Robotic Systems and Autonomous Platforms

Advances in Materials and Manufacturing

Shawn M. Walsh, Michael S. Strano

• Publisher: Woodhead Publishing
• Year: 2018
• eBook ISBN: 9780081020487
• Topic: Technology & Engineering
• Subtopic: Materials Science
• Index: Technology & Engineering

Robotic Systems and Autonomous Platforms: Advances in Materials and Manufacturing showcases new materials and manufacturing methodologies for the enhancement of robotic and autonomous systems. Initial chapters explore how autonomous systems can enable new uses for materials, including innovations on different length scales, from nano, to macro and large systems. The means by which autonomous systems can enable new uses for manufacturing are also addressed, highlighting innovations in 3D additive manufacturing, printing of materials, novel synthesis of multifunctional materials, and robotic cooperation. Concluding themes deliver highly novel applications from the international academic, industrial and government sectors. This book will provide readers with a complete review of the cutting-edge advances in materials and manufacturing methodologies that could enhance the capabilities of robotic and autonomous systems. - Presents comprehensive coverage of materials and manufacturing technologies, as well as sections on related technology, such as sensing, communications, autonomy/control and actuation - Explores potential applications demonstrated by a selection of case-studies - Contains contributions from leading experts in the field

Section C

Control Theory and Algorithms - Logic and Proxy Electronic Functions – Algorithmic Materials

8

Soft timer: Dynamic clock embedded in soft body

Kohei Nakajima⁎^,†; Tao Li^‡; Nozomi Akashi^{§ ⁎} JST PRESTO, Kawaguchi, Saitama, Japan
^† Chair for Frontier AI Education, Graduate School of Information Science and Technology, The University of Tokyo, Tokyo, Japan
^‡ Department of Engineering and Information Technology, Bern University of Applied Sciences, Biel, Switzerland
^§ Graduate School of Informatics, Kyoto University, Yoshida Honmachi, Kyoto, Japan

Abstract

Temporal perceptions, including timing control and interval recognition, are a fundamental constituent for our cognitive functions. Recent neuroscience reveals that the dynamic property of recurrent neural networks—not only attractor states, but also transient dynamics with a fading memory property—is key to understanding such temporal functions. Here, we tackle this issue from the perspective of embodiment, with a special emphasis on the recently emerged field of soft robotics. In particular, we argue that, by using a soft silicone arm, its passive body dynamics can implement a timing/interval control, which is conventionally performed in recurrent neural networks using its intrinsic property of short-term memory. Relationships between the strength of the actuation and the accordingly induced memory profile are further investigated in detail. Finally, we discuss a natural consequence of our results, considering the account of a brain-body-environment interaction to the temporal cognitive functions.

Keywords

Soft robotics; Soft robots; Embodiment; Morphological computation; Reservoir computing; Physical reservoir computing; Timing; Octopus; Nonlinear dynamics

Acknowledgments

This work was supported by JST PRESTO Grant Number JPMJPR15E7, Japan, and KAKENHI No. 15K16076, No. 16KT0019, and No. 26880010.

8.1 Introduction

Animals have certain body morphologies that improve their adaptivity to survive according to their ecological niche. This implies that cognitive function and behavioral control are not only associated with the brain, but that there is a reciprocal coupling between the brain, the body, and the environment, and the cognitive function and behavioral control are dynamically sustained by the interplay among these components [1–6]. Accordingly, there are many robots that are developed based on this concept (e.g., [7–10]). Recently, in the field of bio-inspired robotics, soft robots (robots with soft materials incorporated into their body) are increasingly attracting attention [11–17], because they allow us to naturally integrate an important characteristic ubiquitous in the body structures of living creatures, which is resilience and flexibility.

Compared with a rigid body, a soft body generally exhibits diverse and rich dynamics, including a variety of properties when actuated, such as nonlinearity, elasticity, and high-dimensionality. Although these properties are usually a drawback for control, we previously showed that these complex dynamics can be positively exploited as a computational resource for machine learning purposes [18–20]. This was also demonstrated using a physical platform experiment of a soft silicone arm [21–23]. Our approach was based on a technique called reservoir computing, which is a framework rooted in recurrent neural network learning [24–26]. In this framework, a high-dimensional dynamical system, which is called the reservoir, is driven by the input generating transient dynamics that enhance a separation of the input states. If the dynamics include sufficient short-term memory and ability for the nonlinear processing of the input, then by adjusting a linear and static readout from the high-dimensional dynamics of the reservoir, the required dynamical systems can be learned. Conventionally, neural networks with artificial or spiking neurons running on a program are used as a reservoir, and are implemented in many robotic platforms for behavior controls and motor pattern generations (e.g., [27–30]). Reservoirs do not always have to be driven on a PC but can be implemen...

Table of contents

• Cover image
• Title page
• Table of Contents
• Copyright
• List of contributors
• Foreword
• Preface
• Approaching robotics and autonomous systems as an integrated materials, energy, and control problem
• Section A: Actuation
• Section B: Mobility
• Section C: Control Theory and Algorithms - Logic and Proxy Electronic Functions – Algorithmic Materials
• Section D: Integration
• Section E: Energy
• Section F: Novel Robotics as Material Platforms
• Index