Abstract
In today's rapidly evolving technological landscape, multi-agent systems have become a fundamental paradigm for modeling complex interactions and decision making in various domains. The study of decision making within these systems has gained significant attention across fields like artificial intelligence, robotics, economics, and social sciences. This dissertation explores the intricacies of decision making in multi-agent systems, focusing on the cooperative, non-cooperative and competitive interactions among agents, and addresses the challenge of designing and analyzing multi-agent decision-making algorithms, with the aim to understand how individual agents with diverse capabilities, knowledge, and objectives can collectively achieve desirable outcomes.
For cooperative decision making that involves agents collaborating to achieve common objectives, effective coordination and information sharing are essential. This research investigates decision-making algorithms that facilitate collaboration and improve overall system performance under various challenging scenarios, such as heterogeneity, non-stationarity, and decentralized communication. In domains with limited resources, conflicting objectives, or strategic interactions, non-cooperative and strictly competitive agent behaviors become prevalent. In such settings, agents prioritize self-interest and individual objectives over collaborative efforts. This research analyzes and develops effective decision-making techniques on the system side to account for non-cooperative behaviors and guide agents towards desirable decision-making outcomes.
By understanding the dynamics of decision making in both cooperative and non-cooperative settings, this dissertation aims to enhance the overall performance and efficiency of multi-agent systems across a wide range of applications, and contribute to the advancement of decision-making algorithms, enabling better cooperation and addressing the challenges posed by non-cooperative behaviors in multi-agent systems.
