Advances in 3D Shape Estimation of Soft Manipulators: a Deep Neural Network Perspective

EasyChair Preprint 13189

Abstract

Recent advances in soft robotics have opened up new possibilities for delicate manipulation tasks in unstructured environments. Soft manipulators exhibit highly deformable structures, posing challenges for accurate shape estimation, which is crucial for precise control and interaction with the environment. This paper presents a comprehensive review of the latest developments in 3D shape estimation techniques for soft manipulators, focusing on the application of deep neural networks (DNNs). We analyze various methodologies, including supervised, unsupervised, and semi-supervised learning approaches, highlighting their strengths and limitations in addressing the complexities of soft manipulator shape estimation. Furthermore, we discuss the integration of additional sensory modalities, such as tactile and proprioceptive feedback, to enhance the robustness and accuracy of shape estimation algorithms. Through this review, we aim to provide insights into the current state-of-the-art techniques and identify potential avenues for future research in advancing the field of soft manipulator shape estimation

Keyphrases: 3D shape estimation, Deep Neural Networks, Soft Robotics, soft manipulators, supervised learning, unsupervised learning

@booklet{EasyChair:13189,
  author    = {Kate Chastain and Wahaj Ahmed},
  title     = {Advances in 3D Shape Estimation of Soft Manipulators: a Deep Neural Network Perspective},
  howpublished = {EasyChair Preprint 13189},
  year      = {EasyChair, 2024}}