What are the safety risks with autonomous robots? Are autonomous robot a threat to human workers? How can robots achieve intelligent behavior? What mechanisms are needed to reach human agility?

1. By,
Ken Hua
Vishal Verma
Srinivas Panchapakesan

2. Benefits of robots working these
jobs
 Harsh Environments
 Mitigates risk of death/injury to humans
 More consistent
 Efficient
 Tireless
 Cheaper (w.r.t. maintenance costs)

3. Here are some examples
 Harvesting
 Mining
 Motor industry
 Warehouse
 Search and Rescue

4. Progress
 Advances in Technology
 Paradigm Shifts
 Draftsmen
 Automotive Industry
 Telephone operators

8. Why it is good:
 Time for other activities
 Proven in history
 More intellectual jobs
 General progress and upliftment
 New jobs created automatically
 Change happens gradually

9. Conclusion
 Need to harness potential
 Resilience of humans

10. Robots are stealing human jobs.
Niket Sheth
Kelsey Metzler
Chris Karman

11. What kinds of robots exist?
 Industrial Robots
 Mobile Robots
 Robots Used in Agriculture
 Telerobots
 Service Robots
www.buzzle.com

12. Which Robots are “Okay”
 Industrial Robots – used in large scale manufacturing
 Mobile Robots – moving within large factories
 Telerobots – used in dangerous or difficult to reach
places
 Able to work in dangerous and unreachable places,
efficient, and reliable
 Example: Snake Robot
 Facts: 7484 fatalities in the industrial sector in year 2007.
(Source: US Bureau of Labor Statistics)

13. Which Robots are “Bad”
• Agricultural Robots
• Service Robots
• These tasks are not dangerous
• Do not require superior efficiency
• “Robots could substitute for a workforce of 3.52
million people in Japan by 2025”
‐Japan Today

14. Agricultural Robots
 Main application is Harvesting
 Fruit Picking Robots
 Other applications
 Shearing Sheep, Milking Cows
 Planting seeds
 Pruning
 Weeding
 Spraying
“Oracle” Sheep Shearing Robot

15. Harves?ng Robot
• Use image recognition and processing technology
• “Harvesting is the most labor‐intensive activity for
many crops, but even advocates say that no one has
built a machine that comes close to matching the
sensory motor control of humans.”
–Popular Mechanics

16. Livestock Robots
• Dangerous, frightening, and painful for the animal
– 'Shear Magic' Sheep Shearing Robot
– “robotic system is not significantly faster or more
efficient than having a human do it”
‐ http://www.botjunkie.com

17. Service Robots
I. Housekeeping
• Cleaning
• Washing
• Folding
• Laundry
• Vacuuming
II. Care‐Taking
• Elderly/Handicapped Care
• Pet Care
III. Customer Service
• Retail
• Food Industry
• Voice Service and Recognition

18. Housekeeping & Care‐Taking
Housekeeping Care‐Taking
• Taking jobs away from immigrants.  The elderly/handicapped person’s
• New improved jobs would require health can go wrong suddenly and
higher education which costs the robot would not know what to
money. do.
• Become poor since no job and • i.e.: Heart attack
money to get better education to • Lives are at stake
get a high paying job. • Quick decision making skills
• The gap between rich and poor • Need human‐to‐human interaction,
would get bigger and bigger. the emotional support to get
• America the land of opportunity, through life.
the New World ‐> Non‐Existent. • Automated robot has no human
emotion, can cause some mental
issue/distress on the elderly and
handicapped person.

19. Customer Service
• 60% of the American working sector
Retail employed at $14/hour or less
• Automated Checkout Lines • A chunk of these employees are from
• Automated Inventory the retail and food industry
Management System • A majority of those are teens trying
to pay for tuition and rent (make a
living)
• Automated Inventory Management
System
• Each shelf and product have its
own RFID tags and bar code
• Allow mobile robots to pick‐up,
identify, shelve and re‐shelve,
shuttle, and stack products in
warehouse and shops
• Automation would cause cheap
labor and massive layoffs that
increase unemployment

20. Customer Service Cont…
Food Industry Voice Service & Recognition
• Waiters/Waitresses • Voice operated robotic
• Cooks systems (Airlines /Computer
related questions and
problems)
• Voice Recognition in the years
to come…
• Near perfect
• Be able to understand
multiple languages
• Deployment of robots and
kiosks throughout the
stores integrated with the
voice recognition systems
to help customers with
product location needs.

21. References
 http://www.marshallbrain.com/robots‐in‐2015.htm
 http://www.cribcandy.com/robots/
028775ad7891e934617f172b99203302&pageoffset=0

22. Autonomous Weapons/Cars
Contra
Mike Pack
Tommaso Buvoli
Sean Egan

23. Overview
Buggy Systems
Sensor reliability
Out of Sync
System Maintainability
Is the Public Ready?

24. Sensor Reliability
Sensors degrade over time
Need sensors to detect degradation
And a sensor to read this sensor
And a sensor to read this sensor
And a sensor to read this sensor
And a sensor to read this sensor
And so on...
Question: How many levels of sensors is sufficient?
Answer: Infinite levels. It's not possible.
Human lives are at risk

25. Buggy Systems
Bugs can cause unfavorable results
Improper image recognition
Is it a child or a bomb
shaped like a child?
Blow it up anyway for fun?
Localization errors
Cars drive off the roads!
2 foot difference in localization is extreme

26. Buggy Systems Today
Does anyone recall the hybrid?
If a car cannot control
its breaks, how will it
be able to drive?
Also who is liable in
an accident.

28. Out of Sync
System Maintainability
Out-of-sync real-world motives
Robot believes Iraq is bad
Humans believe Afghanistan is bad
Humans must maintain software
Time consuming task
Accurately describe real-world situation
Robot must fully understand
Humans can make a more accurate, quicker decisions

29. Is the Public Ready?
How do you explain the
advances in layman's terms?
Political and ethical issues
Who is responsible for
accidents?
How do other countries feel?
6-12 THOUSAND weapon-bearing robots in Iraq and
Afghanistan
May 2009 -Plan to use swarm intelligent autonomous robots
Robot dystopia
Scared baby!

30. Autonomous Robots
By‐ Patrick Cromer
Nisheeth Bhat
Sa9sh Rajan

31. Is he your friend?

32. What is Autonomy ?
• Working without Human interven9on.
• It may or may not be a Learning Machine.
• Non‐ Autonomous
– Military controlled UAV’s
• Autonomous
– Robo9cs Arms – without decision making
capability
– Tac9cal Mobile Robots(TMR) ‐with decision
making capability

33. TMR Challenges
• For the TMR’s, error in iden9fica9on of
poten9al threats friends/foes.
• First, the TMR must iden9fy if aggressor is a
human or animal .
• If a human, what type of person is it?
• Does the profile indicate they are carrying a
weapon?
• Will this en9ty compromise the mission? What
defense level is appropriate?

34. Assuming everything worked !
• Loop Holes ?
• Enemies dressed and ac9ng like friendly
soldiers.
• Can you find a way out of this with a
Mathema9cal Model/Algorithm.
• Profile determina9on and whom to aWack?
• Note: The movie Minority Report.

35. Real World Issues
• German soldiers accidentally killed 6 friendly
Afghanistan soldiers (April 04, 2010 – NY
Times)
• Iraq insurgents passed themselves off as U.S.
soldiers and killed 24 Sunni’s who revolted
against al‐Qaida ( April 03, 2010 ‐ NY Times)
• Would a robot be able to differen9ate
between friend/foe in these situa9ons?

36. Error in posi9on es9ma9on
• Assuming the target’s sta9onary and no one is
around it can probably be hit accurately.
• What if the target is moving?
• What if friendly/non‐combatant units are in
the area?
• How fast can the robot make these decisions?
It needs to be done in real‐9me.

37. Autonomous Cars..present scenario
• Have we been able to develop autonomous
cars?
– Stanford’s Junior drove mostly without human
interven9on but there was s9ll human in there!
• Handling of traffic like in New York?
• How will the decisions be made at split second
efficiency ?
• Ques9on of life and death…

38. Loop holes…
• How well the percep9on works in different
environments and at high speeds.
• Do we have the hardware for making split
second decisions
• Is Ar9ficial Intelligence the solu9on?
• If it is ,it can’t even simulate a human brain!
• Then how will they reduce accidents?

39. Machines can be hacked!
• Sogware can be easily modified/corrupted in
real 9me.
• What about robots that communicate and
base their decisions on the accuracy of each
other.(Swarm Intelligence)
• Imagine a machine controlling another
machine and the controller gets
affected(ripple effect)
• Ex: TMR Robo scenario.

40. Conclusions
• How do you want to live your life ?
– Purposeful life where you are free
Or
– As mere Domes9c animals

41. Group Debates (D3-)
Robots Need Human Cognizant Traits
April 19, 2010
Michelle Bourgeois
Brian Brooks
John Martin

42. What is a Robot?

43. Intelligent Environment?
• Cannot control the environment enough
o Interaction with humans
 service robots
 companion robots
 surgical robots
• Environmental control not possible
o Space Exploration
o Underwater Robots
http://www.mnn.com/sites/default/files/0410walle_main.jpg

44. Service and Companion
Robots
humanoid "Twenty-one"
RobotCub / iCub
Care-O-Bot3

45. Human Social Interaction
• Dialogue
• human actions and
behaviors
• Skill and task learning
• Spatial awareness
Care-O-Bot3

46. Surgical Robots
• IEEE Trans. Ultrasonics, Ferroelectrics and Freq. Control
o autonomous robot directing catheters inside synthetic
blood vessels
• IEEE Ultrasonic Imaging
o autonomous robot performing simulated needle biopsy
• Engineering @ Duke
o Real-time 3D imaging via ultrasound technology
o Successful biopsy of a cyst (artificial mtrls)

47. Safety Evaluation [Ikuta]
• Injury
o Mechanical, Electric, Acoustic
• Design
o Reduce weight
o Absorb impact
o Joint compliance
o Shapes / space
o Reduce surface friction

48. Safety Evaluation [Ikuta]
• Control
o Keeping distance
o Approaching Velocity
o Posture / Stiffness

49. Space Exploration
• Numerous places where it is not feasible to manipulate the
environment for a robot
• Deep space missions
o Time delay at speed of light is 6 min to Mars
o If we wanted to explore stars/planets further away the
communication delay could be very large
o Need the robots to be automated and handle issues on
their own
 Need a more cognitive robot in order to be able to
preform this kind of deep space research

50. Space Exploration
• Spirit 1 spacecraft
• Wheels became stuck in
soft soil on May 1, 2009
• Stuck since
• A more cognitive robot may
have been able to detect the
soft soil earlier and decide to
not go further.

51. Underwater
• Another example of an environment which is not
feasible to manipulate enough for a robot
• Current reasearch into underwater automated
robots

52. Middle Ground
• ACM/IEEE Conference on Human-Robot Interaction
o [Wang] Simulated search and rescue swarm.

53. References
http://www.cogniron.org/
http://www.care-o-bot-research.org/
http://www.springerlink.com/content/m137403033644541/
http://www.aaai.org/ojs/index.php/aimagazine/article/view/2066/2053
http://www.springerlink.com/content/r2555625qq788241/
http://www.aass.oru.se/~peis/
http://www.scientistlive.com/.../Autonomous_robot_surgeries...
[Ikuta] http://ijr.sagepub.com/cgi/reprint/22/5/281
[Wang] http://delivery.acm.org/10.1145/......
http://marsrover.nasa.gov/mission/status_spiritAll.html

54. www.xkcd.com

55. Debate
Nicholas Farrow, Ethan Federman,
and Zach Pranger

56. Question:
Robots do not need to be as cognitive as humans in order to be
useful as making the environment intelligent is sufficient.
Our Stance:
Agree, making the environment intelligent
is sufficient to make robots useful.

57. Why? Here are our facts
RFID
GPS
Guiding Flooring
Wireless Information Dispersal

58. Robots using RFID
Radio Frequency Identification tags
Active
Passive
Battery-Assisted Passive - which require an external
source to wake up but have significant higher forward
link capability providing greater range
Improve the efficiency of inventory tracking and
management

59. Robots using RFID

60. Robots using GPS
Global Positioning Satellite
provides reliable positioning, navigation, and timing
services to worldwide users
Each satellite continually transmits messages which
include:
the time the message was transmitted
precise orbital information (the ephemeris)
the general system health and rough orbits of all GPS
satellites (the almanac).
Possible example:
autonomous trains

61. Guiding Flooring
Color coded/information passing flooring
Robots could read floor pattern and know where it was.
Navigational Assistance
Positioning Assistance
Ex: Kiva Mobile Fulfillment System

62. Wireless Information Dispersal
Wirelessly send information to robots
Routing information
Orders
Navigation
Localization
Centralized Robot Wireless Communication

63. Wireless Information Dispersal
Wirelessly send information to robots
Orders

64. Wireless Information Dispersal
Wirelessly send information to robots
Orders

66. Nick Marsiglia
Mason Hauck

67. Topics
 Improved components (Gears)
 Example (high speed robotic hand)
 Improved controllers/algorithms
 Example (violin robot)
 Example (Big Dog)
 Actuator Trade Study
 SMA wire bundle actuators
 Stepper motor/lead screw actuators

68. Improvements In Components
• Gears
– High precision
– Compact (more gears to
improve accuracy of
movement)
– New designs/ design
implementation
• Harmonic drive
– No backlash
– Small
– Light weight
– High gear ratio:
– 100:1 Harmonic drive
– 10:1 Planetary gear

69. Example: High Speed Robo<c Hand
 Uses harmonic drive
gear
 High power mini
actuator
 Allows for high speeds
 Opens and closes hand
at 1/10 second

70. Improvements In Control
Algorithms
• Learned dynamics
– Robot learns by iteration
• Reduction in
computational cost
– Less strain on processor
• Faster processors
– Quicker computation
time=faster movements/
less delay
• Combining these methods
with traditional feedback
control

71. Example: Toyota Robot
• Designed for help in:
hospitals, factories, and
around the home
• Has 17 joints in one hand
and arm (mimics human
arm)
• Combines control and gate
algorithms with high
complexity and accuracy
of components.
• Dexterous enough to play
violin

72. Example: Big Dog
• Coordination provided by
gait algorithm
– Controls posture and
movement
– Feedback from ground
sensors controls load
balance
• Combination of feedback
control and desired
motion algorithm
• Uses hydraulics for
actuation

73. Actuator Trade Study
 Shape Memory Alloy  Stepper Motor/Lead
Wire Bundle Actuators Screw Actuators
 Bundle of two 250 um  Haydon Kerk 35000
Nitinol wires Stepper Motor Lead
Screw Actuator

74. Actuator Trade Study
 Shape Memory Alloy  Stepper Motor/Lead
Wire Bundle Actuators Screw Actuators
 Very Low Actuator  Higher Actuator Weight
Weight
 Very bad actuator speed  Fine actuator speed
control control
 Non‐variable actuator  Speed and force
force inversely proportional
 High waste heat  Lower waste heat

75. Control Complexity
 SMA Actuators  Stepper Motor Actuators
 Low precision, more  High precision, simpler
complex control control algorithms
algorithms  Discrete digital control,
 Highly non‐linear very linear operation
operation  No hysteresis
 High degree of hysteresis

76. Actuator Force Vs. Supplied Power
 Bundle of two 250 um  Haydon Kerk 35000
Nitinol wires Stepper Motor Lead
Screw Actuator
 Actuation Force = 53.4 N  Actuation Force = 175 N
 Supplied Power = 14.5 W  Supplied Power = 11.5 W
=> F/P = 3.68 N/W => F/P = 15.9 N/W
(SMA F/P) / (Motor F/P) = 4.12

77. Overall System Benefits
 Shape Memory Alloy  Traditional Actuators
Actuators  Lower battery mass
 Lower actuator mass  More precise actuator
 More compact actuators control
 Simpler control
algorithms
 More efficient operation

78. Robots Grappling With Human
Agility
Debate Group 4:
Joe McCabe
Mikael Pryor

79. Nao is SLOOOOW
http://www.youtube.com/watch?v=rSKRgasUEko

80. STIFF
• Muscles and tendons
– Humans have joints
– Actuated through co-
contraction and adaptive reflex
gains
– More robust, versatile, and
agile than joints, gears, and
links
• More grace and power
http://stiff-project.eu/

81. Manufacturing of Old
• Robot arms consisting of joints, gears,
and links are dangerous
– Mass and speed result in dangerous
momentum
– Have to be kept in a cage to prevent
injuries to human counterparts
• Designed for specific tasks
– Very precise, but at a high cost of having
few applications
Patrick van der Smagt *, Markus Grebenstein, Holger Urbanek, Nadine Fligge, Michael Strohmayr,
Georg Stillfried, Jonathon Parrish, Agneta Gustus. Robotics of human movements

82. Flexible Manufacturing
• The ability to learn from situation based fine-
tuning of the end-effector position
– Can be used for more applications
• Less Weight, more sensors, and still powerful
– Less dangerous to humans
– Able to interact with the environment and adapt
by reacting to sensor data
• Less prone to failure
Patrick van der Smagt *, Markus Grebenstein, Holger Urbanek, Nadine Fligge, Michael Strohmayr,
Georg Stillfried, Jonathon Parrish, Agneta Gustus. Robotics of human movements

83. Where No Man Can Reach
• Flexible manipulators can be used in nuclear power
plants
– Due to being less prone to failure
– Actuation can be done with parts that are less likely to
fuse
• Sensors can prevent manipulators from running into
themselves
• Manipulators can reach areas unattainable with
joints, gears, and links.
– Since this is done through fine-tuning, singularities in
joint space are not an issue