597F: Intelligent Robotics/Sharma/Fall'98, Notes #1

Action in the world, mechanism for movement

Reprogrammable, information processing

Have controls and sensors

1940s : handle radioactive items, electrically connected to human operator in a Master-Slave relationship.

1950s : Computer Numerically Controlled (CNC) 'robots' used in industrial application - milling aircraft wings; Unimation founded by Joseph Engelberger

1960s : sensors and force integrated; 1969 Kawasaki Corp. founded

1970s : Programmable Universal Machine for Assembly (PUMA)

robots language developed in 1978; Selective Compliant ???

(SCARA) robots begun 1979

1980s : space applications of robotics, including developments in control,

reasoning, sensing and especially vision incorporated; portable

computation becomes much cheaper

1990s : sensing, computation and camera mounted robots -> lots of mobile robots;

reconfigurable and distributed robot systems developed;

new applications: agriculture (sheering sheep, picking fruit)

manufacturing: repetitive tasks; harsh environments(hazardous waste, undersea, space, firefighting, military); medicine: non-invasive surgery, tele-operated surgery, micro-and nano- scale robots

Internal Sensing: sensing of internal parameters (esp. position, velocity, acceleration)

via potentiometer, position encoders, tachometer, accelerometer, odometer

External Sensing:

Contact: tactile, limit switch/touch sensor, pressure sensors (grasping without breaking)

Non contact: vision, radar, infrared, ultrasonic

Design Issues :

Kinematics (position, velocity and acceleration; no force involved) +

Dynamics (force calculations involved)

Sensors: external/internal

Control: feedback, as well as strategic, such as that required for juggling

Planning / Reasoning : example of assembly was cited:

to move an object, the robot needs to

1. Be able to represent the world coordinate system
2. Determine the position and orientation of object in that system
3. Articulate the goal position of that object
4. Deal with constraints, such as the set of available paths and obstacles

Two Related Concepts:

Gross Motion Planning (path planning)

Fine Motion Planning (ex.: "grasp planning" such as how to approach and handle the object at a close distance)

Motion Planning : given an initial state, the robot must determine how to

get to the desired 'final' state

1. Satisfying certain constraints: obstacles, velocity, acceleration
2. Optimizing some criteria: time, distance, energy, risks

Factors of Uncertainty are involved: robots may have complete

information about the world, or incomplete information about the world,

or about its own sensors

Interesting Robotics projects:

Rodney Brooks and the MIT AI Lab can be found at

http://www.ai.mit.edu/

An AI course "Embodied Intelligence"at MIT taught by

someone other than Rodney Brooks can be found at http://www.ai.mit.edu/courses/6.836/6836.html

MIT Press's Intelligent Robotics and Autonomous Agents book offerings

are at http://mitpress.mit.edu/books-in-series.tcl?series=Intelligent%20Robotics%20and%20Autonomous%20Agents

It currently lists a couple of books on "behavioral" robotics.

John Pollock at University of Arizona: a philosopher, he has worked on a project called OSCAR, a "defeasible reasoning": program

(defeasible reasoning refers to the ability to adapt one's reasoning

process and conclusions to new perceptions of the environment).

It is a more philosophical approach to planning under uncertainty and reasoning problem of robots. His web page is located at http://www.u.arizona.edu/~pollock/

Carnegie Mellon's AI home page :

http://www.cs.cmu.edu/Groups/AI/html/repository.html