Deformable Linear Objects Manipulation with Online Model Parameters Estimation

Alessio Caporali 1, Piotr Kicki 2, Kevin Galassi 1, Riccardo Zanella 1, Krzysztof Walas 2 and Gianluca Palli 1

1. University of Bologna2. Poznan University of Technology

Abstract

Manipulating Deformable Linear Objects (DLOs) is a challenging task for a robotic system due to their unpredictable configuration, high-dimensional state space and complex nonlinear dynamics. This paper presents a framework addressing the manipulation of DLOs, specifically targeting the model-based shape control task with the simultaneous online gradient-based estimation of model parameters. In the proposed framework, a neural network is trained to mimic the DLO dynamics using the data generated with an analytical DLO model for a broad spectrum of its parameters. The neural network-based DLO model is conditioned on these parameters and employed in an online phase to perform the shape control task by estimating the optimal manipulative action through a gradient-based procedure. In parallel, gradient-based optimization is used to adapt the DLO model parameters to make the neural network-based model better capture the dynamics of the real-world DLO being manipulated and match the observed deformations. To assess its effectiveness, the framework is tested across a variety of DLOs, surfaces, and target shapes in a series of experiments. The results of these experiments demonstrate the validity and efficiency of the proposed methodology compared to existing methods.

Resources

Paper IEEE Xplore

Code

Experiments

Shape Control Task with Online Parameters Estimation

Outcomes of the shape control task involving online adaptation of model parameters, conducted across various rope types and surfaces. Average results across 5 repetitions per task (standard deviations confidence region intervals).

Impact of Online Model Parameters Estimation

Comparing prediction errors using mid-range, online, and best model parameters across ropes and surfaces.

Within just a few iterations, the proposed method attains parameters that yield a mean error comparable to the best scenario, and in most of the cases significantly better than for the mid-range parameters.

Fixed vs Conditioning Parameters

Prediction error for fix (no params) vs conditioning parameters across different models. For the latter, the symbol (*) denotes that the same shape is used for parameter estimation and forward prediction error, whereas with (**) the 4-fold cross-validation approach is denoted. With fixed mid-range and varied the two employed datasets are indicated.

Parameters vs Weights Update

The mean prediction error (log scale) of the input parameters (ours) vs neural network weights (RBF [2]) update, evaluated on the same shape (*) or on different shapes (**).