Available: https://en.wikipedia.org/wiki/Polylactic_acid
[31] Atmel, “Atmel 8-bit microcontroller,” [Online] Available: http://www.atmel.com/Images/Atmel-
8271-8-bit-AVR-Microcontroller-ATmega48A-48PA-88A-88PA-168A-168PA-328-
328P_datasheet.pdf
[32] Hobbypartz, “Exceed RC Rocket 2205-1400KV Brushless Motor,” [Online] Available:
http://www.hobbypartz.com/86ma01-2205-1400kv.html
[33] Hobbyking, “Turnigy TrackStar 25A Brushless ESC,” [Online] Available:
http://www.hobbyking.com/hobbyking/store/__14630__Turnigy_TrackStar_25A_1_18th_Scale_
Brushless_Car_ESC.html

55 | P a g e