skip to main content

CSL student duo awarded Qualcomm Innovation Fellowship

10/6/2020

Allie Arp, CSL

The Qualcomm Innovation Fellowship is a new program, and one of the most selective. Only teams of graduate students from North America’s top 24 universities in engineering can apply. This year, of the 188 eligible teams, 13 were selected, including a duo from the University of Illinois Urbana-Champaign.

Suraj Jog and Junfeng Guan, both electrical and computer engineering students, were selected for their work on enabling self-driving cars to see in fog a
Junfeng Guan
Junfeng Guan
nd bad weather by enhancing the resolution of radar imaging.

“This a very prestigious fellowship and it speaks to the novelty and importance of Suraj and Junfeng’s research,” said Haitham Al-Hassanieh, the pair’s adviser and CSL faculty member. “There is significant interest in industry towards achieving a high level of autonomy in self-driving cars, i.e., the cars can safely operate in bad weather and low-visibility conditions. However, the technology is not there yet. Suraj and Junfeng are using AI to enhance the resolution of radar images and have managed to build the first system that can achieve such high-quality results.”

The cameras and lidar sensors that self-driving cars use to navigate depend on visible light. As a result, there are problems when bad weather, like fog or rain, drastically reduces visibility. Jog and Guan have developed a system that applies super-resolution algorithms to low-resolution radar. The Hawkeye Millimeter Wave Radar system uses artificial intelligence to reconstruct what is in front of a vehicle and depict the scene on a screen so the human driver can view whatever is in front of the car, even if he or she can’t see it through the windshield.

Views of a parking garage in normal conditions and in foggy conditions, along with images of the duo's sensors.
Views of a parking garage in normal conditions and in foggy conditions, along with images of the duo's sensors.
Suraj Jog
Suraj Jog
The pair believe that by building their own end-to-end system, and proving it could work through experimentation in real-life settings, they gave their fellowship application a big advantage.

“One of the biggest things is that we built a system and showed that it worked in the real world. We even bought a fog machine to emulate realistic low-visibility conditions,” said Jog. “We showed that even through the dense fog, our system could sense and accurately depict the scene in front of us. This showed we made fundamental contributions to the community and enabled a new sensing capability that will prove critical in realizing the vision of fully autonomous driving.”

Other strengths of their project included its interdisciplinary nature, as it made contributions at the intersection of communications, sensing, and artificial intelligence. The two hope to build on those contributions thanks to the fellowship.

“The fellowship is great encouragement and a great honor,” said Guan. “Based on this recognition, we’re able to collaborate with Qualcomm on their commercial self-driving platforms and work with their researchers to take our ideas and push this project forward onto consumer hardware.”

The project was also presented at the 2020 Conference on Computer Vision and Pattern Recognition, and a recording of the full presentation is available.