Learning-Guided Force-Feedback Model Predictive Control with Obstacle Avoidance for Robotic Deburring
Krzysztof Wojciechowski, Ege Gursoy, Arthur Haffemayer, Sebastien Kleff, Vincent Bonnet, Florent Lamiraux, Nicolas Mansard
AI summary
Problem
Classical MPC formulations often neglect real-time force feedback and struggle to maintain stable contact in collision-rich industrial tasks like deburring. This gap leaves robots unable to precisely regulate contact forces and navigate tight clearances without frequent re-fixturing.
Approach
The framework augments force-feedback MPC with a diffusion model that encodes task-specific motion strategies as a reusable prior, providing robust initialization while the controller enforces explicit force tracking and collision avoidance.
Key results
- Formalized deburring optimal control problem with 1D and 3D contact models
- Integrated diffusion-based motion priors for robust MPC warm-starting
- Achieved 100% success rate in multi-hole deburring under reach and force constraints
- Demonstrated that explicit tangential force regulation prevents slip-out during collision avoidance
Why it matters
Provides a scalable solution for contact-rich industrial automation, reducing setup time and enabling reliable execution in complex, obstacle-dense workspaces.
Abstract
Model Predictive Control (MPC) is widely used for torque-controlled robots, but classical formulations often neglect real-time force feedback and struggle with contact- rich industrial tasks under collision constraints. Deburring in particular requires precise tool insertion, stable force regula- tion, and collision-free circular motions in challenging con- figurations, which exceeds the capability of standard MPC pipelines. We propose a framework that integrates force- feedback MPC with diffusion-based motion priors to address these challenges. The diffusion model serves as a memory of motion strategies, providing robust initialization and adapta- tion across multiple task instances, while MPC ensures safe execution with explicit force tracking, torque feasibility, and collision avoidance. We validate our approach on a torque- controlled manipulator performing industrial deburring tasks. Experiments demonstrate reliable tool insertion, accurate nor- mal force tracking, and circular deburring motions even in hard-to-reach configurations and under obstacle constraints. To our knowledge, this is the first integration of diffusion motion priors with force-feedback MPC for collision-aware, contact-rich industrial tasks. The source code can be found at github.com/agimus-project/agimus-demos.