Deformable Linear Object Prediction Using Locally Linear Latent Dynamics

TL;DR

This paper introduces a novel framework for predicting the behavior of deformable linear objects like ropes by transforming their complex non-linear dynamics into a linear latent space, enabling more accurate and efficient predictions and control.

Contribution

It presents a locally linear, action-conditioned dynamics model that maps non-linear object behavior into a linear latent space for improved prediction and control of deformable objects.

Findings

Accurately predicts rope state up to ten steps ahead.

Effectively finds optimal control actions between initial and goal states.

Demonstrates improved prediction efficiency over non-linear models.

Abstract

We propose a framework for deformable linear object prediction. Prediction of deformable objects (e.g., rope) is challenging due to their non-linear dynamics and infinite-dimensional configuration spaces. By mapping the dynamics from a non-linear space to a linear space, we can use the good properties of linear dynamics for easier learning and more efficient prediction. We learn a locally linear, action-conditioned dynamics model that can be used to predict future latent states. Then, we decode the predicted latent state into the predicted state. We also apply a sampling-based optimization algorithm to select the optimal control action. We empirically demonstrate that our approach can predict the rope state accurately up to ten steps into the future and that our algorithm can find the optimal action given an initial state and a goal state.