My research is in the general area of analysis, modeling, and control of multi-agent systems. A major part of my research program is to understand how collective behavior, such as cooperation and coordination, emerges from interactions among individuals in complex systems, which is proving to be an increasingly important research direction in the field of artificial intelligence. My research has been based on mathematical analysis (including stochastic process, matrix analysis, nonlinear systems, graph theory, and game theory), computer simulation techniques, and empirical studies. The main research thrust can be categorized into the following groupings with some overlap:

Evolutionary Game Theory, Collective Behavior

Collective intelligence emphasizes that systems can rely on cooperation and coordination among individuals to achieve goals that are impossible by any individual alone to achieve. In collective intelligence, one of the most cutting-edge questions is how and when cooperation and coordination emerge and maintains, especially when individuals face conflicts between their own and collective interests and simultaneously have the cognitive ability to make their own behavioral decisions. Despite the fact that large-scale cooperation maximizes the collective benefits, the cooperative act always decreases one’s own utility relative to non-cooperative competitors. Classic non-cooperative game theory based on the assumption of perfect rationality has also predicted a convergence towards the Nash Equilibrium state, i.e., the collapse of cooperation.

My research in this part is to explore various mechanisms that support the evolution of cooperation, including economic factors like joint liability, biological factors like demographic stochasticity, social factors like information uncertainty, and ecological factors like environmental feedback.

Network Theory, Data-driven System Analysis and Synthesis

Complex systems are often structured, and individuals geographically close or socially related are connected and more likely to interact with each other. Many empirical studies have evidenced that different systems differ much in structures, which often affects the evolutionary trajectories of systems and the feasibility to control them. Understanding how the structure of a system affects the emergence and maintenance of collective cooperation remains challenging, especially in the absence of mathematical results of general structures. The most practical and promising application in this area is by slightly modifying the local structure to manipulate and control the system to desired states.

My research in this part is to collect and analyze empirical datasets and infer real-world social networks, such as multilayer networks, directed networks, and dynamic networks.

Game-based Learning and Control, Hybrid Systems of Humans and Machines

Another part of my research is game-based control. It is widely accepted that it is impossible for one player or a group of players to unilaterally control others’ utilities in games since utilities are jointly determined by all players’ strategies. Inspired by Zero-Determinant strategies, we propose a new mathematical framework to study ruling strategies (with which a player unilaterally makes a linear relation rule on his own and opponents’ utilities) in repeated games with an arbitrary number of actions or players, and arbitrary continuation probability. We establish an existence theorem of ruling strategies and develop an algorithm to find them. The proposed mathematical framework also enables the search for collaborative ruling strategies for an alliance to control outsiders.

Skills

Coding: C++, Matlab, Julia

Languages: Mandarin Chinese (Mother-tongue), English (Working proficiency)