To each their own – Decision-making in a connected world

Real-time decision-making and coordination using distributed optimization is essential in many engineering problems. Current literature on distributed optimization assumes all agents use model-based optimization for local decision-making. However, in practice, agents might prefer to use different decision-making techniques, such as feedback optimization, or black-box optimization. To account for the heterogenous nature of the subsystems, first part of the talk presents a methodology-agnostic decision-making framework, where each agent can freely choose its local optimization method, yet still take coordinated actions.  

 

Distributed optimization typically requires several iterations/communication rounds to converge to the overall optimal solution. Repeatedly solving optimization problems in each subsystem adds to the overall computational cost, which can hinder real-time decision-making. Noting that the subproblems solved from one iteration to the next differ only by a few variables, the second part of the talk presents a "sensitivity-assisted distributed optimization" framework, where the subproblems are cheaply approximated using parametric sensitivities from one iteration to the next, improving the computation times by orders of magnitude.  

 

Bio: Dinesh Krishnamoorthy is an Assistant professor at the Department of Mechanical Engineering at TU Eindhoven, where he is a part of the Control Systems Technology group. Prior to this, he was a post-doctoral researcher at Harvard University. Dinesh received his PhD in Process Systems Engineering from the Norwegian University of Science and Technology (2019), MSc in Control Systems from Imperial College London (2012), and B.Eng in Mechatronics from the University of Nottingham (2011). Dinesh was also working as a Senior Researcher at Statoil Research centre between 2012-2016, and was also a part-time Senior Data Science consultant for Novo Nordisk R&D (2021). Among others, Dinesh has received the Dimitirs. N. Chorafas Foundation Award, PhD Excellence Award from the European Federation of Chemical Engineers (EFCE), NTNU Faculty of Natural Sciences Best PhD Thesis Award, as well as an IFAC Young author award. His research interests include distributed optimization, optimal control, and data-driven optimization, with applications to energy and biomedical systems.