Download AI needs within the game can include

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts

Artificial intelligence in video games wikipedia , lookup

Transcript
7.1.OSCARS & ARTIFICIAL INTELLIGENCE
Interim awards and introduction to game AI
Mid-way awards
Submission with the most
impressive/complex exploratory code
Submission with the best progress to date
o Team members
o
Game name
o Team members
Team name
Game name
Most authentic rendition of a classic computer game
Most original/fun game design or game play idea
Game I least want to present at the Board of Examiners
Most comprehensive game design including development
plan and contingency planning.
Team/Game with the most inappropriate name
Team/Game with the best name
Category
Team members
Team name – Game name
Introduction to game-oriented artificial intelligence
Artificial Intelligence (academic vs. game)
Artificial intelligence aims to develop
machines that can perform human
‘thinking’ tasks.
Academic research is split into two camps:
• Strong AI – creating systems that model
human thought processes
• Weak AI – creating working systems that
need not be physiologically plausible
Academic AI tends to focus on optimal
problem solving.
Game AI must work within tight
computational constraints, i.e. effort vs.
outcome is central.
The aims of game artificial intelligence
The aims of game-oriented AI can be
summarised as follows. The AI must:
• appear intelligent, yet purposely
flawed (i.e. beatable)
• have no unintended weaknesses (that
can be repeatable exploited)
• provide an entertaining or engaging
experience
• perform within tight CPU/memory
constraints
• be configurable
• not keep the game from shipping
The illusion of intelligence…
Most people assess intelligence (or the lack
thereof) on how an object behaves.
Acting in a complex (human-like) manner is
readily perceived as intelligence behaviour.
One means of enhancing game
AI is to provide visual/auditory
feedback on what the game
object is ‘thinking.’
Often simple or semi-random
behaviour will be
perceived/intepretated by the
player as complex/intelligent.
Aside: Describe the
colour of square A?
What colour is B?
A is exactly the
same colour as B!
A practical definition of game AI
“Game AI is anything that contributes to
the perceived intelligence of an entity,
regardless of what’s under the hood.”
Aside: Searle’s Chinese Room argument
Game AI (is it, or isn’t it)
Which of the following could be classified as
providing an example of AI within the context of a
game?
• Does a single ‘if’ statement constitute intelligence?
• What about scripted behaviour?
• If an NPC selects which animation to play?
(If this is done via a set of if statements?)
• Maybe path-finding?
• If game automatically generates an environment?
The forms of AI found within different types of (2D) game
AI (in general)
AI needs within the game can include:
Perception – determining what can be
seen (other opponents, pick-ups,
incoming projectiles, etc.)
Steering – basic character movement
Action – executing available actions,
e.g. aiming, shooting, etc.
Path-finding – movement route
planning
Decision making – determining
what to do next (dodge, seek health,
ambush, etc.). At higher levels this
becomes tactical AI.
Perception
Line of sight tests
Strategic AI
Influence maps
Decision Making
Steering
State Change
FSM
Kinematic movement
Object update
AI (side-on/top-down shooters)
AI needs within the game can include:
Perception – detecting nearby objects,
incoming projectiles, etc.
Steering – opponent movement, e.g.
player tracking, projectile avoidance,
etc.
Firing – basic control, firing towards
player
Perception
Aside: AI within 2D shooters
may be effectively nonexistent,
i.e. relying on fixed patterns of
movement and opponent
numbers to provide challenge
Object detection
Steering
Path follow, Evade
Shooting
Path projection
State Change
Object update
AI (driving)
AI needs within the game can include:
Perception – detecting other traffic
Steering – driving line, cornering,
breaking
Decision making – overtaking points,
collision avoidance
Perception
Aside: GTA/Driver clones would
also include AI routines to model
other road traffic, etc.
Predictive collision detection
Decision
Making
FSM, Rule-based system
Steering
Path following
State Change
Object update
AI (platform)
AI needs within the game can include:
Perception – determining
actions/movement of player
Steering – moving towards/away from
player
Shooting – basic control, e.g. aiming
Aside: Platform games tend to
have opponents which have
predictable, easily understood
behaviour. Challenge arises from
the need to time jumps, shots,
etc. to overcome such
opponents.
Perception
Player proximity
Steering
Path follow, Pursue
Shooting
Player proximity trigger
State Change
Object update
AI (real time / turn-based strategy)
AI needs within the game can include:
Perception – determining what can be
seen (other opponents, resources)
Steering – group movement, etc.
Path-finding – movement route
planning
Perception
Opponent visibility, Fog-of-war
Tactical and Strategic Analysis
– determining overall strategy
build, attack, etc.
Strategic AI
Tactical analysis, Influence maps
Aside: AI in real-time games is
mostly the same as in turnbased games. Real-time games
must impose tight performance
constraints on the AI.
Path finding
/ Steering
Tactical AI
Decision Making
State Change
Rule-based system/ FSM
Map/local path-finding
Object update
AI (beat-em-up)
AI needs within the game can include:
Decision making – determining
what to do next (block, back-up, attack,
etc.).
Decision Making
Aside: The behaviour can be
adaptive, i.e. reacting to the
player’s patterns of behaviour
State Change
FSM / Rule-based behaviour
Object update
AI (sport)
AI needs within the game can include:
Steering – basic character movement, group
movement, etc.
Decision making – determining what to do
next, selecting plays, formations, etc. from an
available ‘playbook’
Tactical Analysis – determining play
objectives
Aside: Sport AI has the benefit
of drawing upon existing expert
knowledge, but must return
realistic, ‘human-like’ behaviour
Tactical AI
Influence Maps
Decision
Making
Rule-based system
Steering
Chase, Evade, etc.
State Change
Object update
Summary
Today we
explored:
 The role of AI
within games
and the
constraints
game AI must
operate within
 The typical
roles of AI
within 2D
game genres