Tuesday parts A and B: Lecture and Exercises
Download
Report
Transcript Tuesday parts A and B: Lecture and Exercises
Introduction to AI
&
AI Principles (Semester 1)
WEEK 4 – Tuesday parts A and B
(2008/09)
John Barnden
Professor of Artificial Intelligence
School of Computer Science
University of Birmingham, UK
The “Big Picture” for the Module
We’ve done some scene setting, getting an idea of what
AI is, getting to know a little of how to think about AI
situations, and seeing a few broad issues in AI (contextsensitivity, reasoning, uncertainty, default assumptions,
diversity of information, difficulties with language,…)
We’ve established three Case Studies (Hot Drinks party,
Shopping Trip, Crime Prevention) as a framework for
getting down to detail.
The rest of the module will largely be the detail …
involving “serious exercises” later on.
The “Big Picture” contd.
But first … the Big Picture does NOT involve at this
stage worrying about what will be in the exam!
AI is especially diverse and research-orientated.
Themes and techniques will unfold throughout the module,
not occur in neat chunks week by week.
You’ll learn a lot of what you need through the later “serious
exercises” as well as the more free-wheeling ones.
EVERYDAY VISION &
MOVEMENT
Vision Needs in AI
Seeing what objects are around // where they are // how they’re
moving // etc. … to some degree …
That degree being dependent on the system’s purposes.
Seeing what the “affordances” (basically // possibilities of
manipulation) of objects are // even if the objects can’t be
specifically identified.
Some special things:
Recognizing faces.
Recognizing facial expressions // hand gestures // body movements.
Reading text // mathematical formulas // diagrams // etc.
Seeing What’s in (e.g.) this Room
Difficulty of seeing where/what even the basic surfaces and edges
are: very noisy // corrupted signal.
Problem made more complex by shadows // reflections //
transparency // distortion // texture // uneven lighting.
Ambiguity of component shapes—is it a wheel? Is it a head? Is it an
orange? No … it’s …
Occlusion of objects.
Seeing objects from different angles.
Jointed objects (e.g. // people!) being in different configurations.
Soft and/or irregular objects such as clothes // crumpled paper.
Moving Around and Performing Actions
Remembering a “mental map” of some sort and knowing
where oneself is in such a map. Keeping track of
movements. Recognizing landmarks.
Creating such a map.
Moving arms // etc. to reach objects efficiently and safely.
Grasping (etc.) objects safely.
(HOT) DRINKS
Case Study
Planning the Delivery of One Drink
Planning the Delivery of One Drink, 1
Initial Situation: Mike (M) wants beer; Mike is in the sitting
room (S); you are in the kitchen (K); there is an unopened
bottle B of beer in the fridge (F); you are next to F; there is a
bottle-opener O on top of F; doors D-KS, D-KH and DHL[see below] are closed.
Other Facts:
F is in K
K and S are connected by door D-KS
K and the hall (H) are connected by door D-KH
H and S are connected by door D-HS.
Goal Situation (for You to achieve):
M has B; B is open.
Planning the Delivery of One Drink, 2
Actions available (at a high level):
agent a opens a bottle b with a bottle-opener o.
agent a opens a door d.
agent a closes a door d.
agent a goes through a room-door e.
agent a moves from (being next to) an object x to an object y.
agent a picks up an object x with one free hand.
agent a picks up an object x with both hands.
agent a puts down an object y.
agent a gives a held object z to agent p.
NB: In most cases you will act as agent a, but we should
be open to other possibilities.
Planning the Delivery of One Drink, 3
Individual Exercise “One-Drink A” [4 mins]
Write down one sensible sequence of actions that will achieve
the Goal from the Initial Situation. Use only actions from my
list.
Write down another such sequence.
Individual Exercise “One-Drink B” [3 mins]
How did you work out your first sequence, do you think?
Planning the Delivery of One Drink, 4
One sensible sequence of actions (or rather, action instances) is
as follows…. and is there anything missing?:
I move to F.
I open F’s door.
I pick up B with one free hand.
I pick up O with one free hand.
I open B with O.
I put down O.
I move to D-KS.
I open D-KS.
I go through D-KS.
I move to M.
I give B to M.
NB: we instantiated the variables in the action specifications by
replacing them by constants.