Team Member AI in an FPS

Team Member AI in an FPS
and Goal Oriented Action
Planning
Overview
► Correct
Positioning
► Correct Movement
► Correct Behavior
► Supporting the Player
► Implementation
► GOAP
► AIISC
Correct Positioning
► Should
not be in the players line of sight
► Should
not be in the players line of fire
► Should be visible to the player
► Should
give player first opportunity to act
► Positioning depends on body stance
► Should move out of sight if possible
Correct Positioning Continued
Correct Movement
► Find
smallest angle to a clear view
► Move
away from the players current viewing
direction
► Proximity
► Level
of awareness and readiness to fire
► Should use tactical strategies
► Respect current context of the situation
► Move at same speed as player
Movement Continued
Movement Continued
Correct Behavior
► Use
cover if available
► Selective
firing
► Reloading
Supporting the Player
► Capture
► Support
the flag
player in every movement
► Reporting
► In face of danger
► Taking orders
► Protecting the player
Player is the Most Important
► Complement
game play
► Use similar weapons
► Get choice of items after the player
Implementation of Squad Tactics
► Implementation
► 1:
Hierarchy
Player
► 2: Threat
► 3: Environmental
► 4: Team Manager
Implementation Continued
► Finding an Available NPC
► Availability = (1+N)(1+O)(1+P)+(Q∞)
► N = # of enemies in covering area
►O
= # of enemies within range
► P = # of enemies threatening team
► Q = Supporting team member (T/F)
► Fire Staggering
► Should not target same enemy
► Team manager should change cover
area
GOAP
(Goal Oriented Action Planning)
►A
decision making architecture that allows
characters to decide not only what to do,
but how to do it.
► Not a commonly used technique in today’s
games
► Superior to more common techniques such
as Finite State Machines (FSM) and Rulebased Systems (RBS).(AIISC)
Why GOAP
► Less
repetitive than a FSM
► Less predictable than FSM
► Can
adapt action to fit current situation
► A better way to simulate intelligence
Definitions
► Goal:
any condition that an agent wants to
satisfy
► Agent can have many goals
Initial state
A
C
B
A
B
C
Goals
Definitions
► Goal…
► Only
time
one goal active at any one
Sussman Anomaly
www.csee.umbc.edu/471/lectures
Definitions
► Goal
► Three
categories of gaming goals
► Relaxed
► Investigative
► Aggressive
Definitions Cont
► Goals
are not hard coded and do not
contain a plan
► Plan:
a sequence of actions that takes
an agent from a starting state to a
state that satisfies a goal
► Action: a single step within a plan
that make an agent do something
► Actions have effect(s) and may have
precondition(s)
Benefits of GOAP
► Characters
can find alternate solutions to
problems
► Characters can handle dependencies that
may not have been thought of at
development time
► Difficult to manage every possible situation
► Characters do not have to use all possible
actions (creates different behaviors)
Implementing GOAP
► Planner
Search
► World Representation
► Planner Optimization
Implementation Continued
Implementation Continued
AIISC
► AI
Interface Standards Committee
 Their benefits
►Modularity
►Workflow
►Gameplay
► Would
like to define an API that is appealing
to game developers
► Determine
how real-time dynamic planning
can be used in practice
References
► AI
Game Programming Wisdom 2
► AI Game Programming Wisdom
► The 2004 AIISC Report
► Symbolic Representation of Game World
State: Toward Real-Time Planning in Games