Goal-oriented robot navigation learning using a multi-scale space representation

M. Llofriu, G. Tejera, M. Contreras, T. Pelc, Jean-Marc Fellous, A. Weitzenfeld

Research output: Contribution to journalArticle

10 Citations (Scopus)

Abstract

There has been extensive research in recent years on the multi-scale nature of hippocampal place cells and entorhinal grid cells encoding which led to many speculations on their role in spatial cognition. In this paper we focus on the multi-scale nature of place cells and how they contribute to faster learning during goal-oriented navigation when compared to a spatial cognition system composed of single scale place cells. The task consists of a circular arena with a fixed goal location, in which a robot is trained to find the shortest path to the goal after a number of learning trials. Synaptic connections are modified using a reinforcement learning paradigm adapted to the place cells multi-scale architecture. The model is evaluated in both simulation and physical robots. We find that larger scale and combined multi-scale representations favor goal-oriented navigation task learning.

Original languageEnglish (US)
JournalNeural Networks
DOIs
StateAccepted/In press - 2015

Fingerprint

Navigation
Learning
Robots
Reinforcement learning
Cognition
Place Cells
Research

Keywords

  • Hippocampus
  • Multiscale spatial representation
  • Place cells
  • Reinforcement learning
  • Spatial cognition model

ASJC Scopus subject areas

  • Artificial Intelligence
  • Cognitive Neuroscience

Cite this

Goal-oriented robot navigation learning using a multi-scale space representation. / Llofriu, M.; Tejera, G.; Contreras, M.; Pelc, T.; Fellous, Jean-Marc; Weitzenfeld, A.

In: Neural Networks, 2015.

Research output: Contribution to journalArticle

Llofriu, M. ; Tejera, G. ; Contreras, M. ; Pelc, T. ; Fellous, Jean-Marc ; Weitzenfeld, A. / Goal-oriented robot navigation learning using a multi-scale space representation. In: Neural Networks. 2015.
@article{92d475dd337949d6a6bb0bb1d81b2bec,
title = "Goal-oriented robot navigation learning using a multi-scale space representation",
abstract = "There has been extensive research in recent years on the multi-scale nature of hippocampal place cells and entorhinal grid cells encoding which led to many speculations on their role in spatial cognition. In this paper we focus on the multi-scale nature of place cells and how they contribute to faster learning during goal-oriented navigation when compared to a spatial cognition system composed of single scale place cells. The task consists of a circular arena with a fixed goal location, in which a robot is trained to find the shortest path to the goal after a number of learning trials. Synaptic connections are modified using a reinforcement learning paradigm adapted to the place cells multi-scale architecture. The model is evaluated in both simulation and physical robots. We find that larger scale and combined multi-scale representations favor goal-oriented navigation task learning.",
keywords = "Hippocampus, Multiscale spatial representation, Place cells, Reinforcement learning, Spatial cognition model",
author = "M. Llofriu and G. Tejera and M. Contreras and T. Pelc and Jean-Marc Fellous and A. Weitzenfeld",
year = "2015",
doi = "10.1016/j.neunet.2015.09.006",
language = "English (US)",
journal = "Neural Networks",
issn = "0893-6080",
publisher = "Elsevier Limited",

}

TY - JOUR

T1 - Goal-oriented robot navigation learning using a multi-scale space representation

AU - Llofriu, M.

AU - Tejera, G.

AU - Contreras, M.

AU - Pelc, T.

AU - Fellous, Jean-Marc

AU - Weitzenfeld, A.

PY - 2015

Y1 - 2015

N2 - There has been extensive research in recent years on the multi-scale nature of hippocampal place cells and entorhinal grid cells encoding which led to many speculations on their role in spatial cognition. In this paper we focus on the multi-scale nature of place cells and how they contribute to faster learning during goal-oriented navigation when compared to a spatial cognition system composed of single scale place cells. The task consists of a circular arena with a fixed goal location, in which a robot is trained to find the shortest path to the goal after a number of learning trials. Synaptic connections are modified using a reinforcement learning paradigm adapted to the place cells multi-scale architecture. The model is evaluated in both simulation and physical robots. We find that larger scale and combined multi-scale representations favor goal-oriented navigation task learning.

AB - There has been extensive research in recent years on the multi-scale nature of hippocampal place cells and entorhinal grid cells encoding which led to many speculations on their role in spatial cognition. In this paper we focus on the multi-scale nature of place cells and how they contribute to faster learning during goal-oriented navigation when compared to a spatial cognition system composed of single scale place cells. The task consists of a circular arena with a fixed goal location, in which a robot is trained to find the shortest path to the goal after a number of learning trials. Synaptic connections are modified using a reinforcement learning paradigm adapted to the place cells multi-scale architecture. The model is evaluated in both simulation and physical robots. We find that larger scale and combined multi-scale representations favor goal-oriented navigation task learning.

KW - Hippocampus

KW - Multiscale spatial representation

KW - Place cells

KW - Reinforcement learning

KW - Spatial cognition model

UR - http://www.scopus.com/inward/record.url?scp=84949294191&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84949294191&partnerID=8YFLogxK

U2 - 10.1016/j.neunet.2015.09.006

DO - 10.1016/j.neunet.2015.09.006

M3 - Article

C2 - 26548944

AN - SCOPUS:84949294191

JO - Neural Networks

JF - Neural Networks

SN - 0893-6080

ER -