Comparing Approaches for Evolving High-Level Robot Control Based on Behaviour Repertoires

Publikation: Bidrag til tidsskriftKonferenceartikelForskningpeer review

Resumé

Evolutionary robotics approaches have traditionally been focused on monolithic controllers. Recent studies on the evolution of hierarchical control have, however, yielded promising results. Hierarchical approaches typically rely on a repertoire of behaviour primitives (which themselves can be the result of an evolutionary process), and an evolved top-level arbitrator that continually executes primitives from the repertoire to solve a given task. In this paper, we compare different controller architectures for the evolution of top-level arbitrators. We propose two new methods, one based on neural networks and another based on decision trees induced by genetic programming. We compare the new approaches with existing ones, namely neural network regressors and non-hierarchical control, in a challenging simulated maze navigation task that requires a broad diversity of primitives. Based on empirical results, we draw a number of conclusions regarding the strengths and limitations of each of the studied approaches.

OriginalsprogEngelsk
Tidsskrift2018 IEEE Congress on Evolutionary Computation, CEC 2018 - Proceedings
DOI
StatusUdgivet - 28. sep. 2018
Begivenhed2018 IEEE Congress on Evolutionary Computation, CEC 2018 - Rio de Janeiro, Brasilien
Varighed: 8. jul. 201813. jul. 2018

Konference

Konference2018 IEEE Congress on Evolutionary Computation, CEC 2018
LandBrasilien
ByRio de Janeiro
Periode08/07/201813/07/2018
SponsorIEEE, IEEE Computational Intelligence Society (CIS)

Fingeraftryk

Robots
Neural networks
Controllers
Genetic programming
Decision trees
Navigation
Robotics

Citer dette

@inproceedings{616ef3c9d55f4be5ae4fbaea4702248d,
title = "Comparing Approaches for Evolving High-Level Robot Control Based on Behaviour Repertoires",
abstract = "Evolutionary robotics approaches have traditionally been focused on monolithic controllers. Recent studies on the evolution of hierarchical control have, however, yielded promising results. Hierarchical approaches typically rely on a repertoire of behaviour primitives (which themselves can be the result of an evolutionary process), and an evolved top-level arbitrator that continually executes primitives from the repertoire to solve a given task. In this paper, we compare different controller architectures for the evolution of top-level arbitrators. We propose two new methods, one based on neural networks and another based on decision trees induced by genetic programming. We compare the new approaches with existing ones, namely neural network regressors and non-hierarchical control, in a challenging simulated maze navigation task that requires a broad diversity of primitives. Based on empirical results, we draw a number of conclusions regarding the strengths and limitations of each of the studied approaches.",
keywords = "behaviour repertoires, Evolutionary robotics, genetic programming, hierarchical control, neural networks",
author = "Jorge Gomes and Christensen, {Anders Lyhne}",
year = "2018",
month = "9",
day = "28",
doi = "10.1109/CEC.2018.8477699",
language = "English",
journal = "Publishers Weekly",
issn = "0000-0019",
publisher = "PWxyz, LLC",

}

TY - GEN

T1 - Comparing Approaches for Evolving High-Level Robot Control Based on Behaviour Repertoires

AU - Gomes, Jorge

AU - Christensen, Anders Lyhne

PY - 2018/9/28

Y1 - 2018/9/28

N2 - Evolutionary robotics approaches have traditionally been focused on monolithic controllers. Recent studies on the evolution of hierarchical control have, however, yielded promising results. Hierarchical approaches typically rely on a repertoire of behaviour primitives (which themselves can be the result of an evolutionary process), and an evolved top-level arbitrator that continually executes primitives from the repertoire to solve a given task. In this paper, we compare different controller architectures for the evolution of top-level arbitrators. We propose two new methods, one based on neural networks and another based on decision trees induced by genetic programming. We compare the new approaches with existing ones, namely neural network regressors and non-hierarchical control, in a challenging simulated maze navigation task that requires a broad diversity of primitives. Based on empirical results, we draw a number of conclusions regarding the strengths and limitations of each of the studied approaches.

AB - Evolutionary robotics approaches have traditionally been focused on monolithic controllers. Recent studies on the evolution of hierarchical control have, however, yielded promising results. Hierarchical approaches typically rely on a repertoire of behaviour primitives (which themselves can be the result of an evolutionary process), and an evolved top-level arbitrator that continually executes primitives from the repertoire to solve a given task. In this paper, we compare different controller architectures for the evolution of top-level arbitrators. We propose two new methods, one based on neural networks and another based on decision trees induced by genetic programming. We compare the new approaches with existing ones, namely neural network regressors and non-hierarchical control, in a challenging simulated maze navigation task that requires a broad diversity of primitives. Based on empirical results, we draw a number of conclusions regarding the strengths and limitations of each of the studied approaches.

KW - behaviour repertoires

KW - Evolutionary robotics

KW - genetic programming

KW - hierarchical control

KW - neural networks

UR - http://www.scopus.com/inward/record.url?scp=85056265994&partnerID=8YFLogxK

U2 - 10.1109/CEC.2018.8477699

DO - 10.1109/CEC.2018.8477699

M3 - Conference article

JO - Publishers Weekly

JF - Publishers Weekly

SN - 0000-0019

ER -