Evolutionary Multiobjective optimization for automatic generation Of Neural game controller

This research presents the result of implementing evolutionary algorithms towards computational intelligence in Tower Defense game (TD game). TD game is a game where player(s) need to build tower to prevent the creeps from reaching their based. Penalty will be given if player losses any creeps during gameplays. It is a suitable test bed for planning, designing, implementing and testing either new or modified Al techniques due to the complexity and dynamicity of the game. In this research, Genetic Algorithm (GA) is implemented in evolving the required controllers along with two different neural networks used namely: (1) Feed-forward Neural Network (FFNN) and (2) Elman Recurrent Neural Network (ERNN). The NN are used as tuner of the weights. ANN determines the placement of the towers and the fitness scores are calculated at the end of each game. A new fitness function has been proposed as well in this research. As a result, it is proven that the implementation of GA towards FFNN is better compared to GA towards ERNN