Closed-Loop Trajectory Optimization of Deformable Linear Objects for Dynamic Motions
Marc Kilian Klankers, Jochen J. Steil
AI summary
Problem
Tracking dynamic endpoint trajectories of deformable linear objects remains challenging due to complex non-linear behavior and the lack of accurate, computationally efficient dynamic models. Existing methods often rely on quasi-static assumptions or open-loop control, limiting predictive performance.
Approach
The authors model the DLO as a floating-base kinematic chain and train a neural network to approximate its hybrid dynamics, simultaneously predicting joint accelerations and base wrenches. This learned model is embedded in a closed-loop optimal control loop solved via linear MPC and DDP to track dynamic 2D trajectories.
Key results
- Closed-loop trajectory optimization framework for dynamic DLO manipulation
- Data-driven hybrid dynamics model predicting joint accelerations and base wrenches
- Successful simulation and hardware experiments tracking dynamic 2D endpoint trajectories
- Implementation and comparison of gradient descent, linear MPC, and DDP solvers for real-time control
Why it matters
Enables precise, predictive control of deformable objects in dynamic tasks, advancing robotics applications in industrial, medical, and domestic environments.
Abstract
Tracking dynamic endpoint trajectories of de- formable linear objects (DLOs) with a robotic manipulator remains challenging due to their complex non-linear behavior. While closed-loop Model Predictive Control (MPC) can account for these non-linearities, it requires an accurate dynamic model and precise state estimation. This paper introduces a closed-loop approach for controlling a DLO’s endpoint to track dynamic 2D shapes. We model the DLO as a floating-base kinematic chain and present a new perspective on learning its dynamics using a data-driven approximation of its hybrid dynamics. Based on this model, we formulate an Optimal Control Problem (OCP), which we solve within the control loop using both linear MPC and DDP. We validate our approach with simulation and hardware experiments, demonstrating its ability to track dynamic endpoint motions.