Novel Game-Theoretic Tools and Solution Concepts with Applications to Network Dynamics and Control

Massachusetts Institute Of Technology, Cambridge MA

Investigators

Abstract

Objective: The objective of this program is to develop new game-theoretic models and tools for multi-agent networks, with emphasis on dynamics and control aspects. The main research thrusts are: (i) structural approaches based on novel decompositions of the space of games (in terms of potential, harmonic, and nonstrategic components); (ii) new solution concepts (exchangeable equilibria), intermediate between Nash and correlated equilibria. Both techniques can be effectively applied to games of practical interest, since they enjoy convenient computational complexity properties. On the application side, the techniques will be applied to practically relevant models in the context of communication and social networks, with emphasis on the design of new distributed dynamics with desirable convergence properties. The proposed program will enable new network protocols and architectures, as well as improved practical operation with performance guarantees. The intellectual merit: The focus is to apply and extend promising early results on structure, characterization, and solution concepts of multiplayer games, as well as to develop new frameworks towards the understanding and systematic analysis of dynamic networks. This requires the combination of ideas and mathematical tools from game theory, network optimization, systems theory, and control theory. Broader impact: The methods and techniques will contribute to the efficient and reliable design and operation of dynamic networks. This is an area of clear technological interest, wide societal impact, and obvious economic implications. Other key features of this proposal include its integration with curriculum development, undergraduate research projects, new programs on computational optimization, and close collaboration with industrial partners and government research labs.