Research Analyzer
← Back ICRA 2026

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

Hien Bui, Yufeiyang Gao, Haoran Yang, Eric Cui, Siddhant Mody, Brian Acosta, Thomas Stephen Felix, Bibit Bianchini, Michael Posa

PDF

AI summary

Key figure (auto-extracted from paper)
A new contact-implicit MPC algorithm and perception pipeline enable real-time, high-precision pushing of diverse single and multi-object scenes on hardware with a 98% success rate.
Contact-implicit MPC Non-prehensile manipulation Multi-object pushing Real-time control Robot perception ADMM optimization

Problem

Prior contact-implicit MPC methods are restricted to single-object tasks with known geometries and become computationally intractable when reasoning over the exponential number of contacts in cluttered, multi-object environments.

Approach

The authors introduce Push Anything, a pipeline that combines real-time object scanning and mesh reconstruction with C3+, an accelerated CI-MPC controller that uses slack-variable reformulation and ADMM to efficiently solve multi-contact optimization problems in real time.

Key results

  • Push Anything pipeline for real-time planar pushing from first sight
  • C3+ algorithm enabling efficient multi-contact reasoning via slack-variable ADMM
  • 98% success rate across 33 objects in 928 hardware trials
  • Real-time multi-object decluttering with times-to-goal ranging from 0.5 to 5.3 minutes

Why it matters

Bridges the gap between theoretical contact optimization and real-world robotic manipulation by enabling robust, scalable pushing of arbitrary objects in cluttered scenes.

Abstract

Non-prehensile manipulation of diverse objects remains a core challenge in robotics, driven by unknown phys- ical properties and the complexity of contact-rich interactions. Recent advances in contact-implicit model predictive control (CI-MPC), with contact reasoning embedded directly in the trajectory optimization, have shown promise in tackling the task efficiently and robustly. However, demonstrations have been limited to narrowly curated examples. In this work, we show- case the broader capabilities of CI-MPC through precise planar pushing tasks over a wide range of object geometries, includ- ing multi-object domains. These scenarios demand reasoning over numerous inter-object and object-environment contacts to strategically manipulate and de-clutter the environment, which was 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. Project page: https: //dairlab.github.io/push-anything.

Index terms

Dexterous Manipulation Optimization and Optimal Control Contact Modeling

Related papers