My research centers on estimation in autonomous systems. I am particularly interested in:

• Bayesian Estimation: Developing advanced Bayesian methods for state estimation in dynamic environments.
• Data Association: Improving techniques for associating sensor data with known objects in the environment.
• Time Series Classification: Applying machine learning to classify time series data for fault detection in aerospace systems.

Current Projects

mWidar Imaging System Platform

Our research develops intelligent detection and tracking algorithms for the revolutionary mWidar (microwave detection and ranging) system—a cost-effective alternative to traditional radar that uses basic antennas and advanced software instead of expensive hardware to create real-time images of moving objects. Unlike conventional radar systems that require costly components and scan one direction at a time, mWidar captures images of entire areas simultaneously at thousands of frames per second, making it ideal for applications like counter-drone defense and future hypersonic vehicle tracking. Working with realistic mWidar simulations, we tackle three key challenges: detecting objects in microwave images using computer vision and deep learning methods that work even when targets are hidden or interfering with each other; associating detected objects across time to maintain consistent tracking; and predicting target movements using mathematical filters like Kalman filters and particle filters. This research represents a shift from hardware-intensive to software-intensive sensing systems, potentially making advanced surveillance capabilities more accessible and affordable while maintaining the high performance of traditional expensive radar installations.

Time Series Classification for Satellite Fault Detection – Partnership with Trusted Space Inc

Our research develops intelligent systems that help satellites detect and diagnose their own problems automatically. Working in partnership with Trusted Space, Inc., we’re applying advanced machine learning algorithms to analyze spacecraft sensor data in real-time, enabling satellites to identify issues like power failures, attitude control problems, or thruster malfunctions without waiting for ground control intervention. Using realistic spacecraft simulations, we train these systems to recognize fault patterns across multiple satellite subsystems, with the goal of making future space missions more autonomous and reliable. This work could significantly reduce the need for constant human monitoring of satellites and improve response times when problems occur, ultimately leading to safer and more efficient space operations.