Research

My research interests are in the intersection of Systems, Control, Optimisation and Networks. The major focus of my research is to develop the theory required for addressing the limitations of Cyber-Physical systems in being applied to modern day applications. Some of the applications that I work on include Heating, Ventilating and Air Conditioning (HVAC) systems, Power systems, Brain Networks and Dynamic pricing.

Research Areas

Distributed Optimisation

Distributed optimization assumes each agent in a network system has access to a part of a global objective function. Each agent is allowed to communicate with its neighbors and thereby arrive at an optimal value of the global objective function. Since each agent has computing and communicating resources, it is possible to adopt the distributed optimization framework for Cyber-physical systems. There are challenges in practically implementing these algorithms, majorly because of the distributed frame work such as reliability of communication medium across agents, communication overhead, security issues, privacy concerns and failure of sensors.

 

My research addresses the following aspects:

  1. Network challenges such as adversarial agents, directed communication and intermittent communication.

  2. Dynamic environment where the cost function and constraints are varying and not known prior.  This occurs in applications such as energy management for a micro-grid with renewable sources, sponsored ads on search engines, airline ticket sales and stock markets

  3. Zeroth order algorithms where the gradient information is not available and hence an approximate gradient is constructed from an oracle which returns noisy function values.

Control and Optimisation in Complex Dynamical Systems

For large scale systems, the classical notion of controllability is not relevant as huge control effort might be required. Hence various energy based metrics have been introduced for \bf quantifying controllability of complex networks. We chose average controllability, which is a measure of controllability in all directions of state space and address the problem of optimising the network topology. 

For scenarios such as the spread of information in a social network and epidemic spreading where time ordering of contacts influences the information spread, time-varying graphs (  temporal networks) are used instead of static graphs.  Problems pertaining to optimising resources such as inputs, feedback interaction were addressed. Polynomial time algorithms were proposed. The proposed algorithms were implemented on data sets corresponding to  functional brain networks.

Data Driven Control

In the age of big data, data driven control methods where system data are used directly to design controllers instead of using a system model has renewed interest in the control community in recent years. I am working on data driven control for networked systems.

Advanced Controls For LIGO

Advanced gravitational-wave detectors like LIGO, are currently the most sensitive instruments capable of detecting passing gravitational waves. To increase the sensitivity and frequency of detection of gravitational waves, continuous efforts are made to suppress the various forms of noise that hinder gravitational wave detection. Our work mainly focuses on reducing the effects of Seismic Noise and Newtonian Noise at LIGO.

  • Broadband Newtonian noise cancellation is achieved by optimizing the position of sensors for the desired seismic frequency range. The sensor placement optimization problem is formulated as a multi-objective optimization problem and efficient methods are developed to obtain a smart Pareto optimal solution. The proposed method improved the Newtonian noise estimation in comparison with the single objective case.

  • Proposed modern techniques for controlling the multi-stage pendulum systems at LIGO which is an upgrade to the existing control system onsite.

Research Grants

  1. Optimization and Control of Complex Dynamical Systems, funded by DST, 2017-2022.

  2. Ligo R & D for India - Advanced controls, 2018-2022.

  3. New Faculty Initiation Grant, funded by IIT Madras, 2019-2021

  4. New Faculty Seed Grant, funded by IIT Madras, 2022-2025