Sivaranjani Seetharaman, Assistant Professor of Industrial Engineering

This past May, the United States Department of Defense awarded grants to four projects led by Purdue University researchers as part of the Defense Established Program to Stimulate Competitive Research (DEPSCoR).

Chosen from among 115 white paper submissions, Seetharaman is one of two Purdue IE faculty to receive this year’s prestigious award. Her project, titled “Operational Resilience to Extreme Events in Networked Dynamical Systems through Realtime Adaptation,” aims to develop new mathematical tools to control complex system behavior in response to unexpected events and disturbances.  The team plans to use control algorithms to create adaptable, scalable designs to achieve “operational resilience” in networked dynamical systems. 

The DEPSCoR Award was given to 25 teams across 15 states to bolster research benefitting the Defense Department at previously underutilized universities. Among eligible universities, Purdue achieved the greatest number of grant recipients. 

Congratulations to Seetharaman and her team! You can read the full project description below.

Networked dynamical systems such as infrastructure networks, supply chains, biological networks, social networks, and so on, are central to our lives. Yet, they often fail catastrophically when faced with unexpected large disturbances. The payoff if we can ensure that such catastrophic failures do not happen is huge in a variety of contexts. Currently, such a theory is lacking. There are some relevant tools from both systems-theory, such as adaptive control, as well as from AI. However, they do not scale well, lack guarantees on resulting solutions, cannot consider far-from-equilibrium behavior, and are fragile to adversarial decision makers. In most practical settings, assuming away such complexities will be far too limiting. In this context, our project seeks to solve the problem of ensuring operational resilience of networked dynamical systems to extreme events/shocks that push the system far-from-equilibrium. The key idea behind our work is that control algorithms that capture and utilize physical properties or constraints such as energy dissipation, conservation laws, or symmetries, can lead to scalable designs that can adapt to large disturbances, and ensure desired operational resilience in networked dynamical systems. The project marks a basic deviation from the literature and makes the entire problem setting applicable to far more practical settings through development of new mathematical tools to learn physics-informed and control-relevant models of large-scale systems, analyze far-from-equilibrium operation, and control these systems using model-based systems theory.

Source: Siva Seetharaman, sseetha@purdue.edu