In this FlightSim example using Evolutionary Computation, an initial population of game agents Non Player Characters (NPCs, aka: intelligent agents or software agents) are set-up with a set of traits that define physical and cognitive abilities, each starting with random values. The physical traits are for things such as rate of turn, acceleration, maximum and minimum speeds, etc. The cognitive traits are for decision making triggers such as: lead-distance for firing, when to fire, when to flee, which way to turn, etc.
The population is evolved such that the best performers are the ones that survive (don't get shot down), with a secondary importance given to scoring the most hits.
Starting with "dumb" game agents, Evolutionary Computation (EC) is performed on their traits and eventually the evolved population performs similar to as if they had been pre-programmed to do what you are seeing. However, the difference is that these players have learned on their own, self-adapting to the situation.
Visually, you are seeing the game playing with 2 versus 2 evolved agents; two aircraft on each of two teams playing against each other for survival.
This OpenGL demo was created to demonstrate the use of EC on game characters. So, please pay no attention to the quality of the graphics since that is of no importance to the proof of the effectiveness of the ai technology.
These game characters were evolved using distributed computing (several networked computers) in a relatively short period of time and the results showed that, in only a few generations, the best performers were the ones that learned how to optimize their acceleration, deceleration, turning and targeting abilities, and these were the most effective surviving parents that passed-on their digital "genes" to their children that became the new population for the next round. The game will reset automatically and then a new round begins.
Copyright 2003-2005 Natural Selection, Inc. (R), www.natural-selection.com
Protected under U.S. Patent No. 7,025,675
Developer: Tim Hays
Artificial Intelligence Demos 1:
FlightSim example using Evolutionary Computation (above)
