Push Anything: Single- and Multi-Object Pushing from First Sight with Contact-Implicit MPC

September 2025   Bianchini

This big collaboration in my lab is currently under review at a robotics venue. My co-authors are my PhD labmate Hien Bui; three Penn masters students Yufeiyang Gao, Haoran Yang, and Stephen Felix; two Penn undergrads Eric Cui and Sid Mody; my lab alum Brian Acosta; and our advisor, Michael Posa. In this large team, my role was as the primary mentor for and one of many individual contributors on the model-building procedure, the sampling strategy, incorporation with the sampling-based contact-implicit control strategy I previously published, and experimentation. The other aspect of the project crucially is C3+, the enormously faster local contact-implicit controller.

Our official webpage

Our official webpage is located here and embedded below for ease-of-viewing.

Paper abstract

Non-prehensile manipulation of diverse objects remains a core challenge in robotics, driven by unknown physical properties and the complexity of contact-rich interactions. Recent advances in contact-implicit model predictive control (CI-MPC), with contact reasoning embedded within the trajectory optimization, have shown promise in tackling the task efficiently and robustly, yet demonstrations have been limited to narrowly-curated examples. In this work, we showcase the broader capabilities of CI-MPC through precise planar pushing tasks over a wide range of object geometries, including multi-object domains. These scenarios demand reasoning over numerous inter-object and object-environment contacts to strategically manipulate and de-clutter the environment, challenges that were intractable for prior CI-MPC methods. To achieve this, we introduce Consensus Complementarity Control Plus (C3+), an enhanced CI-MPC algorithm integrated into a complete pipeline spanning object scanning, mesh reconstruction, and hardware execution. Compared to its predecessor C3 , C3+ achieves substantially faster solve times, enabling real-time performance even in multi-object pushing tasks. On hardware, our system achieves overall 98% success rate across 33 objects, reaching pose goals within tight tolerances. The average time-to-goal is approximately 0.5, 1.6, 3.2, and 5.3 minutes for 1-, 2-, 3-, and 4-object tasks, respectively.

Citation

@article{bui2025push,
 title={Push Anything: Single- and Multi-Object Pushing From First Sight with Contact-Implicit MPC},
 author={Hien Bui* and Yufeiyang Gao* and Haoran Yang* and Eric Cui and Siddhant Mody and Brian Acosta and Thomas Stephen Felix and Bibit Bianchini and Michael Posa},
 year={2025}, 
 website={https://push-anything.github.io/}
}