Combining Self-Organizing and Graph Neural Networks for Modeling Deformable Objects in Robotic Manipulation.

Angel J Valencia,Pierre Payeur

doi:10.3389/frobt.2020.600584

Abstract

Modeling deformable objects is an important preliminary step for performing robotic manipulation tasks with more autonomy and dexterity. Currently, generalization capabilities in unstructured environments using analytical approaches are limited, mainly due to the lack of adaptation to changes in the object shape and properties. Therefore, this paper proposes the design and implementation of a data-driven approach, which combines machine learning techniques on graphs to estimate and predict the state and transition dynamics of deformable objects with initially undefined shape and material characteristics. The learned object model is trained using RGB-D sensor data and evaluated in terms of its ability to estimate the current state of the object shape, in addition to predicting future states with the goal to plan and support the manipulation actions of a robotic hand.

Highlights

In the context of robotic manipulation, object models are used to provide feedback signals that a robot can control when performing a specific task
Attempts to estimate the object shape in robotic manipulation mainly adopted an analytical approach, which is commonly adjusted in simulation (Nadon et al, 2018)
The contributions of this work can be summarized as follows: First, we develop a method for shape estimation using Self-Organizing Neural Networks (SONNs)

Summary

Introduction

In the context of robotic manipulation, object models are used to provide feedback signals that a robot can control when performing a specific task. Attempts to estimate the object shape in robotic manipulation mainly adopted an analytical approach, which is commonly adjusted in simulation (Nadon et al, 2018) This comes with some drawbacks in real robotic environments, as simulators are currently not sophisticated enough to provide realistic models of non-rigid objects (Billard and Kragic, 2019), and the support for sensor measurements and hardware in simulators is very limited. It is rarely possible to determine these conditions in advance for every new object encountered in the environment This lack of a general-purpose methodology to estimate the object shape makes it difficult to develop more autonomous and dexterous robotic manipulation systems capable to handle deformable objects (Sanchez et al, 2018)

Objectives

Methods

Discussion

Conclusion