Weizhe Chen, a Ph.D. student in the Luddy School’s Intelligent Systems Engineering department, won the Best Student Paper Award at the Robotics: Science and Systems conference.
The award was announced during the recent Robotics: Science and Systems conference in New York City. The conference – considered the most prestigious in the robotics field – brings together researchers from all areas of robotics from around the world.
“I am thrilled and grateful!” Chen said. “The award really means a lot to me. This project took a long time, and had a lot of setbacks when developing and trying other methods. The award makes me feel that all the detours and hard work are finally acknowledged.”
The paper, “Attentive Kernel for Information Gathering,” took two years to research and experiment. It proposes a new and better way for Autonomous Vehicles (unmanned, robot-controlled vehicles for land, air, and water) to collect and analyze data during terrain mapping.
Chen works with Professor Roni Khardon in computer science and Assistant Professor of Intelligent Systems Engineering Lantao Liu, who runs Vehicle Autonomy and Intelligence Laboratory (VAIL). He is interested in the intersection of robot learning and decision making under uncertainty and develop algorithms to help robots make informed decisions by gathering information in ever-changing environments.
The paper was the result of analysis, observation, field experimentation, dedication and even muscle (the vehicle and equipment were carried to a distant lake).
“Field experiments are challenging,” Chen said, “and I’d like to thank ISE Ph.D. students Durgakant Pushp and Mahmoud Ali for their help during the field experiment.”
The paper proposes replacing stationary kernels (which assume the variability of the environment remains constant everywhere) with the new attentive kernel, which can learn and capture phenomena that are smooth in some places, and highly varying elsewhere, and in this way help robots focus on areas of high variability.
“This is an exciting interdisciplinary project that requires innovations in multiple fields and our paper was long in the making,” Khardon said. “It is rewarding that Weizhe’s persistence and hard work is recognized.”
Every year, the conference selects around 70 of the best papers from all robotics research directions in every engineering department.
“The award is a great recognition for our research, and it means we have achieved a good milestone,” Liu said. “It also encourages us to pursue a leading role for this research topic.”
The paper reflects ongoing research to help robots learn the way humans and animals do.
The goal is to endow robots with the ability to intelligently explore the environment for building a better “mental model” that matches the external reality. In robotics, this research topic is called Robotic Information Gathering, which aims to guide the robot to collect informative sensing data for building accurate models of the environment.
“Intuitively, we expect the robot to ‘pay more attention’ to the environmental details that are difficult to model. For example, (in an elevation map) where red means high elevation and blue represents low elevation, (one) part of the environment (can have) more variation and is therefore harder to model,” the group said.
Their research thus far has shown that, when using existing approaches, the robot explores the environment uniformly and ignores terrain variability. The proposed attentive kernel can guide the robot to collect more valuable samples in the complex region of interest.
The proposed approach has potential beyond “active elevation mapping.” It can also be used in autonomous exploration, inspection, surveillance, and 3D reconstruction.
As far as what’s next, they want to continue facilitating education and future research on Robotic Information Gathering.
“We believe that active interaction with the environment is an indispensable capability of the future AI systems, but this field is still in its infancy, so we would like to invite more researchers to work on this topic together.”