Homeostatic robot control using simple neuromodulatory techniques

James C. Finnis*

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference Proceeding (Non-Journal item)

Abstract

The UESMANN (Uniform Excitatory Switching Multifunction Artificial Neural Network) architecture has been shown to produce interesting transitions between multiple behaviours using an extremely simple neuromodulatory regime. Previous work has concentrated on discrete classification tasks. In this work, three different simple neuromodulatory architectures including UESMANN are used to control a robot in a homeostatic task. The experiments show that UESMANN produces interesting and useful transitional behaviour in an embodied system, learning the two tasks in the same number of parameters (i.e. network weights) as networks which learned each individual task.

Original languageEnglish
Title of host publicationTowards Autonomous Robotic Systems - 18th Annual Conference, TAROS 2017, Proceedings
EditorsYang Gao, Saber Fallah, Yaochu Jin, Constantina Lekakou
PublisherSpringer Nature
Pages325-339
Number of pages15
ISBN (Electronic)978-3-319-64107-2
ISBN (Print)9783319641065, 3319641069
DOIs
Publication statusPublished - 19 Jul 2017
Event18th Annual Conference on Towards Autonomous Robotic Systems, TAROS 2017 - Guildford, United Kingdom of Great Britain and Northern Ireland
Duration: 19 Jul 201721 Jul 2017

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume10454 LNAI
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference18th Annual Conference on Towards Autonomous Robotic Systems, TAROS 2017
Country/TerritoryUnited Kingdom of Great Britain and Northern Ireland
CityGuildford
Period19 Jul 201721 Jul 2017

Keywords

  • Backpropagation
  • Long-term autonomy Homeostasis
  • Neural network
  • Neuromodulation
  • Robotics

Fingerprint

Dive into the research topics of 'Homeostatic robot control using simple neuromodulatory techniques'. Together they form a unique fingerprint.

Cite this