Fully Distributed Real-Time MPC for Cooperative Mobile Manipulation Via Box-iLQR and ADMM with an Object-Centric Planar Projection
Jeong tae Lee, Jin Ho Park, Seunghoon Yang, Keun Ha Choi, Kyung-Soo Kim
AI summary
Problem
Cooperative mobile manipulation typically relies on centralized optimization or high-dimensional whole-body models, creating computational bottlenecks that prevent real-time deployment and limit scalability. This paper addresses the need for a lightweight, distributed control strategy that synchronizes multiple robots on a shared object state without centralized computation.
Approach
Each robot solves a local optimal control problem using Box-iLQR while ADMM enforces consensus on a shared object state. The method reduces computational load by projecting the whole-body dynamics onto an object-centric planar orthographic projection.
Key results
- Box-iLQR solver operates at ~6.32 ms per iteration, roughly 4 times faster than a full 6-DoF model
- ADMM-based distributed consensus achieves accurate trajectory tracking without a central server
- Dimension reduction cuts computational cost nearly in half while maintaining comparable tracking accuracy
- Framework scales naturally with robot count and preserves stability under simplified dynamics
Why it matters
Enables real-time, scalable cooperative manipulation for industrial and logistics applications where compute and communication resources are limited.
Abstract
We propose a fully distributed real-time model predictive control framework for transporting a single rigid object with multiple mobile manipulators. Each robot rapidly solves a local optimal control problem via Box-iLQR, while ADMM enforces consensus on the shared object state without centralized computation. The core idea is an object-centric planar orthographic projection that reduces the whole-body state and input dimensions, substantially lowering the com- putational load of linearization and the Riccati backward pass. Simulations demonstrate accurate trajectory tracking and consistent convergence. Specifically, the proposed dimension- reduced Box-iLQR solver operates at an average of 6.32 ms per iteration—approximately 4 times faster than a full 6-DoF model and cutting the computational cost of SQP-based methods nearly in half. Despite this significant reduction, our controller achieves comparable tracking accuracy, offering a practical alternative for real-time cooperative manipulation under limited compute and communication resources. The framework scales naturally with the number of robots and provides a concise and effective design for cooperative mobile manipulation grounded in real-time distributed optimization.