Behaviours for simulated humanoid robots

Detalhes bibliográficos
Autor(a) principal: Mendonça, José Lucas Lemos
Data de Publicação: 2014
Tipo de documento: Dissertação
Idioma: eng
Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
Texto Completo: http://hdl.handle.net/10773/14699
Resumo: This thesis in inserted in the FC Portugal 3D team, which competes in the humanoid simulation league 3D from RoboCup. The objectives of this thesis are to improve the behaviours already created and to develop tools to support the development and debugging of the robotic agent. With this in mind, the process of optimization was improved to make it more efficient and adapted to include the new heterogeneous models. Executing the optimization process, using the state of the art algorithm CMA-ES, the time of the getup was reduced by half. Afterwards, the agent was put running in sync mode, which allows the simulations to run as fast as the computer in use can process, and not the simulation speed of the competion with cycles of 20ms. In the agent posture, it is now used the information from the gyroscope and the euler angles are calculated to get a better estimative of the robot orientation. On the other hand, the agent architecture was updated and new behaviours were created and optimized to support the new heterogeneous models. In relation to the standard model, some behaviours execute faster because of their physical difference. In the slot behaviours, it is now possible to defined preconditions in each step, so the agent can abort the behaviour when any condition does not comply. This change reduces the time wasted executing all the behaviour in situations in which the success is improbable. In terms of tools, a Agent Monitor Window was created for each agent which can: present in runtime variables from the agent code; interact with the code trough widgets; and if the simulation is in sync mode, defined the simulation cycle time, with the possibility to pause it and execute step by step, which gives a great advantage in terms of analysing the agent execution. The second tool was a behaviour testes for behaviours defined in XML, which allows, in runtime, to change the behaviour to test, edit its content, aggregate different files in sequence and finally the tolls can execute various agents in parallel. The last tools is Log Analyser of the logs generated by the agents and the server, which allows: exporting in different formats, see in form of plots the variables parsed, filtrate the simulation information; and create a server simulation which can be used to analyse, in parallel, the plots of chosen variables and the simulation in a monitor.
id RCAP_7384bcc2704aabd2f341a87461764d47
oai_identifier_str oai:ria.ua.pt:10773/14699
network_acronym_str RCAP
network_name_str Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
repository_id_str 7160
spelling Behaviours for simulated humanoid robotsEngenharia de computadoresAndroidesRobots autónomosSistemas de controloThis thesis in inserted in the FC Portugal 3D team, which competes in the humanoid simulation league 3D from RoboCup. The objectives of this thesis are to improve the behaviours already created and to develop tools to support the development and debugging of the robotic agent. With this in mind, the process of optimization was improved to make it more efficient and adapted to include the new heterogeneous models. Executing the optimization process, using the state of the art algorithm CMA-ES, the time of the getup was reduced by half. Afterwards, the agent was put running in sync mode, which allows the simulations to run as fast as the computer in use can process, and not the simulation speed of the competion with cycles of 20ms. In the agent posture, it is now used the information from the gyroscope and the euler angles are calculated to get a better estimative of the robot orientation. On the other hand, the agent architecture was updated and new behaviours were created and optimized to support the new heterogeneous models. In relation to the standard model, some behaviours execute faster because of their physical difference. In the slot behaviours, it is now possible to defined preconditions in each step, so the agent can abort the behaviour when any condition does not comply. This change reduces the time wasted executing all the behaviour in situations in which the success is improbable. In terms of tools, a Agent Monitor Window was created for each agent which can: present in runtime variables from the agent code; interact with the code trough widgets; and if the simulation is in sync mode, defined the simulation cycle time, with the possibility to pause it and execute step by step, which gives a great advantage in terms of analysing the agent execution. The second tool was a behaviour testes for behaviours defined in XML, which allows, in runtime, to change the behaviour to test, edit its content, aggregate different files in sequence and finally the tolls can execute various agents in parallel. The last tools is Log Analyser of the logs generated by the agents and the server, which allows: exporting in different formats, see in form of plots the variables parsed, filtrate the simulation information; and create a server simulation which can be used to analyse, in parallel, the plots of chosen variables and the simulation in a monitor.Esta tese está inserida na equipa FC Portugal 3D, que compete na liga de futebol robótico simulado 3D. Os objetivos da tese são melhorar os comportamentos já existentes e desenvolver ferramentas de suporte ao desenvolvimento e depuração para o agente robótico. Nesse sentido, foi melhorado o processo de optimização de comportamentos de forma a torná-lo mais eficiente e adaptado para incluir os novos modelos heterogéneos disponibilizados. Ao executar o processo de optimização, usando o algoritmo de estado de arte CMA-ES, foi obtido reduções para metade do tempo nos comportamentos de levantar-se. Seguidamente o agente foi colocado a correr em modo síncrono, o que permite que as simulações corram à velocidade de processamento do computador em uso, e não à velocidade da simulação da competição em que cada ciclo demora 20ms. Assim é possível executar simulações e consequentemente inferir conclusões muito mais rapidamente. Passou-se a usar a informação de giroscópio e o cálculo dos ângulos de euler para obter uma melhor estimativa da rotação do robô. Por outro lado, devido ao lançamento de novos tipos de robôs, a arquitectura do agente teve de ser atualizada e novos comportamentos foram criados e optimizados para estes novos modelos. Em relação ao modelo original, alguns comportamentos são executados mais rapidamente e melhor pelos modelos novos, devido às suas alterações físicas. Por fim, nos comportamentos foi dada a possibilidade de definir pré condições em etapa do mesmo, para que possa ser abortado caso as condições não se verifiquem. Esta alteração veio reduzir o tempo desperdiçado a executar a totalidade do comportamento em situações em que não é provável o seu sucesso . Em termos de ferramentas, foi colocada uma Janela de Monitor de Agente para cada agente que, apresenta em tempo de simulação variáveis que o código do agente disponibiliza, interage com código através de widgets de seleção ou preenchimento, e se a simulação estiver a correr em modo síncrono, permite definir o tempo de ciclo da simulação, pausá-la e executar ciclo a ciclo, o que permite vantagens óbvias em termos de análise de execução dos agentes. Seguidamente, foi criada uma ferramenta de teste para comportamentos definidos em XML, que permite, em tempo de execução, alterar o ficheiro a testar, alterar o seu conteúdo, agrupar vários ficheiros em sequências e executar vários agentes em paralelo. Por fim, a última ferramenta é um Analizador de Logs gerados pelos agentes e pelo simulador que permite, entre outras funcionalidades, ver em forma de gráficos variáveis da simulação, exportar para diferentes formatos, filtrar a simulação usando informação da mesma e correr um servidor de forma a ser possível analizar em paralelo, gráficos de variáveis escolhidas e a simulação num visualizador.Universidade de Aveiro2015-09-22T15:51:25Z2014-01-01T00:00:00Z2014info:eu-repo/semantics/publishedVersioninfo:eu-repo/semantics/masterThesisapplication/pdfhttp://hdl.handle.net/10773/14699TID:201581620engMendonça, José Lucas Lemosinfo:eu-repo/semantics/openAccessreponame:Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)instname:Agência para a Sociedade do Conhecimento (UMIC) - FCT - Sociedade da Informaçãoinstacron:RCAAP2024-02-22T11:26:55Zoai:ria.ua.pt:10773/14699Portal AgregadorONGhttps://www.rcaap.pt/oai/openaireopendoar:71602024-03-20T02:50:13.665137Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos) - Agência para a Sociedade do Conhecimento (UMIC) - FCT - Sociedade da Informaçãofalse
dc.title.none.fl_str_mv Behaviours for simulated humanoid robots
title Behaviours for simulated humanoid robots
spellingShingle Behaviours for simulated humanoid robots
Mendonça, José Lucas Lemos
Engenharia de computadores
Androides
Robots autónomos
Sistemas de controlo
title_short Behaviours for simulated humanoid robots
title_full Behaviours for simulated humanoid robots
title_fullStr Behaviours for simulated humanoid robots
title_full_unstemmed Behaviours for simulated humanoid robots
title_sort Behaviours for simulated humanoid robots
author Mendonça, José Lucas Lemos
author_facet Mendonça, José Lucas Lemos
author_role author
dc.contributor.author.fl_str_mv Mendonça, José Lucas Lemos
dc.subject.por.fl_str_mv Engenharia de computadores
Androides
Robots autónomos
Sistemas de controlo
topic Engenharia de computadores
Androides
Robots autónomos
Sistemas de controlo
description This thesis in inserted in the FC Portugal 3D team, which competes in the humanoid simulation league 3D from RoboCup. The objectives of this thesis are to improve the behaviours already created and to develop tools to support the development and debugging of the robotic agent. With this in mind, the process of optimization was improved to make it more efficient and adapted to include the new heterogeneous models. Executing the optimization process, using the state of the art algorithm CMA-ES, the time of the getup was reduced by half. Afterwards, the agent was put running in sync mode, which allows the simulations to run as fast as the computer in use can process, and not the simulation speed of the competion with cycles of 20ms. In the agent posture, it is now used the information from the gyroscope and the euler angles are calculated to get a better estimative of the robot orientation. On the other hand, the agent architecture was updated and new behaviours were created and optimized to support the new heterogeneous models. In relation to the standard model, some behaviours execute faster because of their physical difference. In the slot behaviours, it is now possible to defined preconditions in each step, so the agent can abort the behaviour when any condition does not comply. This change reduces the time wasted executing all the behaviour in situations in which the success is improbable. In terms of tools, a Agent Monitor Window was created for each agent which can: present in runtime variables from the agent code; interact with the code trough widgets; and if the simulation is in sync mode, defined the simulation cycle time, with the possibility to pause it and execute step by step, which gives a great advantage in terms of analysing the agent execution. The second tool was a behaviour testes for behaviours defined in XML, which allows, in runtime, to change the behaviour to test, edit its content, aggregate different files in sequence and finally the tolls can execute various agents in parallel. The last tools is Log Analyser of the logs generated by the agents and the server, which allows: exporting in different formats, see in form of plots the variables parsed, filtrate the simulation information; and create a server simulation which can be used to analyse, in parallel, the plots of chosen variables and the simulation in a monitor.
publishDate 2014
dc.date.none.fl_str_mv 2014-01-01T00:00:00Z
2014
2015-09-22T15:51:25Z
dc.type.status.fl_str_mv info:eu-repo/semantics/publishedVersion
dc.type.driver.fl_str_mv info:eu-repo/semantics/masterThesis
format masterThesis
status_str publishedVersion
dc.identifier.uri.fl_str_mv http://hdl.handle.net/10773/14699
TID:201581620
url http://hdl.handle.net/10773/14699
identifier_str_mv TID:201581620
dc.language.iso.fl_str_mv eng
language eng
dc.rights.driver.fl_str_mv info:eu-repo/semantics/openAccess
eu_rights_str_mv openAccess
dc.format.none.fl_str_mv application/pdf
dc.publisher.none.fl_str_mv Universidade de Aveiro
publisher.none.fl_str_mv Universidade de Aveiro
dc.source.none.fl_str_mv reponame:Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
instname:Agência para a Sociedade do Conhecimento (UMIC) - FCT - Sociedade da Informação
instacron:RCAAP
instname_str Agência para a Sociedade do Conhecimento (UMIC) - FCT - Sociedade da Informação
instacron_str RCAAP
institution RCAAP
reponame_str Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
collection Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
repository.name.fl_str_mv Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos) - Agência para a Sociedade do Conhecimento (UMIC) - FCT - Sociedade da Informação
repository.mail.fl_str_mv
_version_ 1799137552342450176