Chap. 1.3 Solving the Right Problem

1
Chap. 1.3 Solving the Right Problem

• AI Game Programming Wisdom
• ??? ???

2
Abstract

• Solving a different problem
• Might be more effective
• Game AI developer have to do
• Good programming and management habits

3
Solving a Different Problem

• Example
• IF NPC can talk to the player
• Need speech recognition NLP
• In speech recognition
• Little Red Riding Hood
• Ladle Rat Rotten Hut
• In NLP
• U.S.A Pissed Plowed (meaning)
• England Pissed ? Plowed (meaning)
• Taken two technologies together
• Too big for persons computer game
• More Effective
• Solving differently
• Using large motion gestures that are Universal
  across human cultures

4
First Principles

• What is the Real Problem We Are Trying to
  Solve?
• Trying to entertain someone
• Problem it was supposed to solve
• Better at solving the entertainment problem than
  a more realistic solution
• cheating

A (the real problem) B (another
problem that is easier t think about) B ? A (B
solves A many developers stop here) C
(yet another problem that is easier to think
about than A) C ? A (C also solves A)
5
Find the better problem

• Start early in the process, but continue looking
• Get the hooks in early
• Multiple solution to every problem
• Thinking Out-of-the-Box
• Creativity needs practice

6
Find the better problem

• Start early in the process, but continue looking
• Starting early
• Not guarantee that you will find better problems
• Starting late
• Certainly prevent that you will find better
  problems
• The precious resource is time
• Get the hooks in early
• AI programmer
• needs to ensure that access to information
• Early phases of game design, architecture

7
Find the better problem

• Multiple solution to every problem
• A single solution
• It is hard to say good, better, best
• Comparing between solutions
• Being able to Make better solution
• Thinking Out-of-the-Box
• Getting to the better problem
• Thinking outside the normal solution spaces
• To barge in on some other boxes
• Creativity needs practice

8
Chap. 1.412 Tips from the Trenches

• AI Game Programming Wisdom
• ??? ???

9
Abstract

• The course of a game development
• Tips for game AI programmers in course

10
12 Tips

• 1. Do Your Homework
• No such thing as a one size fits all
• Different techniques for different situations
• Consideration for the right solution
• The emphasis of the AI in the game
• The schedule and budget for engineering the AI
• The team make-up
• Choose focus for the AI development effort

11
12 Tips

• 2. Keep It Simple
• K.I.S.S. plan
• Keep It Simple Stupid!!
• In design and in code
• Simplification
• Complex things to simple parts
• Easy to comprehend, reuse, debug, and maintain
• 3. Try It Out on Paper First
• Write an outline of the code
• For a smoother implementation
• Provide a more robust system

12
12 Tips

• 4. Precompute Navigation
• Navigation Pathfinding
• In game, Easier than problem for academics
• Allowing cheat
• For cheaply navigation
• Using precomputed pathfinding data
• Checking Collision
• Using navigational hint information
• Paint floor geometry with levels of movement
  preference
• Place line segments to be used as preset path
  for agents
• Place boxes at least one box with a clear line
  of sight
• Create geometry freely move without checking
  for collisions
• A tool can analyze the level data and the result
  generate a navigational mesh

13
12 Tips

• 5. Put the Smarts in the World
• AI system cant handle every situation
• Simple AI system
• Let agents choose desirable destinations
• The game world gives the agent specific
  instruction
• Putting the intelligence into the world
• 6. Give Every Action a Timeout and a Fallback
• Check for success conditions
• Within a reasonable amount of time

14
12 Tips

• 7. Use a Hierarchy of States
• In finite-state machine(FSM)
• Putting states into a hierarchy
• Simple and reusable
• Facilitate simple lower-level states
• Using high-level for big-picture decision-making
• 8. Agents Dont Interfere with the Crucial
  Storytelling Events
• Using simple mechanism such as boolean flag

15
12 Tips

• 9. Keep Agents Aware of the Global State of the
  World
• To convince the player
• Agents use global flags and data, or employ
  decision-tree for self-learning
• 10. Create Variety through the Data, Not through
  the Code
• If create variety through the code
• Many AI programmers and different behaviors in
  code
• The code provide one or a handful of basic
  behavior types
• Behavior type are infinitely customizable through
  data

16
12 Tips

• 11. Make the Data Easily Accessible to Designers
• Interesting AI and good play balancing
• Designers tweak every value to fine-tune the AI
• Making convenient environment for designers
• Easy access to data
• 12. Factor Stat Formulas into AI
• Factor stat
• Agents attributes and abilities
• The player will easily be able to see characters
  stats
• Making character improvements more rewarding