Chris Metzler Receives $450K Air Force Young Investigator Award

The three-year project seeks to develop artificial intelligence-based multimodal sensor fusion algorithms that are fully self-supervised and do not require training data
Descriptive image for Chris Metzler Receives $450K Air Force Young Investigator Award

A University of Maryland expert in computational imaging has received funding from the Air Force Office of Scientific Research (AFOSR) to develop novel algorithms for fusing multimodal sensing data.

Chris Metzler, an assistant professor of computer science with an appointment in the University of Maryland Institute for Advanced Computer Studies (UMIACS), was one of 36 researchers in the U.S. to receive a $450,000 grant from AFOSR as part of its Young Investigator Program.

The competitive program—175 proposals were submitted this year—was established to foster creative basic research in science and engineering, enhance early career development of outstanding young investigators, and increase opportunities for early-stage scientists working on topics of interest to the Air Force.

Metzler’s three-year project, titled “Adversarial Sensing: A Self-Supervised Approach to AI-Based Multimodal Sensor Fusion,” seeks to develop artificial intelligence-based multimodal sensor fusion algorithms that are fully self-supervised and do not require training data.

These algorithms should be able to reconstruct time-varying scenes using heterogeneous measurements captured over varied and challenging environmental conditions, Metzler says.

To achieve this goal, he plans to parameterize time-varying scenes using scene representation networks; design differentiable multimodal signal propagation models; develop improved multimodal classification algorithms; and combine these components into an adversarial sensing framework that uses distribution matching to perform multimodal sensor fusion.

“We anticipate this project will lead to significantly improved situational awareness and intelligence gathering capabilities for Air Force personnel,” Metzler says.

Time-varying neural signal representations—as opposed to fixed pixel or voxel representations—will allow analysts to effectively fuse information from measurements captured at different times, resulting in a multiplicative improvement in signal-to-noise ratios, Metzler says.

In addition, the overall adversarial sensing framework Metzler will develop can enable effective multimodal sensor fusion in situations where one does not have access to training data. This capability is particularly important, he says, because sensing is arguably most critical when identifying never-before-seen signals and threats.

Metzler says the AFOSR funding has provisions to support graduate students and he encourages UMD grad students interested in computer vision and AI to reach out to him.

Before coming to Maryland, Metzler was an Intelligence Community Postdoctoral Fellow in the Stanford Computational Imaging Lab. He was also an NSF Graduate Research Fellow, a DoD NDSEG Fellow, and a NASA Texas Space Grant Consortium Fellow. He received his doctorate in electrical and computer engineering from Rice University in 2019.

 

The Department welcomes comments, suggestions and corrections.  Send email to editor [-at-] cs [dot] umd [dot] edu.