| Summary: | This thesis explores how to use Multi-Objective Evolutionary Algorithms (MOEA)to solve problems that are not explicitly defined as multi-objective problems. Aneuroevolution technique consisting of combining a multi-objective evolutionaryalgorithm called NSGA-II and artificial neural networks (ANN) based on Neu-roEvolution of Augmented Topoligies (NEAT) were used to develop a systemthat created controllers for a version of the Super Mario Bros game called MarioAI. Experiments were conducted to measure different ways to define objectivesfor MOEAs in Mario AI, how using these objectives as a basis for a scalar fitnessfunction would affect a genetic algorithm and to examine how to use ensemblesto combine individuals of a pareto front into a single controller that would beable to display the strengths of all of the individual controllers.The results show that adding sub-goals as objectives together with the main goalcould have a positive effect for a MOEA and that the same sub-goals could alsogive a positive effect when applied to the scalar fitness of a genetic algorithm.It is however not trivial to decide which sub-goals to use, as most of the chosenobjectives were found to have a negative impact on the controllers, even whenselected based on the authors? expert knowledge about the game domain. Usingbasic behaviours that the controller has to use in order to play well as objectiveshad a negative effect on the controllers and the controllers were able to learnthese behaviors even without using them as objectives.
|
|---|
