Bipedal Robotic System Utilizing Dynamic Balance and Adaptive Walking Gaits

thunderclingΤεχνίτη Νοημοσύνη και Ρομποτική

13 Νοε 2013 (πριν από 3 χρόνια και 7 μήνες)

94 εμφανίσεις


Bipedal Robotic System Utilizing Dynamic Balance and Adaptive Walking Gaits



Applicant:

Edward Hunckler

1829 E. Donald St.

South Bend, IN 46613

(574) 288
-
1004

ehunck@aol.com


Sponsor:

Ken Andrzejewski

Marian High School

1311 S. Logan St.

Mishawaka, IN 46544

(574) 259
-
5257

scires@marianhs.org


Proposed Project Dates:

November 2012 to February 2013

Total Expenses
: $
364
.
43

Amount Requested
: $300


Hypothesis/
Engineering Goal:
To create a
biped walking robot system that relies on dynamic
balance, motion sensors, and tactile sensors for real
-
world data in order to traverse unfamiliar
terrain.



2

Abstract


Numerous forms of motion can be

seen throughout nature. Humans

have a unique
upright bipedal walk.
In order to create robots that more efficiently interact with humans, the
robot must move with a humanlike bipedal walking gait.
The goal of this project is to create a
bipedal robot that utilizes dynamic balance and m
odels after human walking. This will be
approached by analyzing human walking gaits at different speeds and inclines and by developing
a relationship between these variables. Next a bipedal robot, modeled after human legs and torso
in dimensions and weig
ht distributions, will be created for testing and analyzing the effectiveness
of the walking gait.

Finally, stability data will be incorporated into the gait using an
accelerometer and live data analysis for automatic corrections. The robot will be able t
o traverse
at multiple speeds and inclines
. Research into robotics, in particularly the study of bipedal
walking motion, has impacts in many different regions of science. First, biped research can help
in the understanding of how humans learn to
walk. A
rtificial legs have an impact on prosthetics
that
has

both military and civilian applications. Finally, robotics is an area that is developing
quickly and spreading into everyone’s lives.








1

Introduction

Mobility is vital for any creature’s survival. All throughout nature numerous forms of
movement are seen by animals and creatures of all sizes.
Ranging

from

swimming to flying to
walking
to

many others, nature has solved the problems of
in
efficient locom
otion in creatures.
Therefore when implementing movements into robotics, modeling natural locomotion of animals
is beneficial in order to create effective mobility in the device.
Likewise, if the robot is intended
to interact primarily with humans in any

human environment, the most effective solution is a
human
-
like robot utilizing a biped walking system.
“Less than half the Earth's landmass is
accessible to wheeled and tracked vehicles, yet people and animals can go alm
ost anywhere on
Earth” (Raibert

et

al. 1).

Although accommodations can be made for wheeled robots in a human
environment, the most probabl
e

solution would be to design the robot to move in a human
-
like
manner.

Much research has gone into differences in walking patterns and effective walkin
g gaits.
In walking robotic systems, two major techniques can be

implemented for
balance during
mo
vement: dynamic balance or static balance.
These two are based on different principles for
design and function of the robot.

Static balance incorporates mo
ving through several states of
static equilibrium ens
uring that the robot is always at a state of equilibrium

(Miller 1).


When
walking with dynamic balance, the projected center of mass is allowed outside of the area
inscribed by the feet, and the walker
may be falling

during parts of the gait cycle” (Kun et al. 1).
This falling motion is what is translated into the motion because the leg at some
point

breaks this
fall
and
allows the robot to continue onward. This form of control requires greater
sophistication
in the mechanical device as well as in the programs running the robot.

A successful dynamically
balance robot has greater benefits tha
n

those of a statically balanced robot.


2

Many difficulties arise when implementing bipedal motion into a ro
bot. Several reasons
exist for this difficulty. “
The control goals (translating without falling) are not easily
decomposed in terms of the actions of individual actuators.

The system is unstable, or only
marginally stable (depending on foot design).

Time

delays in the control loop amplify stability
problems.

The system nonlinear dynamics and kinematics are difficult to model accurately, and
simplified models are generally not adequate.

Other significant properties are difficult to model
accurately (gear
stiction, gear play, foot flex, etc.).

Since the robot has no direct connection to an
inertial frame of reference, the controller must rely on often noisy sensors (foot
-
force sensors and
accelerometers, for example) to represent the relationship between t
he rob
ot and the external
environment” (Kun et al. 1). Each one of these problems compounds the reasons why so few
successful bipedal robots are built and why even fewer are put into practical applications.

Many different approaches have been
employed

t
o solve

these
difficulties in dynamically
balanced bipedal robots. In the development of a walking gait, approaches such as learning
algorithms in the robots have shown results of success based upon human learning methods for
bipedal control. Several inv
estigations have used computer models to approximate a biped robot
that physically exists in order to develop a walking gait through computer trials before ever
implementing it into the robot. Some research has
utilized neural networks in conjunction with

adaptive control to provide a walking pattern. Finally, all of these studies aim to produce the
most human
-
like and efficient gait and method for finding that pattern that will be implemented
into the biped robots.

This area of research is especially int
riguing for several reasons. The level of personal
interest is very high
because

I want to pursue a career in engineering. Also, the subject is new
and constantly changing and progressing with new ideas and technologies.
Also the applications

3

of this ar
ea of research can be very rewarding. This can include exciting new
developments

into
personalized robots as well as practical applications in the health world through the
developments

of new prosthetics. Pursuing this project is a goal for this year as
well as a
lifelong goal in the field of engineering.

Hypothesis/
Engineering Goal:
To create a biped walking robot system that relies on
dynamic balance, motion sensors, and tactile sensors for real
-
world data in order to traverse
unfamiliar terrain.



















4

Methods


The process of engineering a biped robot must be separated into portions of work that
serve as benchmarks towards a successful outcome. Each phase must be accomplished before
advancing to the next goal. F
ive
benchmark goals are in place in order
to clearly

guide the
research towards the final goal of a bipedal robot that adapts its gait to provide balance and
stabilization over various inclines in the terrain. These intermediate goals are as follows:
1)
analy
ze human walking gaits at various inclines in order to create scalable relationships between
the functions of the gaits of different inclines,
2)
create a biped robot in proportion to a human in
size and mass in order to implement and test the walking gait
s,
3)
write a program for the
computer to send to
the microcontrollers

that performs a basic walking gait, 4) update the
walking gait to change the speed of the legs based on input from the computer, and 5) finally
update the program and hardware to implem
ent sagittal balance into the robot’s walking gait
allowing it to traverse at different inclines.


To analyze the walking gait, the angles of the joints in the legs must be measured

nearly
continuously to gather data of the motion of that joint over a sing
le period in that gait. Physical
data collection can be performed by placing an angle sensor at each joint to collect data
throughout the time of one period. This angle sensor is a variable resistor, or a potentiometer
,
that

can be used to vary a
n output

voltage based on its

angle. This voltage would be re
a
d by a
microcontroller and sent to a desktop computer via serial
connection
where it can
then be
analyzed

and fit to a polynomial function
.
Each joint would be measured and interpreted
separately

for on
e period of the gait
. This process is then repeated for different speeds and
angles of incline of the walking surface.
After fitting a polynomial function to the points, t
hese

5

functions can then be analyzed to find a scalable relationship between differe
nt angles as well as
between different speeds.


The next step of the process is to build
the first pair of bipedal legs. The basic version
needs several components in order to operate
,

which are as follow
s
: a power supply, leg
actuation at the hips and kne
es, a microcontroller, and a computer to operate the functions.
The
power supply will be a
twelve
-
volt

battery with voltage regulators for the microcontrollers and
logic integrated circuits. The leg actuation at the hips and knees will require a large
am
ount of
torque. For these joints, a car windshield wiper motor will be used along with a potentiometer
for position feedback. These large motors will be powered using a circuit based around a
L298
dual motor driver. Control will be provided to these usi
ng a Propeller microcontroller
.

This
allows all of the motors to be controlled simultaneously
. A second Propeller microcontroller will
be used for

interface with the computer as well as the accelerometer. The structure of the legs
will be built using aluminum because of its
light
weight and strength. The form of the robot will
roughly follow the shape of human legs and torso
,

focusing on accuratel
y distributing the weight
across proper dimensions.


The next steps involve creating

and updating

a program on the computer to receive serial
data on the position of the accelerometer and to send updated motor signals based on the
accelerometer readings
and the corresponding scaling factors present.
Using this interaction, the
computer will read the accelerometer data and analyze it based on previous trials. Depending on
its variation

from the standard stabilities at the given time
, it will send updated
motor commands

to correct for the various instabilities in the robot because of the terrain it is encountering. With
all of these goals reached, a biped robot system with capabilities of traversing various terrains
will have successfully been created.


6

Sig
nificance


Research into
bipedal
robotics
influences a broad range of categories.
First of all, it
affects
t
he scientific understanding of human walking gaits. This includes how they function
which can lead to an understanding of
how people learn to walk
. This research also affects
prosthetic devices. Developing walking gaits and robots that can accurately perform a human
walk can lead to new designs of prosthetics that can incorporate automated leg motion allowing
new
prosthetics to be used as treatmen
ts. This can also affect the military giving them new
forms of transportation for carrying heavy loads across uneven terrains. Finally, this has an
everyday application where walking robots will one day be implemented into everyone’s daily
lives.


Educ
ationally, this project is giving me the opportunity to research an area that I have
been interested in my whole life.
Over the past summer, I built and programmed a robotic arm
plotter that utilized inverse kinematics to draw images sent to it by a
computer. This experience
made me want to pursue this field of robotics even more. After completing the bipedal robotic
legs

project,
I will have learned many new useful techniques that I can use in the future for my
career. This will be especially help
ful for me because I plan to pursue a degree in engineering in
college next year. With this project complete, I plan to enter it into the Indiana Junior Science
and Humanities Symposium, the Northern Indiana Regional Science and Engineering Fair, and
any
competitions after these.





7

Budget

Item

Quantity

Cost

Source

12VDC Wiper Motor

4

$65.96

www.monsterguts.com

Parallax Propeller

2

$15.98

www.digikey.com

Connection Socket Vertical 40 position DIP

2

$3.80

www.digikey.com

IC EEPROM 256KBIT 400KHZ 8DIP

3

$3.54

www.digikey.com

Connection IC Socket Vertical 8 position DIP

10

$1.81

www.digikey.com

Rotary Potentiometer 10K OHM

10

$7.26

www.digikey.com

L298 IC Driver Full Dual 15 Multiwatt

3

$14.01

www.digikey.com

Breakaway Connection Header 40

position

5

$9.50

www.digikey.com

Connection Receptors 2MM VERT AU 5POS

5

$7.60

www.digikey.com

Memsic 2125 Accelerometer

1

$29.99

www.digikey.com

Prop Plug (Serial to USB connection)

1

$14.99

www.digikey.com

5.000 MHZ Crystal

3

$2.43

www.digikey.com

3A Schottky

Diode

25

$8.25

www.digikey.com

3.3V Line
a
r Voltage Regulator

3

$1.80

www.digikey.com

Blank Single Sided Printed Circuit Board

2

$7.90

www.jameco.com

Jameco Value Pro Prototyping Kit

1

$46.95

www.jameco.com

Black and Decker 300 Amp portable jump
starter

1

$
49.97

www.homedepot.com

16 Conductor ribbon cable,

28AWG
, 10

Ft

1

$
6.49

www.jameco.com

Aluminum
10 feet

2

$18.00

Menards

Raspberry Pi

(Model B)

1

$35.00

www.bootic.com

Samsung
-

Raspberry
-
pi / prog
-
4gb
-
sdcard

1

$13.20

www.
bootic.
com


Other Equipment and Supplies

Soldering iron

Hack Saw

Screwdrivers

Treadmill



All of the expendable items requested are necessary for the construction of the bipedal
robotic legs. Many of the items have been specifically chosen because of their economic
efficiency such as the 12VDC Motor Wipers and
the Parallax Propeller microcontr
oller. The
microcontroller’s support circuitry was budgeted unassembled so the costs would remain low.



8

Works Cited

Kun, Andrew L., and W. Thomas Miller, III. "Adaptive Dynamic Balance of a Biped Robot
Using Neural Networks."
Robotics Laboratory, ECE
Dept., University of New
Hampshire
.

1
-
5. Print.

Miller, W. Thomas, III. "Real
-
Time Neural Network Control of a Biped Walking Robot."
IEEE

(1994): 41
-
48. Print.

Raibert, Marc, Kevin Blankespoor, Gabriel Nelson, and Rob Playter. "BigDog, the Rough
-
Terrain Qu
adruped Robot."
The International Federation of Automatic Control
.

(2008):
10822
-
0825.
Print
.












i

Appendix

Extra
m
ural Support

During the course of this project, Jim Schmiedeler
Ph.D.
at the University of Notre Dame

will be contacted for advice and consultation
.
He can provide various points of guidance
throughout

this project.