Kaiyu Zheng, George Konidaris, And Stefanie Tellex Of Brown CS Win The IROS RoboCup Best Paper Award

Click the links that follow for more news about George Konidaris, Stefanie Tellex, and other recent accomplishments by our students and faculty.

Every year since 2007, the International Conference on Intelligent Robots and Systems (IROS) has awarded the IROS RoboCup Best Paper Award to the best paper presented at IROS related to RoboCup research. RoboCup is an international scientific initiative with the goal to advance the state of the art of intelligent robots. In 2021, a new paper ("Multi-Resolution POMDP Planning for Multi-Object Search in 3D") by Brown CS doctoral student Kaiyu Zheng and Professors George Konidaris and Stefanie Tellex, along with co-author Yoonchang Sung of MIT CSAIL, took home the honor, as well as a $1,000 prize. 

"Robots operating in human spaces," the researchers explain, "must find objects such as glasses, books, or cleaning supplies that could be on the floor, shelves, or tables. This search space is naturally 3D....Searching for objects in a large search region requires acting over long horizons under various sources of uncertainty in a partially observable environment. For this reason, previous works have used Partially Observable Markov Decision Process (POMDP) as a principled decision-theoretic framework for object search. However, to ensure the POMDP is manageable to solve, previous works reduce the search space or robot mobility to 2D."

In contrast, they introduce 3D Multi-Object Search (3DMOS), a general POMDP formulation for the multi-object search task with 3D state and action spaces, and a realistic observation space in the form of labeled voxels within the viewing frustum from a mounted camera. Demonstrating their results with a torso-actuated mobile robot in a lab environment, they show that as the problem scales, their approach outperforms exhaustive search as well as POMDP baselines without resolution hierarchy under the same computational requirement. They also show that their method is more robust to sensor uncertainty against the POMDP baselines.

More details about the work can be found here or at the following links:

For more information, click the link that follows to contact Brown CS Communication Outreach Specialist Jesse C. Polhemus.