IEEE Journal Publishes Shirantha Welikala’s Innovative Findings on Decentralized Decision-Making
Stevens researcher has teamed up with Notre Dame peers to demonstrate how trading central operations hubs for local operations units can be a smarter way to manage critical networks
Imagine a city’s power grid. Because one central system typically manages all the subsystems, one issue in that hub could spark a massive blackout. The same is true for transportation networks, where one problem in a central hub can instantly result in citywide gridlock. In communication systems and supply chain networks, centralized control system failures can lead to large-scale outages and shortages affecting millions.
"When everything relies on a single control center, one failure can bring down the entire network," noted Shirantha Welikala, assistant professor in the Department of Electrical and Computer Engineering at Stevens Institute of Technology. "And cybersecurity failures at control centers can make the entire network vulnerable to cyberattacks."
Working with colleagues at Stevens and the University of Notre Dame, Welikala is testing how spreading analysis and control systems design processes across local decision-making units can make these networked systems faster, more resilient and better prepared for evolving demands, threats and opportunities.
Their groundbreaking study, "A Decentralized Control Synthesis Approach for Networked Systems With Arbitrary Interconnections," was recently published in the IEEE Transactions on Automatic Control journal. The article looks at the benefits of sharing analysis and planning across local units in a network. It also explores better ways to design and control large networked systems, improve communication among the individual units and turn centralized analysis and control design tasks into simpler, decentralized ones.
Spreading the risk
In contrast to single control centers that have full knowledge of the entire system, decentralized systems distribute control and expertise across smaller units. Each piece makes independent decisions while coordinating with neighboring network subsystems. If one area goes offline, other sections can quickly adjust, minimizing and even preventing full-scale collapse.
"The beauty of decentralization is that it reduces operational delays and enhances stability," said Welikala. "It reduces the threat of cyberattacks or system-wide malfunctions. Each part of the system can respond instantly to changes, making operations smoother and more resilient."
In addition, decentralization allows the network to scale for growth, seamlessly integrating new components without overburdening a central hub.
Welikala’s research goes beyond pinpointing the problem to finding solutions. Leveraging mathematical tools and control theories, he and his team are developing scalable methods for analyzing and designing control systems for power microgrids, autonomous vehicle platoons, multi-robot systems and even supply chain networks.
They have also introduced a novel, structured mathematical framework for systematically decentralizing traditional centralized control solutions in general networked systems.
Reaping the rewards
In addition to theoretical proofs, Welikala and his team have simulated and validated the effectiveness of their proposed solution. Now, they’re working to move it into the real world, where they can better understand its function and benefits and continue introducing improvements in efficiency and resilience.
They are seeking collaborations with industry and academia partners to apply their ideas and explore novel adjacent research avenues. Two graduate students, one at Stevens and one at Notre Dame, have chosen to explore applications in smart grids, platoons and multi-robot systems.
"We developed our approach using a general networked system model, so it is widely applicable for an extensive range of real-world networked systems," Welikala said. "It can be applied to existing control frameworks with minimal modifications. We provide insights on how to adapt to other networked systems and processes in a wide range of networked systems. I hope these tools and techniques will be used in future smart networked systems, particularly for enhancing their analysis and control systems design tasks."
Keeping it real
The conference version of this work earned the Best Paper Award at the 30th Mediterranean Conference on Control and Automation.
It’s a testament to Welikala’s deep passion for solving complex problems. With a background in multi-agent and networked systems control engineering in Sri Lanka and the U.S., he has dedicated his career to making large-scale infrastructure smarter and more efficient. Since joining Stevens in 2023, he has also become active on the Department of Electrical and Computer Engineering Curriculum Committee and the Charles V. Schaefer, Jr. School of Engineering and Science Dean’s Faculty Advisory Council.
Welikala was drawn to electrical and computer engineering when he realized that his favorite subjects — mathematics, physics and computing — could combine to solve critical real-world problems.
"I’ve always been fascinated by how theoretical principles translate into practical solutions for complex engineering problems," he said. "I love using just a pen and paper to develop provably precise and highly practical solutions to challenging engineering problems. Above all, I enjoy seeing a mathematically rigorous solution work successfully in a real-world setting."