Research Analyzer
← Back ICRA 2026

Approximating Global Contact-Implicit MPC Via Sampling and Local Complementarity

Sharanya Venkatesh, Bibit Bianchini, Alp Aydinoglu, William Yang, Michael Posa

PDF

AI summary

Key figure (auto-extracted from paper)
A real-time controller for dexterous manipulation that combines global end-effector sampling with local contact-implicit MPC to escape local minima.
Contact-Implicit MPC Dexterous Manipulation Sampling-based Control Linear Complementarity Systems Real-time Robotics

Problem

Existing model-based controllers cannot globally optimize over the exponential number of possible contact sequences in real time, and local approximations often get stuck in local minima.

Approach

The method splits control into a contact-free repositioning stage and a contact-rich MPC stage, using parallel sampling of end-effector locations to inform global movement.

Key results

  • Precise non-prehensile manipulation of non-convex objects on hardware with a Franka arm
  • Accomplished tasks that local control alone essentially always fails to do
  • Outperformed MuJoCo MPC (MJPC) in simulation success and safety
  • Online execution requiring no offline computation or training

Why it matters

Enables robots to autonomously discover and execute complex, multi-contact behaviors for general-purpose manipulation without manual heuristics.

Abstract

To achieve general-purpose dexterous manipulation, robots must rapidly devise and execute contact-rich behaviors. Existing model-based controllers cannot globally optimize in real time over the exponential number of possible contact sequences. Instead, progress in contact-implicit control leverages simpler models that, while still hybrid, make local approximations. Locality limits the controller to exploit only nearby interactions, requiring intervention to richly explore contacts more broadly. Our approach leverages the strengths of local complementarity- based control combined with low-dimensional, but global, sam- pling of possible end effector locations. Our key insight is to consider a contact-free stage preceding a contact-rich stage at every control loop. Our algorithm, in parallel, samples end effector locations to which the contact-free stage can move the robot, then considers the cost predicted by contact-rich MPC local to each sampled location. The result is a globally-informed, contact-implicit controller capable of real-time dexterous manip- ulation. We demonstrate our controller on precise, non-prehensile manipulation of non-convex objects with a Franka arm. Project webpage: https://approximating-global-ci-mpc.github.io

Index terms

Dexterous Manipulation Optimization and Optimal Control Integrated Planning and Control

Related papers