Research Area: Games


  • Starships: A dynamic scaling heuristic is applied to Neural Fitted Q-Iteration in order to learn a strategy for a simple space ship combat simulation. [Video] pdf
  • Starcraft: Neural Fitted Q-Iteration is employed to control attack and retreat patterns of small groups of units in the popular Starcraft strategy game.
  • Torcs: Learning to control the steering behaviour of a racing car in the TORCS simulation using Neural Fitted Q-Iteration. [Video]
  • Civilization IV: A neural network is trained to choose among a predefined set of strategies based on the current game situation.