son of edgar - mechanical...
TRANSCRIPT
-
Faculty of Engineering, Computer & Mathematical Sciences School of Mechanical Engineering
SON of EDGAR
State-Space Control of Electro-Drive Gravity-Aware Ride
Final Report Oct 20th, 2006
Authors N. P. Baker C. P. Brown D. R. S Dowling J. L. Modra D. J. Tootell Supervisor Dr. B. S. Cazzolato
-
i
Executive Summary
This report, SON of EDGAR: A Self-Balancing Scooter, covers the modelling and
design of a co-axial, two wheeled scooter to provide a method of human transport.
The aim of the project was to design and build a device that behaves in a similar
manner to that of the Segway Human Transporter, the first and only self-balancing
vehicle to be commercially available (Clark et al., 2005).
The outcomes of a range of previous attempts at creating self-balancing devices are
discussed in literature reviews contained within this report. The design of SON of
EDGAR draws upon the advantages and disadvantages of the previous designs in an
attempt to create a robust, easy to use device. Components for the device were
selected after extensive research had occurred and the mechanical and electrical
design was implemented using the characteristics of these components.
The process used to maintain the scooter in the upright position, is similar to that used
by humans to balance. By recognising the angular position the device is from upright,
a correction is made using a state space controller which moves the wheels in the
appropriate direction to return the device to the upright position.
SON of EDGAR has been successfully built and tested and is a robust, easily rideable
scooter. Many enhancements have been made from the 2005 EDGAR model, making
SON of EDGAR a far superior device. The majority of the primary goals set out at the
beginning of the project have been completed and all goals will be completed prior to
the 2006 University of Adelaide Mechanical Engineering project exhibition.
-
ii
Disclaimer
This report is the work of the five authors. Any information that has been obtained
from other authors has been respectively cited where used.
..
Nicholas Baker
Date:
..
Cameron Brown
Date:
..
David Dowling
Date:
..
Justin Modra
Date:
..
Daniel Tootell
Date:
-
iii
Acknowledgements
The SON of EDGAR project would not have been possible without the help of many
individuals. Firstly the project group would like to thank Dr. Benjamin Cazzolato for
his support and help throughout the project, it has been much appreciated and of great
assistance.
The project team would also like to thank the technicians in the Mechanical
Workshop. Richard Pateman and Bill Finch have been helpful in offering advice
however inconvenient it was for them. A special mention must go to Steve Kloeden
who has given up so much of his precious time to discuss design issues with the team.
The project group would also like to thank the team from the Instrumentation
Workshop, in particular Silvio De Ieso, for their guidance in all things electronic.
The project team would like to thank Felix Grasser for the information provided that
assisted us greatly with our understanding of the project as well as Trevor Blackwell
for his correspondence.
mailto:[email protected]
-
Contents
iv
Contents
1 Introduction............................................................................................................1
2 Background............................................................................................................3
2.1 Fundamental Control Principles ....................................................................3
2.2 Recent Research and Development ...............................................................4
2.2.1 The Segway Model ................................................................................4
2.2.2 Blackwells First Model.......................................................................12
2.2.3 Blackwells Second Model ..................................................................15
2.2.4 The EDGAR Model .............................................................................19
2.2.5 RC Control of a Segway ......................................................................23
2.2.6 JOE.......................................................................................................25
2.2.7 Complementary Filters.........................................................................28
3 Project Goals and Component Specifications......................................................33
3.1 Project Goals................................................................................................33
3.2 Specifications Development ........................................................................34
3.3 Basic Component Specification...................................................................36
4 Component Selection ...........................................................................................39
4.1 Motor............................................................................................................39
4.2 Gyroscope ....................................................................................................41
4.3 Accelerometer ..............................................................................................42
4.4 Power Supply ...............................................................................................43
4.4.1 Power Distribution Board ....................................................................46
4.5 Encoders.......................................................................................................46
4.5.1 Encoder Resolution Divider.................................................................48
-
Contents
v
4.6 Motor Controller ..........................................................................................49
4.7 Capacitive Sensor.........................................................................................51
4.8 Steering Mechanism.....................................................................................52
4.9 Wheels..........................................................................................................54
4.10 User Displays ...............................................................................................55
4.11 BluetoothTM Device .....................................................................................57
5 Detailed Hardware Design...................................................................................58
5.1 Motor Bracket ..............................................................................................58
5.2 Support Bracket ...........................................................................................59
5.3 Boss and Flange ...........................................................................................61
5.4 Base Plate.....................................................................................................63
5.5 Electrical Distribution Design......................................................................64
5.6 Battery Mounting .........................................................................................66
5.7 Encoder Mounting .......................................................................................67
5.8 Handle Bars..................................................................................................70
5.8.1 Adjustable Base Angle.........................................................................71
5.8.2 Adjustable Height ................................................................................72
5.9 Motor Controller Mounting .........................................................................72
5.10 Auxiliary Electrical Mounting .....................................................................73
5.11 Switches and Ports Mounting ......................................................................74
5.12 Gyroscope and Accelerometer Mounting ....................................................74
5.13 Steering Mechanism.....................................................................................75
5.14 Finite Element Analysis...............................................................................76
6 Hardware Integration Design...............................................................................81
6.1 Driveline Design ..........................................................................................81
6.2 Aesthetics.....................................................................................................83
-
Contents
vi
6.2.1 Handle Bars..........................................................................................83
6.2.2 Wheels..................................................................................................85
6.2.3 Rear Tail Lights ...................................................................................85
6.2.4 Casing ..................................................................................................86
6.2.4.1 Fenders.................................................................................................86
6.2.4.2 Standing Platform ................................................................................90
6.2.4.3 Grip Tape .............................................................................................91
7 Software Implementation.....................................................................................93
7.1 Software Overview ......................................................................................93
7.2 Input and Output Data Types.......................................................................96
7.3 Interpreting the Inputs and Creating Useful Outputs...................................97
7.3.1 Measuring Pitch Angle ......................................................................101
7.3.2 Complementary Filter ........................................................................102
7.4 dSPACE DS1104 R&D Controller Board .................................................105
7.5 MiniDRAGON+ Development Board .......................................................106
7.6 Final Software Design................................................................................106
8 Control System Design ......................................................................................110
8.1 Introduction of State Space Control...........................................................110
8.2 Mathematical Model of the No Rider System........................................112
8.2.1 System Identification of the No Rider System...............................126
8.2.1.1 DC Motor Identification ....................................................................126
8.3 Mathematical Model of the Rider System .................................................131
8.4 Virtual Reality Model ................................................................................138
8.5 Controller Development.............................................................................140
8.5.1 Decoupling System ............................................................................140
8.6 Controller Development Using Mathematical Model................................143
-
Contents
vii
8.6.1 Control of Unstable Mass ..................................................................143
8.7 Control Development by Trial and Error...................................................147
8.7.1 Yaw Controller...................................................................................150
9 Testing................................................................................................................155
9.1 Open Loop Testing ....................................................................................155
9.2 Controller Testing ......................................................................................157
9.2.1 Linear Controller Testing...................................................................157
9.2.2 Sliding Controller Development and Testing ....................................159
9.3 Complementary Filter Tuning....................................................................163
10 Final Design .......................................................................................................169
10.1 Problems and Solutions..............................................................................173
10.1.1 Handle Bars........................................................................................173
10.1.2 Encoder Mounting .............................................................................175
10.1.3 Fenders...............................................................................................175
10.1.4 Problems with Software.....................................................................176
10.1.5 Mathematical Modelling Problems....................................................177
10.1.6 Problems with Programming the Microcontroller .............................180
11 Cost Analysis .....................................................................................................182
11.1 Component Costing ...................................................................................182
11.2 Group Member Work Hours......................................................................184
11.3 Workshop Costing .....................................................................................186
12 Project Outcomes ...............................................................................................187
12.1 Primary Goals ............................................................................................187
12.2 Future Goals...............................................................................................189
13 Recommendations and Future Work .................................................................190
14 Conclusion .........................................................................................................192
-
Contents
viii
References..................................................................................................................193
Appendix A Component Data Sheets ........................................................................196
Appendix B Design Concepts....................................................................................225
B1 Direct Drive with Bearing and Coupling .....................................................225
B2 Chain Drive and Belt Drive..........................................................................226
B3 Direct Drive with a Lowered Platform.........................................................228
Appendix C CAD Drawings ......................................................................................230
Appendix D Code ......................................................................................................255
-
Table of Figures
ix
Table of Figures
Figure 2-1: Photograph of Segway HT i180 (Segway Inc., 2006). ...............................5
Figure 2-2: Cross Section view of Segway Motors (Segway Inc., 2006)......................6
Figure 2-3: Diagram explaining mesh ratios of Segway gearbox (Segway Inc., 2006).7
Figure 2-4: Diagram of Segway Wheel and Tyre (Segway Inc., 2006). .......................7
Figure 2-5: Photograph of Segway NiMH battery pack (Segway Inc., 2006)...............8
Figure 2-6: Diagram showing the location of both the controller boards and Balance
Sensor Assembly (Segway Inc., 2006). .................................................................9
Figure 2-7: Top-View Diagram of Balance Sensor Assembly (Segway Inc., 2006)...10
Figure 2-8: Photograph of Balance Sensor Assembly (Segway Inc., 2006)................10
Figure 2-9: Diagram showing the location of the Balance Sensor Assembly (Segway
Inc, 2006). ............................................................................................................10
Figure 2-10: Photograph of Segway Fender (Segway Inc., 2006)...............................11
Figure 2-11: Exploded view of Segway HT (Segway Inc., 2006)...............................12
Figure 2-12: Trevor Blackwell on his self-balancing scooter (Blackwell, 2005)........13
Figure 2-13: The undercarriage of the scooter (Blackwell, 2005)...............................14
Figure 2-14: Blackwells Version 2 self-balancing scooter (Blackwell, 2005)...........16
Figure 2-15: The OSMC motor controller (Robot MarketPlace, 2006). .....................17
Figure 2-16: EDGAR fully assembled (Clark et al,. 2005). ......................................20
Figure 2-17: Modelled force distribution (Clark et al., 2005). ....................................21
Figure 2-18: HIPS wheel covers and MDF central platform (Clark et al., 2005)........22
Figure 2-19: JOE undergoing untethered testing (Grasser et al., 2002)...................25
Figure 2-20: A model of JOE with state variables and disturbances (Grasser et al.,
2002). ...................................................................................................................26
-
Table of Figures
x
Figure 2-21: Inclinometer performance versus true orientation (Baerveldt & Klang,
1997). ...................................................................................................................29
Figure 2-22: Rate Gyroscope performance versus true orientation (Baerveldt & Klang,
1997). ...................................................................................................................29
Figure 2-23: Contribution of the two sensors to the filtered output (Baerveldt and
Klang, 1997). .......................................................................................................31
Figure 2-24: Filtered output performance. The solid line is the true orientation and the
dashed line is the output of the complementary filter (Baerveldt and Klang,
1997). ...................................................................................................................31
Figure 4-1: Photograph of NPC-B81 (NPC Robotics, 2006). .....................................39
Figure 4-2: Torque characteristics of the NPC-T64. ...................................................40
Figure 4-3: ADXRS300 iMEMS gyroscope (Spark Fun, 2006)..................................41
Figure 4-4: The Crossbow Accelerometer. (Crossbow Technology Inc. 2006) ..........43
Figure 4-5: Sonnenschein Dryfit 500A 12V, 15Ah, cyclic SLA battery (Sonnenschein,
2006). ...................................................................................................................45
Figure 4-6: Power distribution board. ..........................................................................46
Figure 4-7: US Digital 100 bit incremental (US Digital, 2006) ..................................48
Figure 4-8: Encoder resolution divider (US Digital, 2006). ........................................49
Figure 4-9: The Open Source Motor Controller Board (Robot MarketPlace, 2006)...50
Figure 4-10: The motor controller and cooling fan (Robot MarketPlace, 2006).........51
Figure 4-11: Omron E2K-F Capacitive Proximity Sensor (Omron Corporation, 2006).
..............................................................................................................................52
Figure 4-12: Self-Centring twist grip as implemented on EDGAR (Clark et al., 2005).
..............................................................................................................................53
Figure 4-13: A 45mm linear slide potentiometer as implemented on SON of EDGAR
(Dick Smith Electronics, 2006)............................................................................53
Figure 4-14: 20 inch BMX wheel. ...............................................................................55
Figure 4-15: LED array................................................................................................56
-
Table of Figures
xi
Figure 4-16: Free2move BluetoothTM serial port plug (free2move.se, 2006)..............57
Figure 5-1: Motor bracket. ...........................................................................................59
Figure 5-2: 2006 Support bracket. ...............................................................................60
Figure 5-3: 2005 EDGAR support bracket (Clark et al., 2005)...................................61
Figure 5-4: Flange and boss assembly. ........................................................................63
Figure 5-5: SON of EDGAR on board power distribution. .........................................65
Figure 5-6: SON of EDGAR breakout board. .............................................................66
Figure 5-7: Three dimensional model of encoder without hole in casing (US Digital,
2006). ...................................................................................................................67
Figure 5-8: Photograph of back of NPC-B81 Motor. ..................................................68
Figure 5-9: Three dimensional model showing extension of output shaft and encoder
rotor......................................................................................................................68
Figure 5-10: Three dimensional model showing encoder casing housed inside motor.
..............................................................................................................................69
Figure 5-11: Three dimensional model of encoder enclosed inside motor casing. .....69
Figure 5-12: Exploded view of Encoder Mounting. ....................................................70
Figure 5-13: Adjustable Base Angle............................................................................71
Figure 5-14: Adjustable height bracket........................................................................72
Figure 5-15: Mounting of the motor-controllers..........................................................73
Figure 5-16 Mounting of ports and main power switch. ............................................74
Figure 5-17: Installed self-centring potentiometer. .....................................................75
Figure 5-18: Model used in FEA. ................................................................................76
Figure 5-19: Safety factor - Equivalent stress..............................................................77
Figure 5-20: Area of maximum stress. ........................................................................78
Figure 5-21: Total deflection. ......................................................................................79
Figure 5-22: Graphical representation of stress against cycles to failure for steel
(Linton, 2006). .....................................................................................................80
-
Table of Figures
xii
Figure 6-1: Initial angled bracket design. ....................................................................82
Figure 6-2: Final design. ..............................................................................................83
Figure 6-3: The handlebars. .........................................................................................84
Figure 6-4: Handle bar test rig. ....................................................................................84
Figure 6-5: Taillights to be mounted on rear of fenders with and without plastic
housing.................................................................................................................85
Figure 6-6: Overall casing assembly............................................................................86
Figure 6-7: Stealth design (left) and round design (right). ..........................................87
Figure 6-8: Finished male plug. ...................................................................................88
Figure 6-9: Two female moulds...................................................................................89
Figure 6-10: The final fender.......................................................................................89
Figure 6-11: Painted fenders........................................................................................90
Figure 6-12: Box section..............................................................................................91
Figure 6-13: Grip tape covered standing platform.......................................................92
Figure 7-1: dSPACE Breakout Box (Clark et al., 2005)..............................................95
Figure 7-2: MiniDRAGON+ Development Board (Clark et al., 2005).......................96
Figure 7-3: Simulink model - Steering input subsystem. ............................................98
Figure 7-4: Simulink model - Battery voltage subsystem. ..........................................99
Figure 7-5: Simulink model Encoder subsystem. ...................................................100
Figure 7-6: Simulink Model Motor controller subsystem. .....................................101
Figure 7-7: Tilt angle estimation using an accelerometer and a rate gyro (How & Park,
2004). .................................................................................................................103
Figure 7-8: Example of a complementary filter using a second order filter (Baerveldt
& Klang, 2006). .................................................................................................104
Figure 7-9: Example layout using the dSPACE ControlDesk. ..................................105
Figure 7-10: Simulink model - Overview of final dSPACE model...........................108
Figure 7-11: Simulink model - Sensor subsystem. ....................................................109
-
Table of Figures
xiii
Figure 8-1: Block diagram of basic state space model (Cazzolato, 2005).................111
Figure 8-2: Block diagram of state space system with observer and feedback control
(Cazzolato, 2005)...............................................................................................112
Figure 8-3: Free body diagram for the No Rider system. ......................................114
Figure 8-4: Free Body Diagram of motor connected to wheel. .................................127
Figure 8-5: Comparison of motor data and estimated models. ..................................130
Figure 8-6: Free body diagram of the rider system....................................................133
Figure 8-7: VR model. ...............................................................................................139
Figure 8-8: VR Simulink model. ..............................................................................140
Figure 8-9: Graphical representation of decoupling system. .....................................142
Figure 8-10: Pole zero map of unstable mass system. ...............................................144
Figure 8-11: Simulink model of controller testing. ...................................................145
Figure 8-12: Plot of pitch angle using mathematical model controller. ....................146
Figure 8-13: Simulink diagram of pitch and yaw control..........................................150
Figure 8-14: Simulink model of pitch and yaw control with steering. ......................152
Figure 8-15: Plot of wheel velocities and steering input for scooter turning on the
spot.....................................................................................................................154
Figure 9-1: Comparison of simulated data and measured data..................................156
Figure 9-2: Simulink diagram of pitch control. .........................................................158
Figure 9-3: Performance of linear controller. ............................................................158
Figure 9-4: Initial sliding controller design. ..............................................................160
Figure 9-5: Final sliding controller design.................................................................161
Figure 9-6: Sliding pitch controller performance ......................................................162
Figure 9-7: Photo of test rig. ......................................................................................163
Figure 9-8: Plot of IMU against the first and second order filter. .............................164
Figure 9-9: Performance of final complementary filter implemented on SON of
EDGAR..............................................................................................................167
-
Table of Figures
xiv
Figure 9-10: Contribution of the two sensor branches...............................................168
Figure 10-1: Final Mechanical system design. ..........................................................169
Figure 10-2: SON of EDGAR....................................................................................170
Figure 10-3: Damaged aluminium support bracket and the threaded mild steel plate.
............................................................................................................................174
Figure 10-4 : Re-welded handlebars. .........................................................................174
-
Introduction
SON of EDGAR 1
1 Introduction
In December 2001 a new form of transportation was unveiled. The Segway Human
Transporter (HT) was a revolutionary new way of moving people around. Consisting
of a standing platform between two coaxial wheels with handlebars protruding up
from it, its stability seems an impossible feat. Due to a very robust and responsive
control system coupled with various sensors and actuators, the Segway HT is almost
impossible to fall off.
In 2005, The University of Adelaide sponsored a final year mechanical engineering
project that was to produce a replica of the Segway HT. The project was named
EDGAR a self balancing scooter, which stands for Electro-Drive Grav-Aware Ride.
While the finished EDGAR was functional, there were some major issues with the
design and not all of the extended goals of the project were accomplished. Due to this
the School of Mechanical Engineering at The University of Adelaide decided that a
similar project should be run in 2006 which should address all of the shortcomings
found with EDGAR as well as achieving the goals that were not accomplished in
2005. Hence SON of EDGAR was born.
As with EDGAR, it was important that SON of EDGAR be of easy manufacture and
requires only off-the-shelf parts where possible. It should allow for a person of
average height and weight to safely ride it for over an hour and it should also be
aesthetically pleasing.
The team began the project by conducting research on self-balancing scooters and
related topics followed by critical reviews of the papers, articles and reports. This
allowed the team to then define the goals and specifications of the SON of EDGAR
project. The goals and specifications gave the team a framework from which the
selection of components could be based. A mathematical model of the system was
derived and a number of Simulink (MathWorks Inc., 2006) models were developed
-
Introduction
SON of EDGAR 2
converting the various sensor inputs to meaningful values. A state space controller
was constructed and implemented on the scooter and the control gains were then
manually tuned to improve rider comfort.
Various obstacles were encountered along the way. The project team, with the help of
Dr. Ben Cazzolato, addressed these obstacles in a methodical manner to achieve both
feasible and practical solutions.
Prior to The University of Adelaides School of Mechanical Engineerings 2006 final
year project exhibition all remaining project goals will be completed yielding the
SON of EDGAR project a resounding success.
-
2. Background
SON of EDGAR 3
2 Background
Control has been used in different forms for many thousands of years, as discussed in
Section 2.1. Over the years an increased number of uses for control have been
developed and with ever improving technology it seems the possibilities in the future
are only limited by ones imagination. In recent years, a number of teams and
individuals have developed a new form of transportation where humans travel on a
platform balancing on two coaxial wheels with motion induced by the tilting the
platform. Section 2.1 discusses the principles of control while Section 2.2 consists of
a literature review of the previous attempts.
2.1 Fundamental Control Principles
Control systems can be found all around us. They are a very important part of society
and have been for a long time. Around 300 B.C. the Greeks began using engineering
feedback systems. Ktesibios invented a water clock which operated by measuring the
amount of water which trickled at a constant rate into a measuring container (Nise,
2004). He used a float similar to that used in todays toilets to keep the water level,
which ensured that it flowed at a constant rate. Today control systems are used in such
things as missiles and robots as well as more mundane applications such as the cruise
control in our cars.
The four main reasons for building control systems as stated by Nise (2004) are:
Power amplification.
Remote control.
Convenience of input form.
Compensation for disturbances.
-
2. Background
SON of EDGAR 4
The application for a control system that this report is covering, the self-balancing
scooter, will make use of all four reasons mentioned above. It requires power
amplification, or power gain, to control the amount of power that the motors will be
given. The project team is aiming for remote control capabilities so that the scooter
can move without a rider. The main input of the control system is the tilt angle, while
the desired output from the system is the speed of the scooter. Therefore it is desired
that a convenient input angular position should yield a desired motor speed output.
The group also requires the scooter to be able to compensate for any disturbances it
may receive.
2.2 Recent Research and Development
There have been previous attempts in the field of self-balancing scooters. As there is a
commercially available product, through the Segway Company, others have tried to
understand and duplicate the product. This section covers some of the attempts made
by others and the Segway model.
2.2.1 The Segway Model
The Segway Human Transporter (HT) as shown in Figure 2-1 is the only
commercially available self-balancing vehicle on the market to date. The Segway HT
was unveiled in 2001 on a morning television program in the United States and was
released as a commercial product in 2002 (HowStuffWorks Inc., 2006). The Segway
HT was developed by Dean Kamen and his company DEKA Research and
Development Corporation. At the time of its release Kamen claimed that his machine
will be to the car what the car was to the horse and buggy (HowStuffWorks Inc.,
2006). Since its release however the Segway HT has not provided the revolution in
our travel methods as expected, however there is wide opinion that it offers a superior
option for getting around a city.
-
2. Background
SON of EDGAR 5
Figure 2-1: Photograph of Segway HT i180 (Segway Inc., 2006).
The Segway HT has been described as the worlds first self-balancing human
transporter (HowStuffWorks Inc., 2006). The Segway HT, unlike a car, has only two
wheels and unlike a bike they are axially aligned. To move forward or backward on
the Segway the rider simply leans either forward or backward respectively. To turn
left or right the rider twists the right handlebar the respective way.
It has been described that the balancing system of the Segway HT is similar to that of
the human body (Kamen, 2002). If a person stands up so they are out of balance, the
brain registers this due to a shift in fluid in the inner ear. The brain then triggers the
leg to move forward to prevent a fall. If the person continues to lean forward the brain
will continue to put a leg forward in an attempt to keep the person upright. The
Segway follows this same principle except it has wheels instead of legs, a motor
instead of muscles, microprocessors instead of a brain and a set of tilt sensors instead
of the inner ear balancing system (HowStuffWorks Inc., 2006).
The motors used in the Segway HT (shown in a Section view in Figure 2-2) are
brushless servo motors that are capable of 1.88 kilowatts (kW) or 2.5 horsepower
(HP), which at the time of original production made them the highest powered
motors, mass produced for their size and weight (Segway Inc., 2006). The magnets
http://www.segway.com/segway/view/i180midb.html
-
2. Background
SON of EDGAR 6
used in the motors are made from neodymium-iron-boron driven by a set of twelve
high powered, high voltage field effect transistors (FETS). Each motor is constructed
with two independent sets of windings (which can be seen in Figure 2-2) each driven
by a separate board making them electrically redundant. In normal operation both sets
of windings operate in parallel sharing the load, however, in the event of failure the
motor will instantly disable the faulty side and use the remaining winding to maintain
control of the Segway HT until it can be brought to a safe stop (Segway Inc., 2006).
The motor is designed to operate at levels up to 8000 rpm. This allows for the high
levels of power and torque that the motors can achieve (Segway Inc., 2006).
Figure 2-2: Cross Section view of Segway Motors (Segway Inc., 2006).
The gearbox used by the Segway HT is a two stage reduction system which provides
a 24:1 reduction. This allows the motors to operate at powerful and efficient speeds
throughout the range of the Segway HTs speeds (Segway Inc., 2006). The gears are
cut to a helical profile which both reduce noise and increase the load capacity of each
gear. The number of teeth on each gear (as illustrated in Figure 2-3) is chosen to
produce non integer gear ratios. This was done in an effort to reduce wear and tear on
the teeth. By having a non integer gear ratio between gears the teeth will mesh in a
different location each revolution thus maximizing the life of the gearbox.
-
2. Background
SON of EDGAR 7
Figure 2-3: Diagram explaining mesh ratios of Segway gearbox (Segway Inc., 2006).
The wheels used on the Segway HT (which are depicted in Figure 2-4) are
constructed from an engineering grade thermoplastic. The wheels are moulded around
a forged steel hub which eliminates the use of fasteners (Segway Inc., 2006). Each
wheel is fitted with a specially designed tyre which uses a silica based compound
instead of the usual carbon based compounds. This gives the tyres enhanced traction
and importantly for indoor use, minimises markings on floors. The tyres are also
tubeless which allows low tyre pressure. The wheels are mounted to the transmission
using a taper and hex design. This allows the wheels to be removed or attached using
a single nut while retaining the security of a more complex multiple bolt system
(Segway Inc., 2006).
Figure 2-4: Diagram of Segway Wheel and Tyre (Segway Inc., 2006).
-
2. Background
SON of EDGAR 8
The batteries used by the Segway HT are either twin nickel metal hydride (NiMH) or
lithium-ion (Li-Ion) rechargeable packs (shown in Figure 2-5). The batteries operate
at a nominal 72 volts (Segway Inc., 2006). These packs either consist of sixty 1.2V
NiMH batteries or ninety 0.8V Li-Ion batteries (Segway Inc., 2006). Each pack
contains a custom designed circuit board that constantly monitors the temperature and
voltage of the pack in multiple locations for redundancy. The circuitry of the Segway
HT enables the scooter to be charged by directly connecting it to the mains power.
The battery pack assembly (shown in Figure 2-5) is sealed using a vibration welding
technique that makes the outside of the pack a single continuous structure sealed from
moisture and structurally strong. The type of battery used directly affects the range of
the Segway. The NiMH and Li-Ion are rated to distances of 19 km and 39 km
respectively (Segway Inc., 2006).
Figure 2-5: Photograph of Segway NiMH battery pack (Segway Inc., 2006).
The Segway HT control and processing system is made up of two circuit boards,
housed within the vehicles chassis as shown in Figure 2-6. Each board monitors the
balance sensor assembly 100 times per second (100 Hz) to determine if the rider is
leaning forward or backward. Consequently the output commands are sent to the
motors at 1000 times per second (1000 Hz) with each board being responsible for one
of the two windings in the motors (Segway Inc., 2006). The Segway uses the Texas
Instruments TMS320LF2406A Digital Signal Processor (DSP) which operates at 40
million operations per second, has 32 kilobytes of flash memory and many peripheral
communication ports implemented on board the chip (Segway Inc., 2006). The actual
control and model used for the control of the system is not published as it is a
-
2. Background
SON of EDGAR 9
patented system. However it is known that the Segway uses the DSPs to implement
closed loop motor control and computation (Segway Inc., 2006).
Figure 2-6: Diagram showing the location of both the controller boards and Balance Sensor Assembly
(Segway Inc., 2006).
The balance sensor assembly (BSA) is packed with five solid-state, vibrating-ring,
angular rate sensors (Gyros) and two liquid filled tilt sensors (as shown in Figure 2-7
and Figure 2-8 respectively). The five solid state sensors rings are
electromechanically vibrated such that when they are rotated a small force is
generated which is detected by the internal electronics of the sensor (Segway Inc.,
2006). Each gyro is placed at a different angle which allows the BSA to measure
multiple directions. The data produced by the five gyros is constantly monitored by
the Segways on board computers. The onboard computer determines if any of the
five gyros is supplying false data thus a redundancy system can be put in place if this
occurs (Segway Inc., 2006). The two tilt sensors filled with an electrolyte fluid
provide a reference for the tilt of the Segway with respect to gravity. Only three
gyroscopes are actually needed for normal operation of the Segway however the extra
two are added for extra redundancy. The location of the mounting of the BSA is
shown in Figure 2-9.
-
2. Background
SON of EDGAR 10
Figure 2-7: Top-View Diagram of Balance Sensor Assembly (Segway Inc., 2006).
Figure 2-8: Photograph of Balance Sensor Assembly (Segway Inc., 2006).
Figure 2-9: Diagram showing the location of the Balance Sensor Assembly (Segway Inc, 2006).
The Segway has a weight sensor built into the platform. The weight sensor is used to
tell the Segway computers when the rider has either embarked or disembarked the
vehicle (HowStuffWorks Inc., 2006).
-
2. Background
SON of EDGAR 11
A design feature of the Segway HT not published is its mass distribution. It can be
seen in Figure 2-9 that the main standing platform which contains the majority of the
weight of the Segway HT is situated below the wheel axle line. This design feature
creates a natural pendulum effect which helps stabilize the Segway HT without the
help of control. This design feature also reduces the step up distance to the platform
giving a sense of balance and control to the rider.
The Segway HTs sensitive electronic equipment is housed in a strong die cast
aluminium chassis with a plastic fairing. The chassis is rated to be able withstand 7
tons of force (Segway Inc., 2006). The fairing design can be seen in Figure 2-10.
Figure 2-10: Photograph of Segway Fender (Segway Inc., 2006).
Shown in Figure 2-11 is an exploded view of a basic Segway HT which outlines the
individual components mentioned in this section.
-
2. Background
SON of EDGAR 12
Figure 2-11: Exploded view of Segway HT (Segway Inc., 2006).
2.2.2 Blackwells First Model
Building a Balancing Scooter, written by Trevor Blackwell, describes his successful
attempt at building a self balancing scooter modelled on the Segway HT. It describes
and explains the positive and negative aspects of his model as well as giving a
detailed description of each component within the scooter and inturn comparing them
all to the respective components of the Segway HT.
Blackwells scooter, as seen in Figure 2-12, is constructed from common off-the-
shelf components unlike the Segway which uses components that are custom made.
-
2. Background
SON of EDGAR 13
The motors used within Blackwells model are conventional 24V DC motors coupled
with a gearbox. These motors are used commonly in everyday applications such as in
powered wheelchairs. An advantage of using these motors is that they are designed to
have a wheel mounted straight on the output shaft in a direct drive manner. This
removes the need for bearings as they have already been incorporated in the design
and as a consequence the mechanical design has already been greatly simplified.
Figure 2-12: Trevor Blackwell on his self-balancing scooter (Blackwell, 2005).
A RoboteQ motor controller was used as the motor driver. This particular motor
controller can handle very large currents and is also quite small. Blackwell later
discovered that this motor controller severely limited the performance of his scooter.
The reasons for this are discussed in detail in Section 4.6 of this report.
With the Blackwell model, power is supplied to the motors via Remote Control (RC)
car battery packs as seen in Figure 2-13. The final design included 20 packs of 6
Nickel Metal Hydride (NiMH) AA cells. Although the motors require 24V,
-
2. Background
SON of EDGAR 14
Blackwell supplies them with 36V as he felt the project required more speed. Also it
is important to note that bridge rectifiers were used so that current did not flow
between the packs when the battery voltages were different. The system used
regenerative breaking which helps recharge the batteries when the scooter travels
downhill or when the scooter is decelerating. An advantage of using NiMH batteries,
is that they have a high energy density compared to Nickel Cadmium (NiCd) or
Sealed Lead Acid (SLA) batteries. A disadvantage is that they are substantially more
expensive than SLA batteries. Also the use of AA batteries, which have only a small
capacity, meant that a large number, 120, were required.
The wheels and tyres used in Blackwells first model were small and wide. This
arrangement has resulted in relatively less ground clearance than desired. A
consequence of this arrangement means that small obstacles may hit the bottom of the
scooter as it travels possibly damaging components. Also when comparing the
Blackwell wheels to those on a Segway, Blackwell states that the Segway's wheels
have a large moment of inertia which allows it to apply a reaction torque to the
chassis (Blackwell, 2005). This would allow for a much nicer ride as for small angles
of tilt the motors would act to straighten the chassis instead of moving the wheels to
balance the scooter.
Figure 2-13: The undercarriage of the scooter (Blackwell, 2005).
-
2. Background
SON of EDGAR 15
To detect the tilt motion of the scooter a ceramic rate gyroscope, in conjunction with a
2-axis accelerometer was used. According to Blackwell, the gyroscope was prone to
drifting especially when accelerating hard or going up a ramp (Blackwell, 2005).
The mechanical design of the model was very basic. It consisted of two aluminium
plates and a piece of aluminium tubing for the handle bars. This primitive design
gives a very plain appearance which is not very aesthetically pleasing.
With regards to safety, Blackwell has only two features. The first is a kill switch
which stops the scooter if the rider should fall off. The second safety feature is in the
form of logic control which shuts the scooter down should the tilt angle exceed 45
degrees.
The operation of the control system is centred on an 8-bit micro-controller which was
programmed in C code. A proportional-derivative (PD) controller was implemented
using the error and the change in error of the tilt angle to calculate the required torque
to be applied to the wheels.
According to Blackwell, his scooter works quite well although it is not quite as good
as the Segway HT product. Most of the issues found with the design of this scooter
have been taken into account when Blackwell designed his second model which is
discussed in Section 2.2.3.
2.2.3 Blackwells Second Model
Balancing Scooter Version 2, describes another self-balancing scooter built by Trevor
Blackwell. It tells how he made the scooter such that its performance would not only
exceed its predecessor, but also the commercially available Segway HT.
-
2. Background
SON of EDGAR 16
According to Blackwell the second version has achieved its goals as it was faster,
lighter and smoother than both its predecessor and the Segway i-Series. Also the new
iteration has more range than the first version (Blackwell, 2005). A photo of the
second model can be seen in Figure 2-14.
It is important to note that Blackwell has once again only used off-the-shelf parts that
were all ordered over the internet for the construction of this prototype.
The wheels used were 20 inch bicycle wheels. To attach them to the motor output
shaft Blackwell had to machine a new wheel hub which required him to then string,
tighten and adjust the spokes. He noted that this was a very tedious and time
consuming job. The 20 inch wheels increased the speed of the scooter by 43% and
gave the scooter 3 inches more ground clearance than the 14 inch wheels used on the
first version (Blackwell, 2005). Another additional advantage of using the wheels was
that due to their smaller width it was much easier to ride through a doorway.
Figure 2-14: Blackwells Version 2 self-balancing scooter (Blackwell, 2005).
-
2. Background
SON of EDGAR 17
One of the most important evolutionary changes made to the design was the selection
of different motor controllers. The RoboteQ motor controllers were replaced with
Open Source Motor Controller (OSMC) controllers as shown in Figure 2-15. In the
RoboteQ system, implemented in Blackwells first model, the battery charge could
not be read at a high frequency. As a result of this the battery charge could not be
implemented in the control of the scooter. This meant that the gains in the feedback
loop were dependent on the resistance of the batteries which change depending on
how long they have been running. For fully charged batteries the gain was very high
which caused the scooter to oscillate. Whereas if the batteries had lost the majority of
their charge the gain would be very small and the scooter would become unresponsive
due to a lack of power. With the new motor controllers the battery voltage is
measured 2000 times per second and the PWM signal can be adjusted so that the
motor controllers are sending the desired voltage to the motors.
Figure 2-15: The OSMC motor controller (Robot MarketPlace, 2006).
The old gyroscope was changed to a CRS03-02 gyroscope from Silicon Sensing
Systems which has lower noise and is virtually immune to vibration (Blackwell,
2005). The accelerometer, used to compensate for the gyroscopes drift, was upgraded
to an ADXL105 which has a higher saturation threshold. The higher saturation
threshold meant that the gyroscope is less likely to saturate on bumpy roads
(Blackwell, 2005).
-
2. Background
SON of EDGAR 18
Due to the faster response times of the gyroscope and electronics, as discussed above,
the overall control of the scooters balance is far greater. According to Blackwell the
scooter can be controlled entirely with the feet, even at high speed (Blackwell,
2005).
Blackwell has used 60 Panasonic D-cell NiMH cells which provide 8 horsepower to
the motors. A relay is used so that the battery packs can be charged separately. As
discussed in Section 4.4, NiMH batteries have a much larger energy density to NiCd
or SLA batteries and this time Blackwell has used D-cells which have a greater
current capacity than AA, so less cells were required.
Another feature which makes the new prototype superior is the Bluetooth wireless
connection. The Bluetooth connection allows parameters to be changed on the run,
data logging and remote controlled driving of the scooter. This is an exceptionally
good idea for the design especially at the testing and commissioning stage as the
cables that run from the computer to the scooter could potentially get tangled up. The
Bluetooth connection also aids in making the new design superior as it removes the
difficultly in locating, disconnecting and re-connecting the cables when parameters
need to be changed.
Another adaptation made within the new model is that the steering is controlled by
simply twisting the handlebars. It is interesting to note how Blackwell has made this
work. The handlebars do not move, rather strain gauges detect the torque applied and
send a signal to the micro-controller. There may be a disadvantage in this design as
there is a possibility that the rider may accidentally apply a torque to the handlebars
when they receive a slight jolt. This slight jolt could be caused by many unknown
disturbances including a wheel hitting a bump. The scooter will then unexpectedly
turn causing surprise and irritation to the rider.
-
2. Background
SON of EDGAR 19
Blackwell also includes extra protection for the batteries which is in the form of a
stainless steel plate that has been added to the chassis. Aesthetically it looks slightly
better than the first version but it is still possible to improve further.
Extra safety devices on this model include an on-off switch in addition to the kill
switch mentioned in the previous section of the report. Blackwell also added a beeper
which indicates dangerous battery charge and speed combinations to the rider. This is
a good idea as it allows the rider to know if they keep riding it in that particular
fashion that they will most likely fall off. This will encourage the rider to ease off,
rather than risk being hurt.
Summarising, it is mentioned above that Blackwells second attempt at building a
self-balancing scooter is not quite as safe as a Segway nor is it as aesthetically
pleasing. However despite these things, it is reasonably comparative in terms of ride-
ability. In the future Blackwell suggests that he would only like to improve the
aesthetics of the model, which suggests that he is very happy with the performance of
his scooter.
2.2.4 The EDGAR Model
The report, EDGAR, A Self-Balancing Scooter, gives a detailed analysis of the steps
undertaken by a team of final year engineering students at the University of Adelaide
to successfully build a self balancing scooter in 2005. The aim of the project was to
build a device that behaved in a similar manner to the self balancing scooter
commercially available from the Segway Company (as mentioned in Section 2.2.1).
The team were able to build a coaxial, rideable, self-balancing scooter but there
remained many areas for improvement within the design.
-
2. Background
SON of EDGAR 20
Figure 2-16: EDGAR fully assembled (Clark et al,. 2005).
The vehicle used an inertial measurement unit (IMU) to provide the angular position
information to an onboard microcontroller. While the IMU is able to provide accurate
data as it utilises 9 different sensors, it is an expensive component that could be
replaced with a single axis gyroscope coupled with an accelerometer for the purpose
of the project. The device had open loop steering. A proportional derivative (PD)
controller was introduced for pitch stability which is a common form of control and
relatively simple to implement, however, state space control is often a more attractive
and robust form of control for a multiple input, multiple output (MIMO) system.
Open loop steering requires regular pilot correction for any deviation from a straight
path due to the differing conditions at the wheel and motor on each side, which is
obviously not a desirable feature. This is an area where the project team felt
improvements could easily be made by including closed loop steering using encoders
on the wheels.
The mathematical model that was constructed by the team of university students
included many assumptions and simplifications which greatly reduce the quality and
robustness of the control system. One such simplification was modelling the person
and handlebars as one mass. Another problem with the design, related to the centre of
gravity of the vehicle which finished above the axle, giving the rider a sense of an
unstable vehicle and also led to increased requirements of the motors to maintain
-
2. Background
SON of EDGAR 21
stability when stationary. A basic free body diagram of the system can be seen in
Figure 2-17.
Fp - is the force applied by rider on chassis.
Mpg - is the mass of person x gravity.
V is the reaction force between chassis and wheel.
H is the reaction force between chassis and wheel.
is the torque applied by the motors.
x - is the horizontal axis.
y is the vertical axis.
COG is the centre of gravity of person and chassis combined.
- is the angle between vertical axis and the COG.
R subscript is the right wheel.
L subscript is the left wheel.
Figure 2-17: Modelled force distribution (Clark et al., 2005).
The control system was initially tested tethered to the dSPACE system which gave the
opportunity for quick and simple modifications of the controller and possessed many
useful analytical tools and visual displays to assist in troubleshooting. Once satisfied
and ready for untethered operation, the control system was downloaded to an on-
board micro-controller though an RS232 connection. This method of software
implementation appeared to be successful and relatively simple for troubleshooting.
The mechanical components of the device were of differing standards and obvious
budget constraints were visible in the selection of these components. The DC electric
geared motors were not able to produce an abundance of torque and had limited
power ratings which reduced the overall capabilities of the device. The Ni-MH 9Ah
batteries were able to supply the required power for the device but are not designed
-
2. Background
SON of EDGAR 22
for this type of use and suffer from a memory effect. These two problems led to
reduced capacity and therefore reduced power for the device over time. With no
feedback system in place from the batteries to create a dynamic control system, the
response from the device also varies throughout each battery discharge cycle. Another
component which suffered due to the budgetary constraints were the wheels. The
wheels that were selected failed to compliment the overall polished, aesthetically
pleasing design of the vehicle.
Extensive research was undertaken by the team in relation to the style and dimensions
of the handlebars. However it appeared that the implementation of the design was
incorrect as the angular position was uncomfortable with the handlebars too close to
the body. This could be overcome with an angular adjustment mechanism to allow
riders to find a more comfortable position. The outer casing of the vehicle was a
feature, with vacuum moulded high impact polystyrene (HIPS) wheel covers and a
medium density fibreboard (MDF) central platform. It looked impressive as can be
seen in Figure 2-18. The outer casing looked especially appealing in combination with
the well designed and manufactured handlebars which can be seen in Figure 2-16.
Figure 2-18: HIPS wheel covers and MDF central platform (Clark et al., 2005).
A rotary potentiometer inside a twist grip was used as the steering mechanism for the
device and was effective but not intuitive to use. A bank of LEDs was also placed in
the handlebars to provide the rider with a visual display of the devices current state,
-
2. Background
SON of EDGAR 23
e.g. on/off, batteries charge, balancing on/off. The LED arrangement was a simple,
yet effective, method of display for the rider and added to the great aesthetics of the
vehicle.
An important and effective safety feature that was built into EDGAR was the use of
two capacitive foot sensors to determine if the rider had stepped off or fallen off the
device. When one foot is placed onto the device it begins to balance, it will then
continue balancing until both feet come off. The device will then shut down for the
riders safety. This was implemented effectively and comes into effect regularly.
Overall, the report gives detail on the relative simplicity of building this unique device
but also gives rise to possible areas for improvement. The strengths of the EDGAR
design are the aesthetics (handlebars and outer casing), reasonably simple and stable
control system and that it is a fully functional self balancing scooter. The design is
weak in terms of power (batteries and motors), wheel selection, controller robustness
(PD control, assumptions and simplifications) and the use of open loop steering.
2.2.5 RC Control of a Segway
The RC (Remote Control) Control of a Segway paper (Cardi & Wagner, 2006) covers
the control theory to make a model Segway system stable.
The first step undertaken was the creation of a simplified linear model with full state
feedback. The model created was a great deal more simplified compared to other
models such as JOE as discussed in Section 2.2.6. The model, although rather
simplified was useful in understanding the dynamics of a self balancing scooter. The
model also incorporated the dynamics of the DC motors used on the model Segway
outputting a duty cycle which was what was planned for SON of EDGAR. The
motors were modelled as:
-
2. Background
SON of EDGAR 24
aaout Rr
KtKeDr
KtVs= *
Equation 2-1
Where:
out is the torque supplied.
Kt is the torque constant.
Ke is the back EMF constant.
Vs is the voltage supply.
D is a duty cycle.
ar is the armature resistance.
R - is the radius of wheels.
This modelling of DC motor dynamics was seen as a good basis to build a model of
the motors dynamics to be used on SON of EDGAR.
The next step undertaken in the paper was control of the system using a PD
(Proportional Derivative) controller. It was shown using PD control that the system
could be controlled effectively, however a velocity error was present which increased
with time. The use of PID (Proportion Integral Derivative) control was used to
attempt to alleviate this problem and was successfully implemented. However it was
clearly stated that the use of state space techniques would improve the control of the
system which was reassuring for the current group as state space control was hoped to
be used on SON of EDGAR.
The state space control of the system was implemented using a reduced order
observer which observed the velocity of the system as the pitch angle and the pitch
angular rate of the system were already known using sensors. It was shown that the
model Segway system could be suitably controlled using a reduced order observer and
-
2. Background
SON of EDGAR 25
state space control. A mass was attached to the handle bar of the system which could
be moved by remote control to create a disturbance to the system much like a rider
leaning forward on a real Segway. As anticipated the model Segway drove forward
with this mass imbalance and stopped when the mass imbalance was removed.
The RC Control of a Model Segway paper was very useful in the modelling and
understanding of the system. It was also reassuring to see state space control being
used on a similar system to SON of EDGAR that was implemented successfully.
2.2.6 JOE
Two articles have been published titled, JOE: A Mobile, Inverted Pendulum,
(Grasser et al., 2002) and JOE: A Mobile, Inverted Pendulum, (Grasser et al.,
2001). Both give an overview of the process undertaken by a team at the Industrial
Electronics Laboratory, Switzerland to design and construct a mobile, autonomous,
inverted pendulum. The final untethered prototype can be seen in Figure 2-19. The
design team envisaged a form of human transport whereby the driver is balanced on
two coaxial wheels, however, they decided to begin with a scaled down prototype
with a fixed weight replacing the human driver.
Figure 2-19: JOE undergoing untethered testing (Grasser et al., 2002).
-
2. Background
SON of EDGAR 26
This led to reduced costs and removed the risk to test pilots (Grasser et al., 2002)
whilst the simplified model eliminated many variables in terms of modelling and
controller design. The prototype, named JOE by its creators, was modelled using
modern state space theory instead of the more common classical control, as this
allowed for better control of the linear speed and turning rate of the device. A radio-
control system was implemented to give the team control over JOE during testing.
The mathematical model was simplified significantly by using a fixed weight to
simulate the human driver, eliminating many variables. A free body diagram of the
system can be seen in Figure 2-20.
Variable driver weights no longer needed to be considered; furthermore, the dynamic
loads produced by humans, continuously adjusting the overall system whilst riding
JOE, could be neglected. This simplification leads to significant differences
between the prototype and the final, full scale, rideable device. Numerous plant
changes will be introduced to the system when a human driver is used and the
prototype may not be sufficiently robust to remain stable under the dynamic
conditions.
fdP is the disturbance force on centre of gravity.
fdRL is the disturbance force on left wheel.
fdRR is the disturbance force on right wheel.
d is the disturbance angle.
CL is the Torque applied to left wheel.
CR is the Torque applied to right wheel.
is the yaw angle.
xRM is the straight line trajectory.
p is the pitch angle.
Figure 2-20: A model of JOE with state variables and disturbances (Grasser et al., 2002).
-
2. Background
SON of EDGAR 27
Given the simplified prototype, an accurate model of the device, in terms of forces,
could be created. This led to a relatively simple mathematical model that could be
used to create the State Space model of the system. The two areas of interest in terms
of control were the pitch and yaw or the device. Pitch control was crucial for the
device to remain upright, while the yaw control was needed to control the turning rate.
A single input exists in the system and that is the torque applied to the motors and
both the pitch control and yaw control require use of this input to operate effectively.
To overcome this problem the system is decoupled which allows both pitch control
and yaw control to operate independently when attempting to meet the linear speed or
turning rate commands. The decoupling of the two systems also improves the
designers ability to troubleshoot during the simulation and testing phase as two
independent systems exist instead of a single interlinked system. As pitch control is
far more critical than yaw control because it is controlling the balance of the device, it
is given a higher weighting/priority when requiring control of the motors.
A rate gyroscope was implemented to measure the angular pitch rate and integrated to
give the pitch angle. Encoders were mounted on each of the motors to measure the
speed of the vehicle. Four LEDs were used to give a visual display of the battery
voltage and would turn on in the minutes leading up to complete discharge of the
battery and flash upon the batteries reaching their minimal voltage. The onboard
controller is composed of a Sharc floating-point DSP, a XILINX field-programmable
gate array (FPGA), four 10-bit D/A converters, as well as 14 12-bit A/D converters,
(Grasser et al., 2002). A summary of JOE specifications can be seen in Table 2.1.
-
2. Background
SON of EDGAR 28
Table 2.1 Specification table of Joe.
Height 65cm
Weight 12 kg
Maximum speed 1.5m/s (5.4 km/h)
Maximum incline capable of traversing 30 degrees
Power supply 32V, 1.8Ah
Run time 1 Hour
The device successfully meets the creators design requirements of a weight being
balanced on two co-axial wheels and being controlled autonomously. The project was
well planned and thought out and, after making a few alterations, very accurate as
well as being very robust. The main weakness of the project is that it differs
significantly from the original idea of creating a device for human transport.
2.2.7 Complementary Filters
A wide variety of applications employ different sensors to measure tilt angles. A
Low-cost and Low-weight Attitude Estimation System for an Autonomous Helicopter
(Baerveldt and Klang, 1997) presents a method of combining two different tilt sensor
readings, such that the inaccuracies of each sensor are compensated for by the other.
The method utilized in this paper was a complementary filter which combined the
signals from a rate gyroscope and an inclinometer (an accelerometer) to measure the
tilt angle, or the attitude, of an autonomous helicopter.
The inclinometer was unable to accurately measure the attitude of the helicopter due
to its limited bandwidth. This meant that it could only accurately track slow variances
-
2. Background
SON of EDGAR 29
in the tilt angle. Figure 2-21 shows the inclinometer output compared to the true
orientation.
Figure 2-21: Inclinometer performance versus true orientation (Baerveldt & Klang, 1997).
The signal from the rate gyroscope was integrated to obtain an angle measurement.
This angle measurement was inaccurate at very low frequencies due to drift in the
signal created by very low frequency noise in the angular rate measurement. This drift
can be seen in Figure 2-22 which shows the gyroscopes performance compared to the
true orientation.
Figure 2-22: Rate Gyroscope performance versus true orientation (Baerveldt & Klang, 1997).
An effective attitude estimation system was then devised through the use of
complementary filters which utilised the inclinometer for low frequency signals and
-
2. Background
SON of EDGAR 30
the rate gyroscope for higher frequencies, thereby eliminating the majority of the
inaccuracies in the measured signals.
The filter transfer functions were designed based on the following equation
1)()()()( =+ sGgssHgsGisHi Equation 2-2
Where
( )sHi - is the inclinometer transfer function.
( )sHg - is the rate gyroscope transfer function.
( )sGi - is the inclinometer filter transfer function.
( )sGg - is the rate gyroscope filter transfer function.
Second order filters were then chosen to minimize the influence of offsets of the rate
gyroscope. These filters had the form
2)1(12)(
++
=s
ssGi Equation 2-3
2
2
)1()(
+=
sssGg
Equation 2-4
It was noted that the performance of the filters was increased if the dynamics of the
inclinometer were taken into account during their design. Figure 2-23 shows the
contribution of the two sensors to the filter output and Figure 2-24 shows the
complementary filter performance.
-
2. Background
SON of EDGAR 31
Figure 2-23: Contribution of the two sensors to the filtered output (Baerveldt and Klang, 1997).
Figure 2-24: Filtered output performance. The solid line is the true orientation and the dashed line is
the output of the complementary filter (Baerveldt and Klang, 1997).
From this it can be seen that the complementary filters provide an effective estimate
of the attitude of the helicopter.
-
2. Background
SON of EDGAR 32
This article is relevant to the SON of EDGAR project as the device requires the tilt
angle to be measured. This could be successfully done using an accelerometer and a
gyroscope in a similar method to that discussed within this article.
-
3. Project Goals and Component Specifications
SON of EDGAR 33
3 Project Goals and Component Specifications
The goals, primary and extension, as well as the development of the specifications of
the project are outlined in this chapter. The development of suitable goals and
specifications were crucial to the projects success as they guided both the design and
aims of the project team.
3.1 Project Goals
As part of the requirements of the project a number of goals were established to
measure the success of the project. The goals were divided into two categories,
primary and extension goals. The primary goals were defined as the goals the group
hoped to achieve as a minimum for success. The primary goals of the project are:
To develop an accurate and robust mathematical model of the system.
Convert the mathematical model into a state space plant.
Analyse state space model in MATLAB and Simulink.
Implement closed loop steering and balancing.
Design and build a physical prototype.
Create virtual reality model.
Run prototype tethered, to a computer, using state space model.
Run prototype un-tethered using on board microcontroller.
Implement a BluetoothTM communication system.
The next group of goals were defined as extension goals that were an extension of
what was hoped to be achieved by the group but were not deemed necessary for
success. They included:
-
3. Project Goals and Component Specifications
SON of EDGAR 34
Refinement of the state space model by analysing real time dynamic data.
Personalized driving condition by analysis of real time data.
Regenerative braking/energy system (dependent on the motor controller).
3.2 Specifications Development
The desired behaviour of SON of EDGAR was largely based around the functions and
short comings of the EDGAR (2005) prototype. Therefore it was not the aim of the
group to completely replicate it. An assessment of the EDGAR (2005) models
behaviour was undertaken and desired modifications of the EDGAR prototype were
debated.
Before mounting the vehicle it was desired that the rider would turn a master switch
from off to on at the back of the vehicle chassis. SON of EDGAR would power up
however the control system would not engage. Any self checks required by the
microcontroller or other components would be undertaken at this time and the
readiness of the vehicle would be indicated through an LED display to the rider. At
this stage the vehicle waits for the activation of foot sensors situated on the foot plate
of the vehicle. These foot sensors consisted of capacitive sensors as seen in the
EDGAR (2005) model. This was thought to be an effective method of implementing
an emergency stop device into SON of EDGAR rather than using a dead mans
switch as was used by the Blackwell (2005) series one model. Once one of the
capacitive sensors is active signalling the riders intention to mount the vehicle the
control system is activated.
It was desired that disembarking the vehicle would be similar to mounting the vehicle.
While under normal operation both the capacitive sensors under the feet of the rider
were to be activated signalling full capacity. At the point the rider decided to
disembark and remove a foot from the platform, disengaging a capacitive sensor, the
balancing and control of the system would switch off and remain like this until one of
-
3. Project Goals and Component Specifications
SON of EDGAR 35
the capacitive sensors is turned high again indicating the rider is re-mounting the
vehicle.
The motion of SON of EDGAR was desired to be very similar to the EDGAR (2005)
model. Forward and reverse motion is achieved by leaning forward or backward
respectively. The EDGAR (2005) model assumed a rigid link between the riders
arms and the handle bars of the vehicle; however it was desired in SON of EDGAR
that this link would act like a spring, damper system which was considered a far more
accurate representation of the system. The motion of the vehicle is instigated when an
onboard gyroscope measures a change in pitch from the rider leaning either backward
or forward on the vehicle. The wheels should thus rotate appropriately to try to keep
the wheels of the vehicle under the centre of mass of the rider and vehicle. This
motion is maintained when the rider continues to lean forward or backward.
It was desired that SON of EDGAR should initially run tethered and later as an
independent working model untethered. When running in the tethered state, it was
desired that the power supply for the SON of EDGAR would be provided through an
on board battery supply of 24 Volt. Also in the untethered state 24 Volt power
supplied from a set of rechargeable batteries was to power SON of EDGAR. A
regulated power supply to the other sensors and boards of SON of EDGAR was to be
implemented as required by each parts specification.
The steering of SON of EDGAR was to be closed loop, as desired by project
supervisor Dr. Ben Cazzolato. This closed loop system ensures that SON of
EDGARs steering will be correct during operation and, unlike the EDGAR (2005)
model, would not veer off randomly from the desired direction of travel. It was
desired that this closed loop steering would be implemented through the use of optical
encoders that had been purchased previously but not used by the EDGAR (2005)
group. The use of optical encoders also offered the capability of position control
which was seen by the group as another benefit of the optical encoders. The steering
mechanism of the vehicle was hoped to be something similar to the EDGAR (2005)
model which was thought to be a very sound design. The final specification for the
-
3. Project Goals and Component Specifications
SON of EDGAR 36
steering was that it should be velocity dependent. In other words, when SON of
EDGAR is at a stand still the turning radius of the vehicle should be zero thus
allowing it to turn on the spot. As the velocity of SON of EDGAR increased the
turning radius should increase making it safe to turn at all speeds.
3.3 Basic Component Specification
It was hoped that individual components of SON of EDGAR should meet certain
performance capabilities to ensure overall performance of the entire system. This lead
to the development of desired component specifications, outlined as follows:
Motors
High torque output.
Low backlash.
Bidirectional.
Ease of mounting.
Motor Controllers
Can withstand high current draw.
Fast communication with microcontroller.
Able to drive motors bi-directionally.
Compatible with micro-controller.
-
3. Project Goals and Component Specifications
SON of EDGAR 37
Wheels
Aesthetically pleasing.
Easy to mount.
Able to support persons weight.
Ability to withstand rough/bumpy surfaces.
Gyroscope
Compatible with the microcontroller.
Low power consumption.
High sensitivity.
Easily mounted.
Preferably low cost.
Power source
High power output.
High capacity.
High number of discharge/recharge cycles.
Low cost.
Microcontroller
Accepts compiled Simulink code.
Multiple ADC.
Multiple digital and PWM inputs/outputs.
Low power consumption.
-
3. Project Goals and Component Specifications
SON of EDGAR 38
Chassis
Support a 100kg person.
Adjustable handle bars (height and angle).
Ergonomically and aesthetically pleasing.
Fit through a standard door.
-
4. Component Selection
SON of EDGAR 39
4 Component Selection
The selection of components to be used to create SON of EDGAR was crucial in the
overall performance of the device as a self-balancing scooter. The different
components used in each of the prototypes researched was reviewed and considered.
Also completely different alternative components were analysed to see if they met the
specifications for SON of EDGAR. The advantages and disadvantages of the
possibilities were discussed and selections were made based on how well the criteria
had been met.
4.1 Motor
The motor selection for SON of EDGAR was crucial to the performance and success
of the scooter. One of the main problems with the 2005 EDGAR project was the
limited power of the motors used in the scooter (Clark et al 2005). In accordance with
these findings, more powerful motors were sought for their possible application in the
project.
Figure 4-1: Photograph of NPC-B81 (NPC Robotics, 2006).
-
4. Component Selection
SON of EDGAR 40
The motors selected to be used for SON of EDGAR were the NPC-B81 sourced from
National Power Chair (NPC). The NPC-B81 (Figure 4-1) is a four pole, 24 Volt, DC
motor originally designed for use in electric powered wheelchairs. They are capable
of producing 0.7 HP, 95 Nm stall torque and a speed of 180 RPM.
Shown in Figure 4-2 is a curve illustrating the torque characteristics of the NPC T64
which is a very similar motor to those used for the project. One of the main features of
the NPC-B81 is the right angled gearbox seen in Figure 4-1. This allows the motors to
be used in the driveline concept as described in Section 6.1. Another feature of the
NPC-B81 is the weight bearing capacity of the output shafts. The output shaft is
capable of holding approximately 137 kg (NPC Robotics, 2006) which gives a load
bearing capacity for both the motors of 274 kg. This will more than suffice for this
application and allows for direct connection to the wheels of SON of EDGAR. It must
be noted that these motors were compared to others similar for this application
however the NPC-B81 were the selected motor.
Torque (Nm) vs RPM
0
20
40
60
80
100
120
0 50 100 150 200 250
RPM
Torq
ue(N
m)
Figure 4-2: Torque characteristics of the NPC-T64.
-
4. Component Selection
SON of EDGAR 41
4.2 Gyroscope
An important requirement when creating a self-balancing scooter is the ability to
d