Neural Networks in Games

The Book

Four Main Ways to Use Neural Networks in Games

  • Control

  • Threat Assessment

  • Attack or Flee

  • Anticipation

Control

  • Input
    • 5 sensors in front
  • Output
    • a left and right steering force

Threat Assessment

  • Input
    • Number of enemy ground units
    • Number of enemy aerial units
    • Indication if the grounds/aerial units are on the move.
    • the range to the ground/aerial units.
  • ​Output
    • ​Aerial threat
    • Ground threat
    • Both an aerial and ground threat
    • No threat

Attack or Flee

  • Inputs 
    • Number of like-creatures in proximity
    • Health Points;
    • whether there is a combat or not;
    • the range to the enemy;
    • Enemy Class
  • ​Outputs
    • ​Attack
    • Evade
    • Wander

Anticipation

References

Neural Networks in Games

By Eric Moura Guimaraes