Strategies for Evolving Diverse and Effective Behaviours in Pursuit Domains

Abstract

Evolutionary algorithms have a tendency to overuse and exploit particular behaviours in their search for optimality, even across separate runs. The resulting set of monotonous solutions caused by this tendency is a problem in many applications. This research explores different strategies designed to encourage an interesting set of diverse behaviours while still maintaining an appreciable level of efficacy. Embodied agents are situated within an open plane and play against each other in various pursuit game scenarios. The pursuit games consist of a single predator agent and twenty prey agents, with the goal always requiring the predator to catch as many prey as possible before the time limit is reached. The predator's controller is evolved through genetic programming while the preys' controllers are hand-crafted. The fitness of a solution is first calculated in a traditional manner. Inspired by Lehman and Stanley's novelty search strategy, the fitness is then combined with the diversity of the solution to produce the final fitness score.The original fitness score is determined by the number of captured prey, and the diversity score is determined through the combination of four behaviour measurements. Among many promising results, a particular diversity-based evaluation strategy and weighting combination was found to provide solutions that exhibit an excellent balance between diversity and efficacy. The results were analyzed quantitatively and qualitatively, showing the emergence of diverse and effective behaviours.Writing a thesis during a pandemic is no easy task. As the lines between work, school, and life became blurred, focusing on what needed to be done became increasingly difficult.Fortunately, I had the assistance of many wonderful people throughout my research. It's thanks to all of you that I was able to complete this thesis. First, I'd like to thank my supervisor, Dr. Brian J. Ross, for recommending that I consider enrolling as a Master of Science. I am forever grateful for him taking me under his wing, teaching me what he knows about the concepts used throughout this research, and providing me with continual support and assistance. I'd like to thank Illya Bakurov, Michael Gircys, and Doug Ord for their contributions toward the implementation of my systems. Their experience in genetic programming, multi-objectivization, and game-development saved me plenty of time during the development stages of this research. I'd like to thank my friends, Mitchell Clark and Jordan Maslen. The three of us met within our first month as post-secondary students and we road out the entire journey together. I have absolutely no idea where I would be today had I not met you two, so, for the many all-nighters and crunch sessions, thank you. I'd like to express heartfelt gratitude to my significant other, the most incredible woman I have ever met: Sara Testani. Your endless support and encouragement has helped me in so many ways throughout this thesis, and you've helped me find myself throughout this incredibly stressful pandem...