About Our Research

We are interested in controlling dynamic systems so they behave the way we want. To achieve this, we develop fundamental theories, design efficient numerical algorithms, and test our ideas in the real world. Our research bridges deep theoretical insights with practical implementation.

We tackle several critical challenges in control and decision-making. These include achieving real-time computation for complex, high-dimensional systems; making robust decisions with inaccurate or insufficient information, often under various types of uncertainties; and ensuring theoretical guarantees for safety and performance — a necessity in mission-critical applications.

Research Areas

Scalabile Algorithms

Our lab develops scalable control algorithms for high-dimensional, multi-agent systems operating in dynamic and uncertain environments. As these systems grow in complexity, they must make real-time decisions using limited onboard resources—including constrained processors, strict latency requirements, limited memory, and local sensing and communication capabilities. These constraints present major challenges for traditional control approaches, which often rely on centralized computation, global state information, or oversimplified dynamics that fail to reflect the realities of embedded autonomy. Bridging the gap between growing system complexity and practical deployment requires new computational frameworks that are both efficient and robust under these limitations.

We aim to develop a real-time, decentralized decision-making framework that is inherently compatible with the constraints of embedded platforms. Our vision is to enable scalable, reliable coordination in high-dimensional systems using only onboard resources, by integrating principles from control theory, machine learning, and mathematical foundations.

Relevant Publications:

• ZSG-MPPI: Robust Model-Predictive Path-Integral Method for Disturbance Handling, IEEE Robotics and Automation Lettres, 2025, Accepted.

• Efficient Computation of State-Constrained Reachability Problems Using Hopf-Lax Formulae, IEEE Transactions on Automatic Control, 2023.

• Real-time Robust Receding Horizon Planning using Hamilton-Jacobi Reachability Analysis, IEEE Transactions on Robotics, 2022.

Safe Automation

Ensuring safety and reliable performance in autonomous systems remains a central challenge due to uncertainties arising from imperfect models, sensor noise, and dynamic or unpredictable environments. Traditional model-based control methods offer formal safety guarantees but often rely on system models and structured assumptions that limit their applicability in real-world settings. Meanwhile, learning-based methods are more flexible and data-driven but typically lack theoretical assurances on safety and performance. Additionally, many existing approaches struggle to meet real-time constraints, scale to high-dimensional systems, or handle partial observability and intermittent communication—factors that are essential for embedded and distributed autonomy.

This research direction focuses on establishing foundational principles for ensuring both safety and performance in autonomous systems operating under uncertainty. Our aim is to explore scalable, theoretically grounded approaches that integrate tools from control theory, machine learning, and applied mathematics to enable reliable decision-making with safety and performance guarantees in complex, dynamic environments.

Relevant Publications:

• Certifiable Reachability Learning Using a New Lipschitz Continuous Value Function, IEEE Robotics and Automation Letters, 2025

• Solving Reach-Avoid-Stay Problems Using Deep Reinforcement Learning, Under Review

• On the Hamilton-Jacobi Reachability of Systems with Input Delay, Under Review