Bio-inspired adaptive feedback error learning architecture for motor control

Silvia Tolu, Mauricio Vanegas, Niceto R. Luque, Jesus A. Garrido, Eduardo Ros

Research output: Contribution to journalJournal articleResearchpeer-review

Abstract

This study proposes an adaptive control architecture based on an accurate regression method called Locally Weighted Projection Regression (LWPR) and on a bio-inspired module, such as a cerebellar-like engine. This hybrid architecture takes full advantage of the machine learning module (LWPR kernel) to abstract an optimized representation of the sensorimotor space while the cerebellar component integrates this to generate corrective terms in the framework of a control task. Furthermore, we illustrate how the use of a simple adaptive error feedback term allows to use the proposed architecture even in the absence of an accurate analytic reference model. The presented approach achieves an accurate control with low gain corrective terms (for compliant control schemes). We evaluate the contribution of the different components of the proposed scheme comparing the obtained performance with alternative approaches. Then, we show that the presented architecture can be used for accurate manipulation of different objects when their physical properties are not directly known by the controller. We evaluate how the scheme scales for simulated plants of high Degrees of Freedom (7-DOFs).
Original languageEnglish
JournalBiological Cybernetics
Volume106
Issue number8-9
Pages (from-to)507-522
ISSN0340-1200
DOIs
Publication statusPublished - 2012
Externally publishedYes

Keywords

  • Artificial Intelligence
  • Feedback
  • Models, Neurological
  • Robotics
  • COMPUTER
  • NEUROSCIENCES
  • INTERNAL-MODELS
  • CEREBELLAR CORTEX
  • MECHANISMS
  • REGRESSION
  • CONSOLIDATION
  • MEMORY
  • ARM
  • Adaptive filter
  • Feedforward scheme
  • Cerebellum
  • Motor control
  • Machine learning
  • Internal model
  • Biomedicine
  • Statistical Physics, Dynamical Systems and Complexity
  • Computer Appl. in Life Sciences
  • Neurobiology
  • Bioinformatics
  • Neurosciences
  • SC3
  • Biotechnology
  • Computer Science (all)
  • Adaptive control architecture
  • Adaptive error
  • Adaptive feedback
  • Alternative approach
  • Bio-inspired
  • Control schemes
  • Control task
  • Feed-Forward
  • Hybrid architectures
  • Internal models
  • Learning modules
  • Locally weighted projection regressions
  • Proposed architectures
  • Reference models
  • Regression method
  • Computer science
  • Cybernetics
  • Learning systems
  • Adaptive filters
  • article
  • artificial intelligence
  • biological model
  • feedback system
  • methodology
  • robotics
  • bio-inspired adaptive feedback error
  • learning architecture
  • motor control
  • sensorimotor space
  • 04500, Mathematical biology and statistical methods
  • 20504, Nervous system - Physiology and biochemistry
  • Computational Biology
  • Neural Coordination
  • cerebellum nervous system
  • cerebellar-like engine mathematical and computer techniques
  • control task mathematical and computer techniques
  • Locally Weighted Projection Regression LWPR mathematical and computer techniques
  • Mathematical Biology
  • Nervous System
  • ADAPTIVE computing systems

Cite this