CYBER-PHYSICAL SYSTEMS AND ML LAB

Our lab sits at the intersection of artificial intelligence, physical systems, and human health. We develop machine learning methods that bridge the digital and physical worlds, creating systems that sense, understand, and respond to real-world conditions in meaningful ways.

The work here ranges from detecting subtle health changes through everyday technology to coordinating teams of robots in challenging environments. Each project pushes the boundaries of what intelligent systems can achieve when they interact directly with the physical world around them.

Health Monitoring Through Visual Analysis

One major research direction involves extracting physiological signals from facial videos. Remote photoplethysmography, or rPPG, lets us measure heart rate and blood flow patterns without any physical contact. The technology works by detecting tiny color changes in skin that occur with each heartbeat.

We're particularly interested in what these signals reveal about cognitive health. Subtle changes in cardiovascular patterns can indicate underlying impairments long before they become obvious through other means. By studying the relationship between daily activities and these physiological markers, we aim to develop early warning systems for cognitive decline.

The challenge lies in making this technology work in real-world conditions. Lighting changes, movement, and camera quality all affect the signal. Our research tackles these problems through robust algorithms that can extract reliable measurements even when conditions are far from ideal.

Real-Time IoT Deployments

Taking rPPG from the laboratory to practical use requires building complete systems that operate in real time. We develop Internet of Things platforms that can capture video, process it on the spot, and deliver health insights without sending sensitive data across networks.

These systems balance competing demands. They need enough computing power to run sophisticated algorithms, but they also need to be small, affordable, and energy-efficient. Our work focuses on optimizing this tradeoff, finding ways to deploy advanced machine learning on resource-constrained devices.

Sports Analytics

Athletic performance generates enormous amounts of data. We apply machine learning to extract meaningful patterns from this information, helping coaches and athletes understand what separates good performance from great performance.

The analysis goes beyond simple statistics. We look at movement patterns, decision-making under pressure, and how different factors interact to produce outcomes. Machine learning excels at finding these complex relationships in ways that traditional analysis might miss.

Resilient Multi-Agent Robotics

When multiple robots need to work together in contested or unpredictable environments, communication becomes critical. Our research develops networks that keep ground and aerial robots coordinated even when connections are unreliable or under attack.

The key challenge is building systems that gracefully handle failure. If one robot loses contact, the team needs to adapt. If part of the network goes down, traffic needs to reroute. We create algorithms that optimize resource use while maintaining fault tolerance, ensuring missions can continue despite disruptions.

These contested environments might involve jamming, physical obstacles, or rapidly changing conditions. The systems we build need to recognize problems, adjust strategies, and maintain coordination without constant human oversight.

Model-Based Systems Engineering for Cybersecurity

Securing complex systems requires understanding how all their parts interact. We use model-based systems engineering approaches to map out these relationships and identify vulnerabilities before they can be exploited.

This work involves creating formal models of system behavior, then using those models to reason about security properties. Where might an attacker find a way in? How could a compromise in one component affect others? By answering these questions during design, we build more secure systems from the ground up.

Funding and Partnerships

Our research receives support from the Naval Air Systems Command through NAVAIR funding. This partnership focuses on practical applications for naval aviation systems and operations.

We also participate in a five-year cooperative agreement between the University of Maryland and the US Army Research Laboratory. The ArtIAMAS program, which stands for AI and Autonomy for Multi-Agent Systems, brings together researchers across institutions to advance the science of coordinated autonomous systems.

These partnerships ensure our work addresses real operational needs while advancing fundamental understanding. The problems we tackle come from actual mission requirements, and the solutions we develop get tested in realistic scenarios.

Looking Forward

The convergence of machine learning and cyber-physical systems opens new possibilities across domains. Health monitoring becomes less intrusive and more continuous. Robot teams become more capable and resilient. Security improves through better understanding of system complexity.

Our lab continues pushing these boundaries, developing the algorithms, systems, and methods that will define the next generation of intelligent physical systems. The work requires expertise across multiple disciplines, from computer vision and machine learning to networking, robotics, and human factors.

Each project contributes to a larger vision: creating intelligent systems that enhance human capabilities, protect human health, and operate reliably in the complex, unpredictable conditions of the real world.