1. Course Objectives
Know what it takes to make a robust autonomous robot work:
Sense/Think/Act
Understand the important, approaches, research issues and challenges in autonomous robotics.
Know how to program an autonomous robot.
Just showed you a robotic example of agents interacting with agents
we want to develop CS to observe and model multiagent systems
we need an example system
Collaboration
Ant algorithms
network routing
robot navigation
scheduling
Interested in application to CS and biology
Veloso modeling soccer agents for a while
Observing ants just began recently
Biologists currently use pencil and paper
Introduction to AI Robotics (MIT Press)
Chapter 1

2. What Can Robots Be Used For?
Manufacturing
3 Ds
Dirty
Dull
Dangerous
Space
Satellites, probes, planetary landers, rovers
Military
Agriculture
Construction
Entertainment
Consumer?
Introduction to AI Robotics (MIT Press)
Chapter 1

3. History of Intelligent Robotics
First remote manipulators for hazardous substances
1950s
Industrial manipulators: “reprogrammable and multi-functional mechanism designed to move materials, parts, tools…”
Closed loop control
Introduction to AI Robotics (MIT Press)
Chapter 1

4. History Continued Introduction to AI Robotics (MIT Press) Chapter 1
1955 – term “AI” coined
1960s manufacturing robots
Automatic guided vehicles (AGVs)
Precision, repeatability
Emphasis on mechanical aspects
1970s
Planetary landers
Machine vision research expands
1980s
Black factory
First intelligent autonomous robots:
Shakey, Stanford Cart, etc
Introduction to AI Robotics (MIT Press)
Chapter 1

5. History Continued 1990s 2000s Introduction to AI Robotics (MIT Press)
Symbolic AI/Robotics stalls
Reactive/Behavior-based robotics emerges
2000s ?
Introduction to AI Robotics (MIT Press)
Chapter 1

6. Intelligent Robot Mechanical creature which can function autonomously
Mechanical= built, constructed
Creature= think of it as an entity with its own motivation, decision making processes
Function autonomously= can sense, act, maybe even reason; doesn’t just do the same thing over and over like automation
Introduction to AI Robotics (MIT Press)
Chapter 1

7. “Intelligent” Robotics
Basic robot primitives : Sense/Think/Act
Three paradigms (architectures): Hierarchical (Deliberative): Sense ->Plan ->Act ; Reactive: Sense -> Act; Hybrid (Deliberative/Reactive): Plan -> Sense -> Act
Introduction to AI Robotics (MIT Press)
Chapter 1

8. Ways of Controlling a Robot
“RC-ing”
you control the robot
you can view the robot and it’s relationship to the environment
ex. radio controlled cars, bomb robots
operator isn’t removed from scene, not very safe
teleoperation
you can only view the environment through the robot’s eyes
don’t have to figure out AI
semi- or full autonomy
you might control the robot sometimes
ex. Sojouner with different modes
human doesn’t have to do everything
Introduction to AI Robotics (MIT Press)
Chapter 1

9. Teleoperation Human controls robot remotely Considerations
Hazardous materials
Search and rescue
Some planetary rovers
Considerations
Feedback (video, tactile, smell?)
User interfaces (cognitive fatigue, nausea)
Time/distance
Introduction to AI Robotics (MIT Press)
Chapter 1

10. Components of a Telesystem (after Uttal 89)
Local
display
Local control device
Communication
Remote
sensor
mobility
effector
power
Display
Control
Sensor
Mobility
Effector
Power
Communi-
cation
Local
Remote
Introduction to AI Robotics (MIT Press)
Chapter 1

11. Example Remote Local Introduction to AI Robotics (MIT Press) Chapter 1
This is a picture of Dr. Robin Murphy operating Cowboy, one of the first iRobot Urbans developed for the DARPA Tactical Mobile Robot program. The scene is the SRDR training site in Miami, Florida, and the task is to explore a collapsed building for survivors (urban search and rescue).
Remote
Local
Introduction to AI Robotics (MIT Press)
Chapter 1

12. Typical Run
Introduction to AI Robotics (MIT Press)
Chapter 1

13. Common problems with remote-controlled robots
no feedback, couldn’t really tell that the robot was stuck but finally got free
robot doesn’t have “proprioception” or internal sensing to tell you what the flippers were doing. No crunching noises, no pose widget to show the flippers
no localization, mapping-> no idea how far traveled
partial solution: better instrumentation (but can’t do dead reckoning well)
operator doesn’t have an external viewpoint to show itself relative to the environment
solution: two robots, one to spot the other
communications dropout, even though ~3 meters away
lighting conditions went from dark to very bright
hard for computer vision or human to adjust
Introduction to AI Robotics (MIT Press)
Chapter 1

14. DarkStar+7 seconds=DarkSpot
7 second lag time
also, though this didn’t play a role in this case, there’s a lot of cognitive literature on how people react when having to change tasks– when they’re doing one thing and then get interrupted to do something else. Usually takes a while to context-switch (page in)
7 second communications lag (satellite relay)
“interruption” lag on part of operator
Introduction to AI Robotics (MIT Press)
Chapter 1

15. Predator: ~7:1 human to robot ration
Leo’s unofficial
Predator page
4 people to control it (52-56 weeks of training)
one for flying
two for instruments
one for landing/takeoff
plus maintenance, sensor processing and routing
lack of self-awareness– in Kosovo, come along side in helicopter and shoot down
Introduction to AI Robotics (MIT Press)
Chapter 1

16. Teleop Problems cognitive fatigue communications dropout
communications bandwidth
communications lag
too many people to run one robot
Introduction to AI Robotics (MIT Press)
Chapter 1

17. Telesystems Best Suited For:
the tasks are unstructured and not repetitive
the task workspace cannot be engineered to permit the use of industrial manipulators
key portions of the task require dexterous manipulation, especially hand-eye coordination, but not continuously
key portions of the task require object recognition or situational awareness
the needs of the display technology do not exceed the limitations of the communication link (bandwidth, time delays)
the availability of trained personnel is not an issue
Introduction to AI Robotics (MIT Press)
Chapter 1

18. Teleop Solutions Telepresence Semi-autonomous
improves human control, reduces simulator sickness and cognitive fatigue by providing sensory feedback to the point that teleoperator feels they are “present” in robot’s environment
Semi-autonomous
Supervisory Control
human is involved, but routine or “safe” portions of the task are handled autonomously by the robot
Shared Control
human initiates action, interacts with remote by adding perceptual inputs or feedback, and interrupts execution as needed
Traded Control
human initiates action, does not interact
Mixed Initiative (Guarded Control)
robot doesn’t let the operator injure the robot (without override)
“whoever figures it out first”
Introduction to AI Robotics (MIT Press)
Chapter 1

19. Collaborative Teleoperation1 3
mpg: June 2, 2000 SRDR Miami Beach: view from Inuktun as it falls
mpg: June 2, 2000 SRDR Miami Beach: view from Inuktun from hoisted position 2
Urban is stuck, Inuktun can’t help from current perspective
Driven off 3rd floor
Hoisted to 2nd floor by tether
Has better view, changing configuration & rocking extend view
still: June 2, 2000 SRDR Miami Beach
Introduction to AI Robotics (MIT Press)
Chapter 1

20. 2000 AAAI Mobile Robot
2 robots helping each other reduced collision errors, sped up time navigating confined space, righting
Introduction to AI Robotics (MIT Press)
Chapter 1

21. Example: Mixed-Initiative & Collab. Teleop
9/2000 DARPA Tactical Mobile Robots demonstration
Robot used an intelligent assistant agent to look for signs of snipers hiding in urban rubble
motion
skin color
difference in color
thermal (IR camera)
Human navigated mother robot using viewpoint of 2nd robot (not in picture)
Once deposited the human moved the daughter robot, and either saw a sniper or was alerted by the agent
This is USF’s marsupial team– a customized RWI ATRV robot that can carry three Urban robots. The mother is being teleoperated using a second daughter off to the side (collaborative teleoperation) that isn’t in the picture.
Note: the idea of the robot doing the perception and the human doing the navigation is pretty much the opposite of conventional semi-autonomy thinking where people do the perception and the robot tries to figure out where to go next. USF’s work in USAR with fire rescue workers suggested that they felt comfortable joysticking the robot around but joysticking and looking through a regular camera and an IR camera simultaneously was too much. They voted for having the robot do the vision part, trying to detect people.
Introduction to AI Robotics (MIT Press)
Chapter 1

22. AI provides the “other stuff”
knowledge representation
understanding natural langugage
learning
planning and problem solving
inference
search
vision
Introduction to AI Robotics (MIT Press)
Chapter 1

23. Summary
Teleoperation arose as an intermediate solution to autonomy, but it has a number of problems:cognitive fatigue, high comms bandwidth, short delays, and many:one human to robot ratios.
Telepresence tries to reduce cognitive fatigue through enhanced immersive environments
Semi-autonomy tries to reduce fatigue, bandwidth by delegating portions of the task to robot
Introduction to AI Robotics (MIT Press)
Chapter 1