'Computational Intelligence and Game Design'
Center for Computer Games Research IT university of Copenhagen, Denmark.
Julian demonstrated an A* Algorithm developed for playing a Super Mario Bros Game, which was derived as part of a competition. another such example was the Torque racing game, which also had algorithms developed for it and the most successful, the wons who won the race, where the winners.
The algorithms were tested and were trained using a perceptron, a learning feedback program which takes inputs, generates an output and the most valued inputs get the 'weights', the importance, increased. this way, the most successful methods are evolved from initial inputs.
Representation of Game content > Mutation operator > Fitness function
What this means, it that the inputs are taken, the weights applied and various mutations applied, like in natural evolution, and the fitness of each method, the successfulness, is taken
Level, tracks and so on are also evolving content. Parameters from user input are tkane and fed into the fitness function based on the experience of the player.
No comments:
Post a Comment