Research

My current research interests center on the control of autonomous learning systems, which lies at the intersection of supervisory control, learning, robotics, and autonomous systems. The overarching goal is to develop autonomous systems that can adapt to uncertain and changing environments while preserving formal guarantees of safety, correctness, and task accomplishment. The foundation of this research is supervisory control theory and discrete-event systems. My early work focused on the synthesis and implementation of supervisors for manufacturing systems, real-time computing systems, and autonomous vehicles. This research established a rigorous framework for decision-making in complex event-driven systems and demonstrated how formal supervisory architectures can enforce desired system behavior under operational constraints.

Building on this foundation, my research expanded into learning and adaptive control. By integrating neural networks, iterative learning methods, and adaptive control techniques into robotic systems, I investigated how controllers can improve their performance through experience while maintaining stability and reliability. These studies represented an early effort to bridge model-based control and data-driven adaptation. Another research direction focuses on autonomous and multi-agent robotic systems operating in uncertain environments. This includes robot navigation, formation control, cooperative robotics, and human-inspired decision-making strategies. The objective is to enable autonomous agents to make intelligent decisions, coordinate with one another, and accomplish complex tasks in dynamic settings where complete system models may not be available.

Although the methodologies employed have evolved over time and extended beyond the traditional supervisory control framework, a common thread throughout this research has been the use of control-theoretic principles to guide decision-making, adaptation, and autonomy in complex systems. More recently, these research threads have converged into a broader agenda on learning-enabled autonomy. The central idea is that supervisory control provides a principled framework for constraining and guiding learning processes, thereby combining the adaptability of modern machine learning with the safety, reliability, and explainability of formal methods. Current work investigates constrained exploration, automata-based monitoring, and model-guided learning architectures for autonomous systems. Future research will focus on deeper integration of supervisory control principles with learning-based decision-making methods to enable autonomous systems that are both adaptive and formally trustworthy.