Today in CS 154

1
Today in CS 154

• Notes hw 1 and the project websites
• Short assignment 2 due next Wednesday, Feb. 13

Getting moving with your project
move / sense / get pictures or a movie
Some short problems on motors/sensors/kinematics
All goes to updating your project's web archive

• Interacting with the world actuators and sensors

animals don't have wheels
2
Grading
The GOR system Generalized O'Neill
Rubric A represents outstanding work B
represents solid work C dialed it in D what
happened? F you didn't show up
good things to have
updated website progress pictures and
movies errors and problems!
flexibility?
mapping onto publication possibilities
A ICRA '08, IROS '07, AAAI, SIGCSE, (RSS) B
Robotics and Automation, FLAIRS, workshops, C
SCI, WMSCI and siblings D F
nope
3
Lab Write-ups

• Web-based write up
• with an eye toward other possibilities, e.g.,
  conference paper
• cite all of the papers weve read up to those
  weeks (challenge!)
• link to raw data and code, as well

• Conference papers rough outline
• Introduction what are you doing? (citations)
• Background narrow/better define the problem
  (citations)
• Approach what were your key decisions (and
  difficulties) ?
• Progress/Performance what are your results ?
  pictures/movies
• Perspective what conclusions can you draw and
  what else might be done ?
• Bibliography contrasting with related previous
  work
• Media fun ?

there is a detailed outline linked from HW 2
Experiments in Automatic Flock Control,
R.Vaughan, N. Sumpter, A. Frost, and S. Cameron
4
Example Write-ups
5
CS 154 Topic Outline
In general, how should we design autonomously
reasoning systems?

• Low-level robotics
• architecture
• motors/actuators
• sensors
• probabilistic modeling
• Spatial Reasoning
• reasoning with uncertainty
• filtering and state estimation
• localization
• mapping
• landmarks and vision
• Vision
• Spatial Planning
• configuration space
• kinematics, dynamics
• path planning
• Additional Topics
• navigation with uncertainty
• active spatial reasoning

3wks
Robots have bodies
they want to know where they are
5wks
2wks
taking images can help
they need to choose where to go
3wks
1wk
and may have to be flexible about how and when
they get there.
6
Robot Architecture
how much / how do we represent the world
internally ?
As much as possible!
SPA paradigm
Not at all
Reactive paradigm
Task-specific
Motor Schemas
different ways of composing behaviors
Behavior-based architecture
Subsumption/State machine
As much as possible.
Hybrid approaches
7
Polly (Horswill, 1992-5)
The Polly System specializing perception to a
single task/environment
a vision-based tour guide at MIT
subsumption a layered state machine
Polly tested the limits of reactivity.
8
Rover, the Robot Sheepdog
Richard Vaughan
Oxford USC (HMC) HRL Simon Fraser
Modeling animal intelligence in more than one way
a potential field application
9
Ethological Inspiration
Toads (Arbib 87)
Sheep (and ducks) (Vaughan 95)
vowels and elevators
10
Rover set-up
Overhead camera Circular pen Easily extracted
robot and ducks! - all developed in simulation
first
11
Simulation
simulation run of the system the goal is to get
the flock to the bottom of the pen
12
Simulation results
Results
13
Quiz p. 1
Describe (in English) what the two terms in the
controller below are saying about Robot Rs
behavior in relation to flock F .
the hat is a unit vector
the robot equation
robot Rs movement
14
Quiz p. 2
Describe (in English) what each of the four terms
below are saying about duck Ds behavior in
relation to duck Dn and the Robot R .
Are there any errors here?
the hat is a unit vector
the duck equation
duck Ds movement
15
Quiz p. 1
Describe (in English) what the two terms in the
controller below are saying about Robot Rs
behavior in relation to flock F .
the hat is a unit vector
the robot equation
robot Rs movement
toward flock proportional to the flocks
lostness
away from goal
16
Quiz p. 2
Describe (in English) what each of the four terms
below are saying about duck Ds behavior in
relation to duck Dn and the Robot R .
the hat is a unit vector
the duck equation
duck psychology
?
Im close to my friends
but not too close!
Im claustrophobic
Whats that thing?
duck Ds movement
17
Quiz p. 2
Describe (in English) what each of the four terms
below are saying about duck Ds behavior in
relation to duck Dn and the Robot R .
in the journal version
Oops!
Journal of Autonomous and Robotic Systems
18
The cameras-eye view of Rover
19
Real results!
20
Real results!
Robot and flock paths
21
Robot Architecture
how much / how do we represent the world
internally ?
As much as possible!
SPA paradigm
Not at all
Reactive paradigm
Task-specific
Motor Schemas
different ways of composing behaviors
Behavior-based architecture
Subsumption paradigm
As much as possible.
rest of the course
Hybrid approaches
22
Homework 2

• Project "Getting moving"
• build/familiarize yourself with the platform
• be able to write and run programs move around
• be able to access and print sensor values

thinking of using the Wiimote for a robotics
project ?
23
Robot of the day WiiBot

• A sword-wielding, tennis-playing industrial arm!

What resulted was an algorithm described by
probably the loosest use of the term pattern
recognition that isnt punishable by instant
banishment to programmer Hell.
KR-16 by Kuka
http//www.usmechatronics.com/usmgarage/WiiBot.htm
l
24
CS 154 Topic Outline

• Low-level robotics
• architecture
• motors/actuators
• sensors
• probabilistic modeling
• Spatial Reasoning
• reasoning with uncertainty
• filtering and state estimation
• localization
• mapping
• landmarks and vision
• Vision
• regions and recognition
• feature extraction
• 3d reconstruction
• Spatial Planning
• configuration space
• kinematics, dynamics
• path planning

How are we getting around? And sensing?
3wks
What am I? robots bodies
5wks
Where am I?
3wks
Is seeing believing?
3wks
How do I get there?
25
Robot Actuation Motors
Stepper motors
Servo motors
DC motors
Physics principles
Things seek lowest energy states.
Nature is lazy.

• iron core vs. magnet

N
S

• magnetic fields tend to line up

Electric fields and magnetic fields are the same
thing.
Torque is a good scrabble word.
26
Stepper Motors
variable reluctance
S
stator
rotor
N
electromagnets
27
Stepper Motors
S
stator
rotor
N
aligns the rotor
electromagnets
For this set of energized teeth, how does rotor
angle affect the torque?
28
Stepper Motors
S
stator
rotor
N
aligns the rotor
electromagnets
torque
angle
29
Stepper Motors
S
stator
N
S
rotor
N
on to the next teeth
aligns the rotor
electromagnets
printers computer drives machining

• Direct control of rotor position (no sensing
  needed)
• Strongest when stationary!
• Low resolution (so far)

can we increase our resolution?
30
Increasing Resolution
More teeth
on the rotor and/or stator
How many full steps per revolution does this
motor permit (R6, S4)?
motors ! position next time
31
Increasing Resolution
Hw 2 question
torque
S
N
S
angle
More teeth
N
on the rotor and/or stator
Half-stepping
energizing more than one pair of stator teeth
How many full steps per revolution does this
motor permit (R6, S4)?
motors ! position next time
32
(No Transcript)
33
Controlling Stepper Motors
http//www.cs.uiowa.edu/jones/step/types.html
S 6, R 4
variable-reluctance stepper motor (rotor is not
magnetized)
34
Controlling Stepper Motors
http//www.cs.uiowa.edu/jones/step/types.html
S 6, R 4
variable-reluctance stepper motor (rotor is not
magnetized)
S 4, R 6
variable-reluctance stepper motor (rotor is
magnetized)
tradeoff is in cost and control complexity
going to one extreme of this tradeoff
35
Spherical Stepper Motor !
http//www.me.jhu.edu/r_kine.html
complete motor
stator
half of the rotor
G. Chirikjian David Stein, JHU
applications?
36
Spherical Stepper Motor !
http//www.me.jhu.edu/r_kine.html
applications envisioned
complete motor
some are a bit scary
37
DC motors -- exposed !
or DC servo motor...
One of the problems involve the gear trains of a
DC servo motor and a Roomba wheel
38
DC motor basics
permanent magnets
N
S
rotor
stator
brushes

V
commutator on the rotor
-
39
DC motor basics
permanent magnets
N
S
rotor
N
S
stator
brushes


V
V
commutator on shaft
-
-
40
DC motor basics
permanent magnets
smoother with more rotor teeth
N
S
rotor
N
S
N
S
stator
brushes



V
V
V
commutator on shaft
-
-
-
41
Who pulls more weight?
electro-magnets
stator
S
rotor
N
Stepper motor
DC motor

• Energy used is prop. to speed
• Higher torque at faster speeds
• More popular, so theyre cheaper
• Smoother at low speeds
• DO need to sense axle position!

• Position control
• High holding torque
• Durability (no brushes)
• Don't need to sense axle position!

42
Proprioceptive Sensing
A 40 word yours for only 1.94!
How might we measure the angle of our motor's
axle?

• physical sensing
• electric fields
• variable resistance
• optical sensing

43
Physically counting rotations
You can also use the touch sensor as a rotation
sensor. Attach a cam to the rotating axle, and
then position the LEGO sensor so that it is hit
by the cam as the axle rotates. Counting the
number of hits, combined with the wheel radius,
can give you a distance reading. Measuring the
time between hits can give you a speed
measurement. Note that this only works well for
axles that don't rotate very fast.
llk.media.mit.edu/projects/cricket/doc/LEGO-touchs
ensors.html
44
Measuring axle orientation

• Resolver
• measures absolute shaft orientation

via electric fields

• Potentiometer
• measures orientation by varying resistance, it
  has a range of motion lt 360º

via resistance
power/contact
45
Servomotors
Hw 2 question
potentiometer
Direct position control in response to the width
of a regularly sent pulse. A potentiometer is
used to determine the motor shaft angle.
modified to run continuously
46
Optical Encoders in Lego!
from the FIRST Lego league
47
Optical Encoders 1 absolute

• Detecting motor shaft orientation

Binary encoding of shaft rotation
via light patterns
potential problems?
48
Optical Encoders 1 absolute

• Detecting motor shaft orientation

Alternative encodings are also popular !
49
Gray Code

Binary
0 1 2 3 4 5 6 7 8
9
0 1 10 11 100 101 110 111 1000 1001
000 001 011 010 110 111 101 100 ? ?
What is important about each of these transitions?
what comes next?
50
Gray Code

Binary
0 1 2 3 4 5 6 7 8
9
0 1 10 11 100 101 110 111 1000 1001
000 001 011 010 110 111 101 100 1100 1101
neighboring representations differ by only 1 bit
51
Tradeoffs?

Binary
Gray
0 1 2 3 4 5 6 7 8
9
0 1 10 11 100 101 110 111 1000 1001
000 001 011 010 110 111 101 100 1100 1101
other applications?
52
Tradeoffs?

Binary
Gray
0 1 2 3 4 5 6 7 8
9
0 1 10 11 100 101 110 111 1000 1001
000 001 011 010 110 111 101 100 1100 1101
with important applications
wires?
53
Tradeoffs?

Binary
Gray
0 1 2 3 4 5 6 7 8
9
0 1 10 11 100 101 110 111 1000 1001
000 001 011 010 110 111 101 100 1100 1101
with important applications
not found even in The Official Duke Nukem 3d
Strategies and Secrets (1.94)
wires?
54
Absolute Optical Encoders

• Complexity of distinguishing many different
  states -- high resolution is expensive!

something simpler ?
55
Optical Encoders relative

• Track position changes

light sensor
decode circuitry
light emitter
grating
56
Relative Encoders

• Relative position

- calibration ?
- resolution ?
light sensor
- direction ?
decode circuitry
light emitter
grating
57
Relative Encoders

• Relative position

- calibration ?
- resolution ?
light sensor
- direction ?
decode circuitry
light emitter
grating
A
58
Relative Encoders

• Relative position

- calibration ?
- resolution ?
light sensor
- direction ?
decode circuitry
light emitter
grating
A
A
A vs. B
B
B
59
Relative Encoders

• Relative position

- calibration ?
- resolution ?
light sensor
- direction ?
decode circuitry
light emitter
grating
A
B
A leads B
quadrature encoding
100 lines -gt how much resolution? more lines?
60
Relative Encoders
mask/diffuser

• Relative position

light sensor
A
decode circuitry
light emitter
grating
B
A diffuser tends to smooth these signals Two
thresholds are used hysteresis
Another 40 word free with the purchase of both
proprioceptive and "exteroceptive"!
Ideal
Real
From proprio- to exteroceptivity! First, a note
on building materials...
61
Lego-based quadrature encoding!
http//maven.smith.edu/anevin/Robotics/FinalProje
ct/Lecture05-AdvancedSensors.pdf
62
Lego-based quadrature encoding!
http//maven.smith.edu/anevin/Robotics/FinalProje
ct/Lecture05-AdvancedSensors.pdf
63
What's Next?
Motors geometry kinematics
Next time
as soon as possible!
Starting with the hardware
HW 1 due by 1159pm on Sunday, February 3
Reminder
Richard Vaughan and his 36 iRobot Creates
64
Have a great weekend!
65
Lab / Office Layout
Beckman B111
Legos
The Gator
Desk
Clinic Stuff
to file away
The Gator!
A million Roombas
Desk
The Nomad
Stuff I need
Nomad Reborn
RIP Nomad
Pan/tilt Camera for Duck Hunt
Lots o stuff (sensors, controllers) on shelves
door
door combination telephone number...
equipment access take care!
66
Right now Get Equipment
See you on Monday!
Next time
Motors geometry kinematics
Next official lab meeting
to be determined
Reminder
HW 2 due on Wed. January 31
67
Fictional Robots

• B9 from Lost in Space Danger, Will Robinson!
  (to be 1997)

• The Iron Giant (from the movie of the same name)

• Marvin the paranoid android (1989) from HHG

• Rosie (Jetsons) (to be 2062)

• Santabot and Bender (Futurama 3000) started off
  bending girders

• Agent Smith from The Matrix (2200)

• robots from Fritz Lang's 1927 movie Metropolis

• walking bombs from World of Warcraft

• R. Daneel Olivaw ( Isaac Asimov )

• Unicron and Optimus Prime

• Replicants, Blade Runner (2019 in LA)

68
Real Robots

• Groundhog, CMU's mine-exploring robot

• Roboraptor and Pleo robotic dinosaur
  toy/companion

• RHex, rugged hexapod robot that runs really fast
  (2002-)

• Paro, a robotic pet seal companion

• RiSE, a climbing robot very cool!

• The Segway, a two-wheeled balancing conveyance

• Honda's ASIMO perhaps the most capable humanoid

• Dragon Runner robot - a two-sided military
  robot by iRobot

• The Hubble telescope, the Stardust probe
  (aerogel!)

• Google (perhaps they're interested in funding us
  )

69
Homework 2
due 2/13/08

• Lab Project "getting moving"
• build/familiarize yourself with the platform
• be able to write and run programs move around
• 3 short problems on motor systems, as well

Barbie Jeep teams move motors Lego team look
into computer NXT control
the language is your choice
don't bother with a GUI until next week! (we have
a python GUI)
include one or more pictures/movies!

• be able to access and print sensor values

Barbie Jeep teams decide on / obtain initial
sensors Other teams use the built-in sensors
for this week (but consider others)
one on stepper motors, one on servo motors, one
on kinematics
just post one set of answers to these per team