GRASP lab paper finalist at ICRA

The University of Pennsylvania’s GRASP Laboratory said its DAIR Lab paper “Push Anything: Single- and Multi-Object Pushing from First Sight with Contact-Implicit MPC” was named a finalist for the Best Paper Award on Robot Manipulation and Locomotion at IEEE’s International Conference on Robotics and Automation, or ICRA, in 2026. The conference’s awards page lists “Push Anything” among the finalists for that category. (events.infovaya.com) The paper sits in a part of robotics that gets less public attention than grasping or humanoid demos: non-prehensile manipulation, where a robot moves objects by pushing them rather than picking them up. (dair.seas.upenn.edu) In the project materials, the Penn team said the system is designed for “real-time planar pushing of a wide variety of objects, including multi-object scenes.” ### What exactly did Penn say happened at ICRA? GRASP Lab said in a June 3 social post that the DAIR Lab paper was a finalist for the manipulation-and-locomotion best paper award at ICRA 2026. (dair.seas.upenn.edu) A DAIR Lab conference schedule page published May 27 also said PhD student Xuan Hien Bui would present the paper on Wednesday, June 3, and called it a finalist in that award category. ICRA’s official awards page lists finalists for several best paper categories, including Robot Manipulation and Locomotion. (events.infovaya.com) That page confirms “Push Anything” was in the award session lineup for finalists. ### What is “Push Anything” trying to do that is hard for robots? The Penn paper targets contact-rich pushing tasks involving unfamiliar objects and cluttered scenes. The GRASP Lab project page says non-prehensile manipulation remains difficult because robots often do not know an object’s physical properties in advance and must reason through complex contact interactions. (dair.seas.upenn.edu) The authors said their pipeline combines object scanning, mesh reconstruction, tracking and a controller based on contact-implicit model predictive control, or CI-MPC. (2026.ieee-icra.org) The paper abstract says the method is meant to handle both single objects and multi-object scenes, where the robot must reason about object-to-object and object-to-environment contact. ### What numbers made the paper stand out? The paper abstract reports a 98% overall hardware success rate across 33 objects. (grasp.upenn.edu) It also says the system reached pose goals within tight tolerances and recorded average times to goal of about 0.5 minute for one-object tasks, 1.6 minutes for two-object tasks, 3.2 minutes for three-object tasks and 5.3 minutes for four-object tasks. The same abstract says the team introduced an updated algorithm called “Consensus Complementarity Control Plus,” or C3+, which it described as a faster version of its earlier approach. (events.infovaya.com) The authors said those faster solve times enabled real-time performance even in multi-object pushing tasks. ### Who is behind the work? DAIR Lab’s ICRA page named Hien Bui as the presenter and congratulated Hien, Yufeiyang, Haoran, Eric, Sid, Brian, Thomas and Bibit. GRASP Lab profile pages and publication listings tie the project to Penn’s DAIR Lab within the broader GRASP Laboratory. (events.infovaya.com) DAIR Lab’s homepage says the group focuses on the interplay of contact dynamics, machine learning and numerical optimization, with work spanning legged robots and robotic manipulation. Michael Posa, a Penn faculty member, leads the lab. (events.infovaya.com) ### What does the finalist slot mean in conference terms? The IEEE Robotics and Automation Society says the Best Paper Award on Robot Manipulation and Locomotion recognizes the top paper in that area presented at ICRA. The society says the award has been given since 2000. (dair.seas.upenn.edu) ICRA 2026’s awards program placed the finalists in a formal award session on Tuesday, June 2, in Hall A2, according to the conference site. DAIR Lab separately said Bui’s full paper presentation was scheduled for 11:40 a.m. on Wednesday, June 3, in Vienna. (dair.seas.upenn.edu) (2026.ieee-icra.org) (ieee-ras.org)