see, aibo; run! aibo motion and vision algorithms team advisor: ethan j. tira-thompson teaching...
TRANSCRIPT
![Page 1: See, AIBO; Run! AIBO Motion and Vision Algorithms Team Advisor: Ethan J. Tira-Thompson Teaching Assistant: P. Matt Jennings Team Members: Haoqian Chen,](https://reader030.vdocuments.mx/reader030/viewer/2022032723/56649d0e5503460f949e3e17/html5/thumbnails/1.jpg)
See, AIBO; Run!AIBO Motion and Vision Algorithms
Team Advisor: Ethan J. Tira-ThompsonTeaching Assistant: P. Matt Jennings
Team Members: Haoqian Chen, Elena Glassman, Chengjou Liao,
Yantian Martin, Lisa Shank, Jon Stahlman
![Page 2: See, AIBO; Run! AIBO Motion and Vision Algorithms Team Advisor: Ethan J. Tira-Thompson Teaching Assistant: P. Matt Jennings Team Members: Haoqian Chen,](https://reader030.vdocuments.mx/reader030/viewer/2022032723/56649d0e5503460f949e3e17/html5/thumbnails/2.jpg)
AIBO:
The Legend.The Dog…
![Page 3: See, AIBO; Run! AIBO Motion and Vision Algorithms Team Advisor: Ethan J. Tira-Thompson Teaching Assistant: P. Matt Jennings Team Members: Haoqian Chen,](https://reader030.vdocuments.mx/reader030/viewer/2022032723/56649d0e5503460f949e3e17/html5/thumbnails/3.jpg)
History of AIBO
• Sony’s Entertainment Robot
• AIBO – Artificially Intelligent roBOt
• Or aibo – Japanese for “companion”
• Originally intended for purchase by home users
• Found to be a relatively cheap, versatile platform that could be used by educators and researchers
![Page 4: See, AIBO; Run! AIBO Motion and Vision Algorithms Team Advisor: Ethan J. Tira-Thompson Teaching Assistant: P. Matt Jennings Team Members: Haoqian Chen,](https://reader030.vdocuments.mx/reader030/viewer/2022032723/56649d0e5503460f949e3e17/html5/thumbnails/4.jpg)
AIBO: Technical Specifications• 64 bit Processor• 20 Degrees of
Freedom• Microphone• Accelerometer• Infrared Distance
Sensor• Pressure Sensors• The Kitchen Sink
![Page 5: See, AIBO; Run! AIBO Motion and Vision Algorithms Team Advisor: Ethan J. Tira-Thompson Teaching Assistant: P. Matt Jennings Team Members: Haoqian Chen,](https://reader030.vdocuments.mx/reader030/viewer/2022032723/56649d0e5503460f949e3e17/html5/thumbnails/5.jpg)
Tekkotsu• Application framework for AIBO
• Under development at CMU’s Robotics Lab
• TekkotsuMon - server-side interface to code running on robot
• To accomplish our goals, we built upon Tekkotsu platform
• Our advisor, Ethan Tira-Thompson, is a chief researcher for the project
![Page 6: See, AIBO; Run! AIBO Motion and Vision Algorithms Team Advisor: Ethan J. Tira-Thompson Teaching Assistant: P. Matt Jennings Team Members: Haoqian Chen,](https://reader030.vdocuments.mx/reader030/viewer/2022032723/56649d0e5503460f949e3e17/html5/thumbnails/6.jpg)
Run, AIBO, Run!
• Originally attempted to stabilize image from AIBO’s camera– Software methods– New walking motions
• Modified walk parameters
• Measured performance using accelerometers
• Stability vs. Speed
![Page 7: See, AIBO; Run! AIBO Motion and Vision Algorithms Team Advisor: Ethan J. Tira-Thompson Teaching Assistant: P. Matt Jennings Team Members: Haoqian Chen,](https://reader030.vdocuments.mx/reader030/viewer/2022032723/56649d0e5503460f949e3e17/html5/thumbnails/7.jpg)
How Our Walk Was Developed
• Tekkotsu’s default walk is taken from CMU’s RoboCup soccer team
• Our new motion was created by modifying this walk’s parameters
![Page 8: See, AIBO; Run! AIBO Motion and Vision Algorithms Team Advisor: Ethan J. Tira-Thompson Teaching Assistant: P. Matt Jennings Team Members: Haoqian Chen,](https://reader030.vdocuments.mx/reader030/viewer/2022032723/56649d0e5503460f949e3e17/html5/thumbnails/8.jpg)
Get Your Move On!• Using the
TekkotsuMon GUI
![Page 9: See, AIBO; Run! AIBO Motion and Vision Algorithms Team Advisor: Ethan J. Tira-Thompson Teaching Assistant: P. Matt Jennings Team Members: Haoqian Chen,](https://reader030.vdocuments.mx/reader030/viewer/2022032723/56649d0e5503460f949e3e17/html5/thumbnails/9.jpg)
Changing Things Up:The Parameters
• Lift Velocity• Down velocity• Lift Time• Down time• Body Height/Angle• Period• Position Coordinates• Hop and Sway
![Page 10: See, AIBO; Run! AIBO Motion and Vision Algorithms Team Advisor: Ethan J. Tira-Thompson Teaching Assistant: P. Matt Jennings Team Members: Haoqian Chen,](https://reader030.vdocuments.mx/reader030/viewer/2022032723/56649d0e5503460f949e3e17/html5/thumbnails/10.jpg)
Accelerometers
• Each consists of a mass/spring system
• Measure force and displacement on joints as robot walks
• We graphed the way force varies with time to evaluate stability of new walking algorithm
![Page 11: See, AIBO; Run! AIBO Motion and Vision Algorithms Team Advisor: Ethan J. Tira-Thompson Teaching Assistant: P. Matt Jennings Team Members: Haoqian Chen,](https://reader030.vdocuments.mx/reader030/viewer/2022032723/56649d0e5503460f949e3e17/html5/thumbnails/11.jpg)
Our Upright Walk vs.
CMU RoboCup Walk
• RoboCup Square Sum of Acceleration vs. Time
-5
0
5
10
15
20
25
0 2000 4000 6000 8000 10000 12000
Time (msec)
Sq
uar
e S
um
of
Acc
eler
atio
n
(mm
^2/
mse
c^4)
square sum of acceleration
• Upright
Square Sum of Acceleration vs. Time
0
5
10
15
20
25
30
35
0 2000 4000 6000 8000 10000 12000
Time (msec)S
qu
are
Su
m o
f A
ccel
erat
ion
(m
m^
2/m
sec^
4)Square Sum of Acceleration
Quarter Speed
![Page 12: See, AIBO; Run! AIBO Motion and Vision Algorithms Team Advisor: Ethan J. Tira-Thompson Teaching Assistant: P. Matt Jennings Team Members: Haoqian Chen,](https://reader030.vdocuments.mx/reader030/viewer/2022032723/56649d0e5503460f949e3e17/html5/thumbnails/12.jpg)
Our Upright Walk vs.
CMU RoboCup Walk• RoboCup
Square Sum of Acceleration vs. Time
-5
0
5
10
15
20
25
30
35
0 2000 4000 6000 8000 10000 12000
Time (msec)
Sq
uar
e S
um
of
Acc
eler
atio
n
(mm
^2/
mse
c^4)
Square Sum of Acceleration
• Upright
Square Sum of Acceleration vs. Time
0
5
10
15
20
25
30
35
0 2000 4000 6000 8000 10000 12000
Time (msec)
Sq
uar
e S
um
of
Acc
eler
atio
n
(mm
^2/
mse
c^4)
Square Sum of Acceleration
Half Speed
![Page 13: See, AIBO; Run! AIBO Motion and Vision Algorithms Team Advisor: Ethan J. Tira-Thompson Teaching Assistant: P. Matt Jennings Team Members: Haoqian Chen,](https://reader030.vdocuments.mx/reader030/viewer/2022032723/56649d0e5503460f949e3e17/html5/thumbnails/13.jpg)
Our Upright Walk vs.
CMU RoboCup Walk
• RoboCup
Square Sum of Acceleration vs Time
-20
0
20
40
60
80
100
120
0 2000 4000 6000 8000 10000 12000
Time (msec)
Sq
ua
re S
um
of
Ac
ce
lera
tio
n
(mm
^2
/ms
ec
^4
)
Square Sum of Acceleration
• Upright
Square Sum of Acceleration vs. Time
0
5
10
15
20
25
30
35
40
45
0 2000 4000 6000 8000 10000 12000
Tim e (m sec)
Sq
uar
e S
um
of
Acc
eler
atio
n (
mm
^2/m
sec^
4)
Square Sum of Acceleration
Three Quarter Speed
![Page 14: See, AIBO; Run! AIBO Motion and Vision Algorithms Team Advisor: Ethan J. Tira-Thompson Teaching Assistant: P. Matt Jennings Team Members: Haoqian Chen,](https://reader030.vdocuments.mx/reader030/viewer/2022032723/56649d0e5503460f949e3e17/html5/thumbnails/14.jpg)
Our Upright Walk vs.
CMU RoboCup Walk
• RoboCup
Square Sum of Acceleration vs Time
-20
0
20
40
60
80
100
120
140
0 2000 4000 6000 8000 10000 12000
Time (msec)
Sq
ua
re S
um
of
Ac
ce
lera
tio
n
(mm
^2
/ms
ec
^4
)
Square Sum of Acceleration
• Upright
Square Sum of Acceleration vs. Time
-10
0
10
20
30
40
50
60
70
80
0 2000 4000 6000 8000 10000 12000
Time (msec)
Sq
ua
re S
um
of
Ac
ce
lera
tio
n (
mm
^2
/ms
ec
^4
)
Square Sum ofAcceleration
Full Speed
![Page 15: See, AIBO; Run! AIBO Motion and Vision Algorithms Team Advisor: Ethan J. Tira-Thompson Teaching Assistant: P. Matt Jennings Team Members: Haoqian Chen,](https://reader030.vdocuments.mx/reader030/viewer/2022032723/56649d0e5503460f949e3e17/html5/thumbnails/15.jpg)
See, AIBO, See!
• Developed an algorithm that allows AIBO to follow a pink line
• Gradually improved algorithm based on perceived weaknesses
![Page 16: See, AIBO; Run! AIBO Motion and Vision Algorithms Team Advisor: Ethan J. Tira-Thompson Teaching Assistant: P. Matt Jennings Team Members: Haoqian Chen,](https://reader030.vdocuments.mx/reader030/viewer/2022032723/56649d0e5503460f949e3e17/html5/thumbnails/16.jpg)
How does AIBO see?
CCD Camera
(YUV)
Raw Image Translated
on Computer
(RGB)
AIBO
![Page 17: See, AIBO; Run! AIBO Motion and Vision Algorithms Team Advisor: Ethan J. Tira-Thompson Teaching Assistant: P. Matt Jennings Team Members: Haoqian Chen,](https://reader030.vdocuments.mx/reader030/viewer/2022032723/56649d0e5503460f949e3e17/html5/thumbnails/17.jpg)
Segmented Vision
• AIBO camera captures YUV format
• Bitmapped images are too large to send over network efficiently, so the images must be compressed
![Page 18: See, AIBO; Run! AIBO Motion and Vision Algorithms Team Advisor: Ethan J. Tira-Thompson Teaching Assistant: P. Matt Jennings Team Members: Haoqian Chen,](https://reader030.vdocuments.mx/reader030/viewer/2022032723/56649d0e5503460f949e3e17/html5/thumbnails/18.jpg)
RLE (Run-Length Encoding)
• The data is then converted into a series of color run triplets
• Triplets are sent to the computer, where they are placed into an array
• Vision segmentation and run-length encoding is performed on board the AIBO
![Page 19: See, AIBO; Run! AIBO Motion and Vision Algorithms Team Advisor: Ethan J. Tira-Thompson Teaching Assistant: P. Matt Jennings Team Members: Haoqian Chen,](https://reader030.vdocuments.mx/reader030/viewer/2022032723/56649d0e5503460f949e3e17/html5/thumbnails/19.jpg)
Line Following Algorithms
Hough Transformation
VS.
Hack Algorithm
![Page 20: See, AIBO; Run! AIBO Motion and Vision Algorithms Team Advisor: Ethan J. Tira-Thompson Teaching Assistant: P. Matt Jennings Team Members: Haoqian Chen,](https://reader030.vdocuments.mx/reader030/viewer/2022032723/56649d0e5503460f949e3e17/html5/thumbnails/20.jpg)
Hough Transformation
• Represent a line in an image in a different way: Parameter Space
• Example – a line in image space can be represented as:
y = mx + y0
• The Hough method “transforms” this equation to parameter space:
y0 = y - mx
![Page 21: See, AIBO; Run! AIBO Motion and Vision Algorithms Team Advisor: Ethan J. Tira-Thompson Teaching Assistant: P. Matt Jennings Team Members: Haoqian Chen,](https://reader030.vdocuments.mx/reader030/viewer/2022032723/56649d0e5503460f949e3e17/html5/thumbnails/21.jpg)
Hough Transformation (cont’d)• A discrete parameter space called an accumulator is
created. All points in the original image are converted using Hough Transform into this parameter space.
• Points with the same slope and y-intercept are accumulated into the same cell of the accumulator.
• The highest cell is found, thus finding the most prominent line (marked in red).
![Page 22: See, AIBO; Run! AIBO Motion and Vision Algorithms Team Advisor: Ethan J. Tira-Thompson Teaching Assistant: P. Matt Jennings Team Members: Haoqian Chen,](https://reader030.vdocuments.mx/reader030/viewer/2022032723/56649d0e5503460f949e3e17/html5/thumbnails/22.jpg)
Hough Transformation (cont’d)
• Advantages:– Will almost always find the most
prominent line– Ignores static and foreign objects (unless
they have defined edges)
• Disadvantages:– Implementation in a robot is too computationally
expensive– Processes too slow for a real-time image, like the
AIBO’s
![Page 23: See, AIBO; Run! AIBO Motion and Vision Algorithms Team Advisor: Ethan J. Tira-Thompson Teaching Assistant: P. Matt Jennings Team Members: Haoqian Chen,](https://reader030.vdocuments.mx/reader030/viewer/2022032723/56649d0e5503460f949e3e17/html5/thumbnails/23.jpg)
A Hack Gains Greatness
• A hack is defined as an inelegant and usually temporary solution to a problem
• Ironically, the hack line following algorithm we developed (with the help of Alok Ladsariya) became our best solution.
![Page 24: See, AIBO; Run! AIBO Motion and Vision Algorithms Team Advisor: Ethan J. Tira-Thompson Teaching Assistant: P. Matt Jennings Team Members: Haoqian Chen,](https://reader030.vdocuments.mx/reader030/viewer/2022032723/56649d0e5503460f949e3e17/html5/thumbnails/24.jpg)
The Basic Line Following Algorithm
• Take in decoded segmented vision
• Create RegionMap– First each pixel is set
to pink or not-pink– Give each pink
region a unique number… remember PaintBucket?
• Turn toward largest region on horizontal center line
![Page 25: See, AIBO; Run! AIBO Motion and Vision Algorithms Team Advisor: Ethan J. Tira-Thompson Teaching Assistant: P. Matt Jennings Team Members: Haoqian Chen,](https://reader030.vdocuments.mx/reader030/viewer/2022032723/56649d0e5503460f949e3e17/html5/thumbnails/25.jpg)
The Basic Line-Following Algorithm
• Primary target – displayed as blue dot– at intersection of largest region & center row
• Direction adjustment– Turns L if blue dot to L of center column– Turns R if blue dot to R of center column
• A few fundamental assumptions allow us to keep the algorithm simple– The line will be the largest region– The line will cross the center row once
![Page 26: See, AIBO; Run! AIBO Motion and Vision Algorithms Team Advisor: Ethan J. Tira-Thompson Teaching Assistant: P. Matt Jennings Team Members: Haoqian Chen,](https://reader030.vdocuments.mx/reader030/viewer/2022032723/56649d0e5503460f949e3e17/html5/thumbnails/26.jpg)
Improvements
Definition: Lost = no regions exist in center row
• If lost stop forward motion & rotate
determine rotation direction by which side of image contained dot more commonly
• In last 15 frames
• Adjust direction with speed proportional to blue dot’s distance from center column
![Page 27: See, AIBO; Run! AIBO Motion and Vision Algorithms Team Advisor: Ethan J. Tira-Thompson Teaching Assistant: P. Matt Jennings Team Members: Haoqian Chen,](https://reader030.vdocuments.mx/reader030/viewer/2022032723/56649d0e5503460f949e3e17/html5/thumbnails/27.jpg)
Algorithm Issues
![Page 28: See, AIBO; Run! AIBO Motion and Vision Algorithms Team Advisor: Ethan J. Tira-Thompson Teaching Assistant: P. Matt Jennings Team Members: Haoqian Chen,](https://reader030.vdocuments.mx/reader030/viewer/2022032723/56649d0e5503460f949e3e17/html5/thumbnails/28.jpg)
Algorithm Issues• A Dashed Line
– The breaks in the line trigger the ‘lost’ behavior briefly
– But even so, the break is not long enough to lose the dog completely.
Walking along happily…
AHH!! BREAK IN THE LINE!!
Found line, walking along happily…
![Page 29: See, AIBO; Run! AIBO Motion and Vision Algorithms Team Advisor: Ethan J. Tira-Thompson Teaching Assistant: P. Matt Jennings Team Members: Haoqian Chen,](https://reader030.vdocuments.mx/reader030/viewer/2022032723/56649d0e5503460f949e3e17/html5/thumbnails/29.jpg)
Algorithm Issues• Branching lines:
– branch recognized as one region– the average taken of all x values of
region’s pixels in center row– once the shorter branch gets below center
row, target jumps to longer branch (line)
![Page 30: See, AIBO; Run! AIBO Motion and Vision Algorithms Team Advisor: Ethan J. Tira-Thompson Teaching Assistant: P. Matt Jennings Team Members: Haoqian Chen,](https://reader030.vdocuments.mx/reader030/viewer/2022032723/56649d0e5503460f949e3e17/html5/thumbnails/30.jpg)
Algorithm Issues• Dog can lose line when
– Special case: line slopes toward center of image
• Slope-Opposite-Direction Method– Two more blue dots
• Above and below center
– used to find slope of the line– If special case true:
• Compromise and go straight!
![Page 31: See, AIBO; Run! AIBO Motion and Vision Algorithms Team Advisor: Ethan J. Tira-Thompson Teaching Assistant: P. Matt Jennings Team Members: Haoqian Chen,](https://reader030.vdocuments.mx/reader030/viewer/2022032723/56649d0e5503460f949e3e17/html5/thumbnails/31.jpg)
Unsolved Algorithm Issues
• Extremely sharp curves (< 90 deg.)
• Similar to branching issue (but more troublesome)
• Current algorithm averages X values• Heads for center between two intersections of
line with center row
![Page 32: See, AIBO; Run! AIBO Motion and Vision Algorithms Team Advisor: Ethan J. Tira-Thompson Teaching Assistant: P. Matt Jennings Team Members: Haoqian Chen,](https://reader030.vdocuments.mx/reader030/viewer/2022032723/56649d0e5503460f949e3e17/html5/thumbnails/32.jpg)
Problem: Follow the Line!
• …that means don’t follow the square.
• A possible solution: Shape Recognition– Feature Extraction– Describe regions
FAILURE!
![Page 33: See, AIBO; Run! AIBO Motion and Vision Algorithms Team Advisor: Ethan J. Tira-Thompson Teaching Assistant: P. Matt Jennings Team Members: Haoqian Chen,](https://reader030.vdocuments.mx/reader030/viewer/2022032723/56649d0e5503460f949e3e17/html5/thumbnails/33.jpg)
In the Future: Feature Extraction
• Features easy to measure:– Area– Perimeter– Granularity
Measurement
• And how can their measurements be combined decision?
• Hypothesis: AIBO should choose– Largest– Most elongated– Least grainy
region in image.
![Page 34: See, AIBO; Run! AIBO Motion and Vision Algorithms Team Advisor: Ethan J. Tira-Thompson Teaching Assistant: P. Matt Jennings Team Members: Haoqian Chen,](https://reader030.vdocuments.mx/reader030/viewer/2022032723/56649d0e5503460f949e3e17/html5/thumbnails/34.jpg)
Elongation
• Ratio of Perimeter to Area of region
• In computer vision, called “Compactness”
• C = (P^2) / A
• Circle = most compact• Way to differentiate
betweenmore compact shapes
and the line.
![Page 35: See, AIBO; Run! AIBO Motion and Vision Algorithms Team Advisor: Ethan J. Tira-Thompson Teaching Assistant: P. Matt Jennings Team Members: Haoqian Chen,](https://reader030.vdocuments.mx/reader030/viewer/2022032723/56649d0e5503460f949e3e17/html5/thumbnails/35.jpg)
Granularity Measurement
• Purpose: distinguish between – clear lines and grainy
regions
• From each pixel at the boundary of a region– a ray extends until
another pink pixel is encountered
• Granularity Measurement = sum of all rays’ lengths
![Page 36: See, AIBO; Run! AIBO Motion and Vision Algorithms Team Advisor: Ethan J. Tira-Thompson Teaching Assistant: P. Matt Jennings Team Members: Haoqian Chen,](https://reader030.vdocuments.mx/reader030/viewer/2022032723/56649d0e5503460f949e3e17/html5/thumbnails/36.jpg)
Standardization
• To pick the best region, one can combine the three measurements
• Largeness, Compactness, & Grainy Measure
• All have their own scales standardization necessary
• Relevant measurement for a region relative to all other regions. www.neiu.edu/~lruecker/ smrm.htm
•How different is this point?
![Page 37: See, AIBO; Run! AIBO Motion and Vision Algorithms Team Advisor: Ethan J. Tira-Thompson Teaching Assistant: P. Matt Jennings Team Members: Haoqian Chen,](https://reader030.vdocuments.mx/reader030/viewer/2022032723/56649d0e5503460f949e3e17/html5/thumbnails/37.jpg)
Performance vs. Complexity
![Page 38: See, AIBO; Run! AIBO Motion and Vision Algorithms Team Advisor: Ethan J. Tira-Thompson Teaching Assistant: P. Matt Jennings Team Members: Haoqian Chen,](https://reader030.vdocuments.mx/reader030/viewer/2022032723/56649d0e5503460f949e3e17/html5/thumbnails/38.jpg)
LiveDemo
![Page 39: See, AIBO; Run! AIBO Motion and Vision Algorithms Team Advisor: Ethan J. Tira-Thompson Teaching Assistant: P. Matt Jennings Team Members: Haoqian Chen,](https://reader030.vdocuments.mx/reader030/viewer/2022032723/56649d0e5503460f949e3e17/html5/thumbnails/39.jpg)
How Different WalksAffect Line-Following
• Low velocity vs. high velocity
• Center of gravity affects slip potential
• Camera level affects:– Ease of getting lost– Precision of line following
![Page 40: See, AIBO; Run! AIBO Motion and Vision Algorithms Team Advisor: Ethan J. Tira-Thompson Teaching Assistant: P. Matt Jennings Team Members: Haoqian Chen,](https://reader030.vdocuments.mx/reader030/viewer/2022032723/56649d0e5503460f949e3e17/html5/thumbnails/40.jpg)
Special Thanks to:• Ethan• Alok Ladsariya• Dr. David Touretzky• Greg Kesden • Sony• CMU School of
Computer Science and Robotics Institute
• Manuela Veloso & CMU RoboCup
• PMatt
![Page 41: See, AIBO; Run! AIBO Motion and Vision Algorithms Team Advisor: Ethan J. Tira-Thompson Teaching Assistant: P. Matt Jennings Team Members: Haoqian Chen,](https://reader030.vdocuments.mx/reader030/viewer/2022032723/56649d0e5503460f949e3e17/html5/thumbnails/41.jpg)
Bye!
![Page 42: See, AIBO; Run! AIBO Motion and Vision Algorithms Team Advisor: Ethan J. Tira-Thompson Teaching Assistant: P. Matt Jennings Team Members: Haoqian Chen,](https://reader030.vdocuments.mx/reader030/viewer/2022032723/56649d0e5503460f949e3e17/html5/thumbnails/42.jpg)
Image Sourceswww.dai.ed.ac.uk/HIPR2/hough.htmwww.dai.ed.ac.uk/CVonline/LOCAL_COPIES/MARBLE/medium/contours/feature.htmwww.physik.uni-osnabrueck.de/nonlinop/Hough/LineHough.htmlwww.ri.cmu.edu/labs/lab_60.htmlwww.nynatur.dk/artikler/artikel_tekster/robo_pets.htmlwww003.upp.so_net.ne.jp/studio_mm/kameo/kameo.htmlwww.opus.co.jp/products/aibo/www.generation5.org/aibo.shtmlwww.ohta-kyoko.com/aibo/aibo.htmlwww.vdelnevo.co.uk/www.d2.dion.ne.jp/~narumifu/diary.htmlwww.21stcentury.co.uk/robotics/aibo.aspdigitalcamera.gr.jp/html/HotNews/backno/HotNews001001-31.htwww.seaple.icc.ne.jp/~somari/aibo.htmwww-2.cs.cmu.edu/afs/cs/project/robosoccer/www/legged/legged-team.html#publicationsccrma-www.stanford.edu/CCRMA/Courses/252/sensors/node9.htmlhttp://www.icaen.uiowa.edu/~dip/LECTURE/Shape3.html#scalar