REU-CAAR Student Researchers Earn Acceptance to International Robotics Conference

Student researchers who participated in the University of Maryland’s Research Experiences for Undergraduates (REU) program, known as REU-CAAR, have had their work accepted to the 2026 IEEE International Conference on Robotics and Automation, scheduled for June 1 through 5 in Vienna, Austria.

The paper was authored by Yufan Chen, an undergraduate student at the University of Southern California; Arnav Bali, a student at South River High School in Edgewater, Maryland; UMD alumni Angela Liu (M.S. ’24, computer science) and Laura Zheng (Ph.D. ’25, computer science); and Distinguished University Professor of computer science Ming Lin.

Their paper, SSQA: Sibling-Selective Quadtree Attention for Hierarchical Modeling in Perception Tasks, presents a method to help robots and autonomous systems interpret their environment more efficiently. These systems rely on software to recognize roads, obstacles and other important details in their surroundings, but this process can require substantial computing power and energy.

The researchers developed a method that enables these systems to focus on the most important regions of an image first, thereby accelerating the process without reducing accuracy. This could make it easier for robots, drones and autonomous vehicles to operate effectively, especially on devices with limited processing capacity.

Chen said the acceptance reflects months of iterative work refining the method and validating its performance.

“It’s both rewarding and kind of a relief to see research that involves many iterations, challenges and refinements really reach something,” Chen said. “It’s especially meaningful since it’s my first paper, and it’s completely worked out through our own effort, even from the initial proposal.”

Chen said improving efficiency in perception models could help make robotics and AI systems more practical in real-world environments.

“A lot of AI data centers may have soaked up half a city’s energy,” Chen said. “By reducing computational and energy requirements, this approach might help enable both cost-efficient and robust perception on resource-constrained platforms such as drones, robots and autonomous vehicles.”

The paper will be presented at ICRA, one of the leading international conferences in robotics and automation. The 2026 conference received 4,947 submissions from 86 countries and accepted 1,882 papers, representing an acceptance rate of 38.04%.

The project originated through REU-CAAR, a summer research program that brings undergraduate and select high school students to UMD to work alongside faculty and graduate mentors. Participants conduct research in small teams across areas including artificial intelligence, machine learning, quantum simulation and parallel algorithms. Most participants receive a stipend and housing during the program.

Bill Gasarch, a professor of computer science who has overseen UMD’s involvement with the REU program since 2013, said the program is designed to give students hands-on experience with the research process.

“The program gives students a chance to do research from soup to nuts: reading papers, working on problems whose answers are not known, working with other people and producing papers,” Gasarch said.

The program also includes students at different academic stages, including select high school participants. Bali, a high school senior who contributed to the project, said the experience provided insight into how research develops over time.

“As a high schooler, I have been able to see what real research looks like day to day,” Bali said. “It’s helped me build stronger habits around experimentation, documentation and communicating what I am doing and why.”

Bali, a three-time World of 8 Billion contest winner and a staunch advocate for addressing pressing issues with sustainable solutions, said the collaborative environment helped him refine his ability to communicate technical ideas.

“Our discussions helped me learn how to explain ideas clearly, test assumptions and iterate in a focused way while still being creative,” Bali said. “It has given me confidence that I can contribute in a research environment and keep growing through feedback.”

Lin said the students’ progress reflects the research opportunities available through the program and ongoing work in the field.

“The depth of their research is not much different from the research by starting Ph.D. students, which is reflected in their accepted paper in IEEE ICRA 2026,” Lin said. “And we’re already discussing what’s next and how to extend this work.”

—Story by Samuel Malede Zewdu, CS Communications