Decision-Centric Foundations for Modeling and Analysis of Complex Networked Systems

Purdue University, West Lafayette IN

Investigators

Abstract

The research objective of this award is to establish foundational techniques for modeling and analyzing the evolutionary dynamics of complex networked systems in terms of node-level agents' decisions. The research objective will be achieved through an integration of discrete-choice random-utility theory and network science-based approaches to model the evolutionary characteristics of complex networked systems. The research plan consists of three parts: a) inferring agents' unobserved payoffs from network structure data, b) inferring agents' unobserved payoffs in strategic interaction scenarios, and c) determining ways to improve network performance by influencing node-level decisions. Specific examples of complex networked systems, including air-transportation networks and autonomous system level Internet, will be used for validation. The research will result in novel approaches to model network evolution resulting from decisions made by independent or competing entities, and to evaluate mechanisms for steering the evolution towards higher performance, such as robustness to node failure and targeted attacks. The results of this research hold promise for accurate system performance prediction, and for forecasting how the complex networked systems would evolve in the future. The resulting approaches would enable the development of better surrogate models of networks for efficient design of processes and protocols on networked systems. The approaches will also enable policy and incentive design to guide the restructuring of existing networks for improved performance. The outcomes of the research will be integrated into graduate and undergraduate courses. The results will be disseminated through industrial partners, policy makers, open-source software tools, journal publications and conference presentations.