Definition and composition of motor Primitives using latent force models and hidden markov models

The movement representation problem is at the core of areas such as robot imitation learning and motion synthesis. In these fields, approaches oriented to the definition of motor primitives as basic building blocks of more complex movements have been extensively used because they cope with the high...

ver descrição completa

Detalhes bibliográficos
Autor: Agudelo España, Diego Alejandro
Formato: tesis de maestría
Estado:Versión publicada
Fecha de publicación:2017
País:Colombia
Recursos:Universidad Tecnológica de Pereira
Repositorio:Repositorio Institucional UTP
Idioma:inglés
OAI Identifier:oai:repositorio.utp.edu.co:11059/8055
Acesso em linha:https://hdl.handle.net/11059/8055
Access Level:acceso abierto
Palavra-chave:Procesos de markov
Procesos de gauss
Fuerza y potencia física
id CO_db9c4e11b576ddc6ef16a8aae955cd07
oai_identifier_str oai:repositorio.utp.edu.co:11059/8055
network_acronym_str CO
network_name_str Colombia
repository_id_str
spelling Definition and composition of motor Primitives using latent force models and hidden markov modelsAgudelo España, Diego AlejandroProcesos de markovProcesos de gaussFuerza y potencia físicaThe movement representation problem is at the core of areas such as robot imitation learning and motion synthesis. In these fields, approaches oriented to the definition of motor primitives as basic building blocks of more complex movements have been extensively used because they cope with the high dimensionality and complexity by using a limited set of adjustable primitives. There is also biological evidence supporting the existence of such primitives in vertebrate and invertebrate motor systems. Traditional methods for representing motor primitives have been purely data-driven or strongly mechanistic. In the former approach new movements are generated using existing movements and these methods are usually very flexible but their extrapolation capacity is limited by the available training data. On the other hand, strongly mechanistic models have a better generalization ability by relying on a physical description of the modeled system, however, it may be hard to fully describe a real system and the resulting differential equations are usually expensive to solve numerically. Therefore, the motor primitive parameterization used in this work is based on a hybrid model which jointly incorporates the flexibility of the data-driven paradigm and the extrapolation capacity of strongly mechanistic models, namely the latent force model framework. Moreover, the sequential composition of different motor primitives is also addressed using Hidden Markov Models (HMMs) which allows to process movement realizations efficiently. The resulting joint model is an HMM with latent force models (LFMs) as emission process which is an unexplored combined probabilistic model to the best of our knowledge.Pereira : Universidad Tecnológica de PereiraFacultad de Ingenierías Eléctrica, Electrónica y Ciencias de la ComputaciónMaestría en Ingeniería EléctricaÁlvarez López, Mauricio Alexander2017-09-07T16:24:53Z2021-11-02T20:34:06Z2017-09-07T16:24:53Z2021-11-02T20:34:06Z2017masterThesisacceptedVersioninfo:eu-repo/semantics/masterThesisinfo:eu-repo/semantics/publishedVersionapplication/pdfapplication/pdfhttps://hdl.handle.net/11059/8055T519.233 A282;6310000119736 F5334engAttribution-NonCommercial-NoDerivatives 4.0 Internationalhttp://creativecommons.org/licenses/by-nc-nd/4.0/info:eu-repo/semantics/openAccessreponame:Repositorio Institucional UTPinstname:Universidad Tecnológica de Pereirainstacron:Universidad Tecnológica de Pereira2024-09-05T22:23:15Z
dc.title.none.fl_str_mv Definition and composition of motor Primitives using latent force models and hidden markov models
title Definition and composition of motor Primitives using latent force models and hidden markov models
spellingShingle Definition and composition of motor Primitives using latent force models and hidden markov models
Agudelo España, Diego Alejandro
Procesos de markov
Procesos de gauss
Fuerza y potencia física
title_short Definition and composition of motor Primitives using latent force models and hidden markov models
title_full Definition and composition of motor Primitives using latent force models and hidden markov models
title_fullStr Definition and composition of motor Primitives using latent force models and hidden markov models
title_full_unstemmed Definition and composition of motor Primitives using latent force models and hidden markov models
title_sort Definition and composition of motor Primitives using latent force models and hidden markov models
dc.creator.none.fl_str_mv Agudelo España, Diego Alejandro
author Agudelo España, Diego Alejandro
author_facet Agudelo España, Diego Alejandro
author_role author
dc.contributor.none.fl_str_mv Álvarez López, Mauricio Alexander
dc.subject.none.fl_str_mv Procesos de markov
Procesos de gauss
Fuerza y potencia física
topic Procesos de markov
Procesos de gauss
Fuerza y potencia física
description The movement representation problem is at the core of areas such as robot imitation learning and motion synthesis. In these fields, approaches oriented to the definition of motor primitives as basic building blocks of more complex movements have been extensively used because they cope with the high dimensionality and complexity by using a limited set of adjustable primitives. There is also biological evidence supporting the existence of such primitives in vertebrate and invertebrate motor systems. Traditional methods for representing motor primitives have been purely data-driven or strongly mechanistic. In the former approach new movements are generated using existing movements and these methods are usually very flexible but their extrapolation capacity is limited by the available training data. On the other hand, strongly mechanistic models have a better generalization ability by relying on a physical description of the modeled system, however, it may be hard to fully describe a real system and the resulting differential equations are usually expensive to solve numerically. Therefore, the motor primitive parameterization used in this work is based on a hybrid model which jointly incorporates the flexibility of the data-driven paradigm and the extrapolation capacity of strongly mechanistic models, namely the latent force model framework. Moreover, the sequential composition of different motor primitives is also addressed using Hidden Markov Models (HMMs) which allows to process movement realizations efficiently. The resulting joint model is an HMM with latent force models (LFMs) as emission process which is an unexplored combined probabilistic model to the best of our knowledge.
publishDate 2017
dc.date.none.fl_str_mv 2017-09-07T16:24:53Z
2017-09-07T16:24:53Z
2017
2021-11-02T20:34:06Z
2021-11-02T20:34:06Z
dc.type.none.fl_str_mv masterThesis
acceptedVersion
info:eu-repo/semantics/masterThesis
info:eu-repo/semantics/publishedVersion
format masterThesis
status_str publishedVersion
dc.identifier.none.fl_str_mv https://hdl.handle.net/11059/8055
T519.233 A282;6310000119736 F5334
url https://hdl.handle.net/11059/8055
identifier_str_mv T519.233 A282;6310000119736 F5334
dc.language.none.fl_str_mv eng
language eng
dc.rights.none.fl_str_mv Attribution-NonCommercial-NoDerivatives 4.0 International
http://creativecommons.org/licenses/by-nc-nd/4.0/
info:eu-repo/semantics/openAccess
rights_invalid_str_mv Attribution-NonCommercial-NoDerivatives 4.0 International
http://creativecommons.org/licenses/by-nc-nd/4.0/
eu_rights_str_mv openAccess
dc.format.none.fl_str_mv application/pdf
application/pdf
dc.publisher.none.fl_str_mv Pereira : Universidad Tecnológica de Pereira
Facultad de Ingenierías Eléctrica, Electrónica y Ciencias de la Computación
Maestría en Ingeniería Eléctrica
publisher.none.fl_str_mv Pereira : Universidad Tecnológica de Pereira
Facultad de Ingenierías Eléctrica, Electrónica y Ciencias de la Computación
Maestría en Ingeniería Eléctrica
dc.source.none.fl_str_mv reponame:Repositorio Institucional UTP
instname:Universidad Tecnológica de Pereira
instacron:Universidad Tecnológica de Pereira
instname_str Universidad Tecnológica de Pereira
instacron_str Universidad Tecnológica de Pereira
institution Universidad Tecnológica de Pereira
reponame_str Repositorio Institucional UTP
collection Repositorio Institucional UTP
_version_ 1825053376657752064
score 15,812429

Registos relacionados