A simple mechanical system for studying adaptive oscillatory neural networks

Guillaume Jouffroy, Jerome Jouffroy

Research output: Contribution to conference without publisher/journalPaperResearchpeer-review

Abstract

Central Pattern Generators (CPG) are oscillatory systems that are responsible for generating rhythmic patterns at the origin of many biological activities such as for example locomotion or digestion. These systems are generally modelled as recurrent neural networks whose parameters are tuned so that the network oscillates in a suitable way, this tuning being a non trivial task. It also appears that the link with the physical body that these oscillatory entities control has a fundamental importance, and it seems that most bodies used for experimental validation in the literature (walking robots, lamprey model, etc.) might be too complex to study. In this paper, we use a comparatively simple mechanical system, the nonholonomic vehicle referred to as the Roller-Racer, as a means towards testing different learning strategies for an Recurrent Neural Network-based (RNN) controller/guidance system. After a brief description of the Roller-Racer, we present as a preliminary study an RNN-based feed-forward controller whose parameters are obtained through the well-known teacher forcing learning algorithm, extended to learn signals with a continuous component.

Original languageEnglish
Publication date2006
Publication statusPublished - 2006
EventIEEE Int. Conference on Systems, Man, and Cybernetics (SMC'06) - Taipei, Taiwan, Province of China
Duration: 24. Aug 2010 → …

Conference

ConferenceIEEE Int. Conference on Systems, Man, and Cybernetics (SMC'06)
CountryTaiwan, Province of China
CityTaipei
Period24/08/2010 → …

Fingerprint

Recurrent neural networks
Neural networks
Controllers
Bioactivity
Learning algorithms
Tuning
Robots
Testing

Cite this

Jouffroy, G., & Jouffroy, J. (2006). A simple mechanical system for studying adaptive oscillatory neural networks. Paper presented at IEEE Int. Conference on Systems, Man, and Cybernetics (SMC'06), Taipei, Taiwan, Province of China.
Jouffroy, Guillaume ; Jouffroy, Jerome. / A simple mechanical system for studying adaptive oscillatory neural networks. Paper presented at IEEE Int. Conference on Systems, Man, and Cybernetics (SMC'06), Taipei, Taiwan, Province of China.
@conference{0fc906103df411dda26c000ea68e967b,
title = "A simple mechanical system for studying adaptive oscillatory neural networks",
abstract = "Central Pattern Generators (CPG) are oscillatory systems that are responsible for generating rhythmic patterns at the origin of many biological activities such as for example locomotion or digestion. These systems are generally modelled as recurrent neural networks whose parameters are tuned so that the network oscillates in a suitable way, this tuning being a non trivial task. It also appears that the link with the physical body that these oscillatory entities control has a fundamental importance, and it seems that most bodies used for experimental validation in the literature (walking robots, lamprey model, etc.) might be too complex to study. In this paper, we use a comparatively simple mechanical system, the nonholonomic vehicle referred to as the Roller-Racer, as a means towards testing different learning strategies for an Recurrent Neural Network-based (RNN) controller/guidance system. After a brief description of the Roller-Racer, we present as a preliminary study an RNN-based feed-forward controller whose parameters are obtained through the well-known teacher forcing learning algorithm, extended to learn signals with a continuous component.",
author = "Guillaume Jouffroy and Jerome Jouffroy",
year = "2006",
language = "English",
note = "null ; Conference date: 24-08-2010",

}

Jouffroy, G & Jouffroy, J 2006, 'A simple mechanical system for studying adaptive oscillatory neural networks' Paper presented at, Taipei, Taiwan, Province of China, 24/08/2010, .

A simple mechanical system for studying adaptive oscillatory neural networks. / Jouffroy, Guillaume; Jouffroy, Jerome.

2006. Paper presented at IEEE Int. Conference on Systems, Man, and Cybernetics (SMC'06), Taipei, Taiwan, Province of China.

Research output: Contribution to conference without publisher/journalPaperResearchpeer-review

TY - CONF

T1 - A simple mechanical system for studying adaptive oscillatory neural networks

AU - Jouffroy, Guillaume

AU - Jouffroy, Jerome

PY - 2006

Y1 - 2006

N2 - Central Pattern Generators (CPG) are oscillatory systems that are responsible for generating rhythmic patterns at the origin of many biological activities such as for example locomotion or digestion. These systems are generally modelled as recurrent neural networks whose parameters are tuned so that the network oscillates in a suitable way, this tuning being a non trivial task. It also appears that the link with the physical body that these oscillatory entities control has a fundamental importance, and it seems that most bodies used for experimental validation in the literature (walking robots, lamprey model, etc.) might be too complex to study. In this paper, we use a comparatively simple mechanical system, the nonholonomic vehicle referred to as the Roller-Racer, as a means towards testing different learning strategies for an Recurrent Neural Network-based (RNN) controller/guidance system. After a brief description of the Roller-Racer, we present as a preliminary study an RNN-based feed-forward controller whose parameters are obtained through the well-known teacher forcing learning algorithm, extended to learn signals with a continuous component.

AB - Central Pattern Generators (CPG) are oscillatory systems that are responsible for generating rhythmic patterns at the origin of many biological activities such as for example locomotion or digestion. These systems are generally modelled as recurrent neural networks whose parameters are tuned so that the network oscillates in a suitable way, this tuning being a non trivial task. It also appears that the link with the physical body that these oscillatory entities control has a fundamental importance, and it seems that most bodies used for experimental validation in the literature (walking robots, lamprey model, etc.) might be too complex to study. In this paper, we use a comparatively simple mechanical system, the nonholonomic vehicle referred to as the Roller-Racer, as a means towards testing different learning strategies for an Recurrent Neural Network-based (RNN) controller/guidance system. After a brief description of the Roller-Racer, we present as a preliminary study an RNN-based feed-forward controller whose parameters are obtained through the well-known teacher forcing learning algorithm, extended to learn signals with a continuous component.

M3 - Paper

ER -

Jouffroy G, Jouffroy J. A simple mechanical system for studying adaptive oscillatory neural networks. 2006. Paper presented at IEEE Int. Conference on Systems, Man, and Cybernetics (SMC'06), Taipei, Taiwan, Province of China.