Comunicação de expressões espaciais em sistemas multi-agentes utilizando a ego-esfera qualitativa
Carregando...
Arquivos
Citações na Scopus
Tipo de produção
Dissertação
Data
2015
Autores
Rodrigues, F. M. E.
Orientador
Santos, Paulo Eduardo
Periódico
Título da Revista
ISSN da Revista
Título de Volume
Citação
RODRIGUES, F. M. E. Comunicação de expressões espaciais em sistemas multi-agentes utilizando a ego-esfera qualitativa. 2015. 121 f. Dissertação (Mestrado em Engenharia Elétrica) - Centro Universitário FEI, São Bernardo do Campo, 2015 Disponível em: . Acesso em: 27 set. 2017.
Texto completo (DOI)
Palavras-chave
Raciocínio qualitativo,Comunicação humano-robô,Ego-esfera qualitativa
Resumo
A necessidade de representações espaciais e raciocínio espacial é onipresente em Inteligência Artificial (IA). Investigação em raciocínio espacial qualitativo (REQ) (sub-área da IA) tem como desafio criar formalismos que permitam uma máquina representar e raciocinar sobre entidades espaciais. Atualmente é desejável criar um sistema robótico capaz de substituir voluntários humanos em situações de busca e resgate, principalmente quando há algum risco à vida humana. Nesses casos, é importante que o sistema artificial seja capaz de se comunicar
com os agentes humanos da forma mais natural possível. Desse modo este trabalho desenvolve uma forma de realizar comunicação através de expressões espaciais em ambientes homogêneos (somente com robôs) e ambientes heterogêneos (com humanos e robôs). A fim de atingir este objetivo, o presente trabalho utiliza os conceitos da Ego-Esfera Qualitativa como base para possibilitar a comunicação por meio de expressões espaciais. A Ego-Esfera Qualitativa possibilita discretizar qualitativamente o espaço ao redor de um agente. Foram estabelecidas as definições para as regiões ao redor da Ego-Esfera Qualitativa de modo a dividir o equador da esfera em oito regiões angulares (ou setores), longitudinalmente em três distâncias e verticalmente em
quatro alturas. Com essas definições foi possível a um agente robótico classificar a localização de objetos em relação a outro agente ou em relação a si próprio. Assim foram estudados quatro casos representando possíveis situações de comunicação entre agentes em um ambiente. Neste trabalho inicialmente vários objetos foram dispostos em um ambiente e a localização dos objetos foi obtida manualmente e inserida no sistema de modo a criar o padrão ouro, livre de ruído. Em seguida a localização dos objetos foi obtida utilizando o sistema de visão que implementa a Ego-Esfera Qualitativa. Por fim foram realizados testes com humanos, a fim de determinar
como as pessoas localizam objetos dispostos no ambiente. Os valores do padrão ouro, os valores obtidos pelo sistema de visão e os valores obtidos pelas respostas dos humanos foram comparados. (com base na Ego-Esfera Qualitativa) é compatível com aquela executada pelo agente humano. Os experimentos demonstraram a existência de ambiguidade na conceitualização humana do espaço, fazendo-se necessário adaptar as definições da Ego-Esfera Qualitativa. Observou-se também que as definições de altitude e longitude da Ego-Esfera Qualitativa são vagas, dado a necessidade de informar ao humano uma referência ou critério para classificar um objeto em
uma das distâncias da Ego-Esfera Qualitativa. Em relação aos setores da esfera, observou-se que as respostas humanas estão próximas quanto as definições da Ego-Esfera Qualitativa.
The need for spatial representations and spatial reasoning is omnipresent in Artificial Intelligence. The task of the qualitative spatial reasoning (QSR) subfield of AI is to provide formalisms that allow a machine to represent and reason about spatial entities. Currently it is desirable to create a computational system that allows the replacement of human professionals in search and rescue situations, especially when there is some risk to the professional. To this end, it is important that humans and machines are able to communicate using common concepts. This work develops a way to perform communication through spatial expressions in homogeneous environment (only with robots) and heterogeneous environments (with humans and robots). In order to perform this communication, the present work uses the concept of Qualitative Ego-Sphere as the basis for enabling the communication of spatial expressions. The Qualitative Ego-Sphere is a discretization of the space around an agent. Definitions have been established for the regions around the Qualitative Ego-Sphere so as to divide the sphere s equator into eight regions (or sectors), along three distances from the observer. Vertically the sphere is divided into four heights. With these definitions of space it was possible for a robotic agent to classify the location of objects in relation to another agent or itself. In the present dissertation four cases were investigated, representing potential situations of communication between agents in an environment. The situations involved the placement of various objects in an environment; the object s locations were initially obtained manually to create the Gold Standard in order to evaluate the system developed. Then the location of each object was achieved by using a vision system implemented with a version of the Qualitative Ego-Sphere. Finally, human trials were performed by asking the position of objects arranged on the environment. The values of the Gold Standard, the values obtained by the vision system and the values obtained by the responses of human were compared. It was thus possible to evaluate the extent of which the qualitative classification of a robotic agent (based on Qualitative Ego-Sphere) agrees with that of a human agent. The experiments demonstrated the existence of ambiguity in the human conceptualization of space making it necessary to adapt the Qualitative Ego-Sphere settings. It was also observed vagueness in the altitude and longitude Qualitative Ego-Sphere definitions, given the necessity to inform to the human some reference or criteria to classify an object in one of the distances, accordingly to the definitions of Qualitative Ego-Sphere. Regarding the sectors, we observed that the human answers are close to the definitions from Qualitative Ego-Sphere.
The need for spatial representations and spatial reasoning is omnipresent in Artificial Intelligence. The task of the qualitative spatial reasoning (QSR) subfield of AI is to provide formalisms that allow a machine to represent and reason about spatial entities. Currently it is desirable to create a computational system that allows the replacement of human professionals in search and rescue situations, especially when there is some risk to the professional. To this end, it is important that humans and machines are able to communicate using common concepts. This work develops a way to perform communication through spatial expressions in homogeneous environment (only with robots) and heterogeneous environments (with humans and robots). In order to perform this communication, the present work uses the concept of Qualitative Ego-Sphere as the basis for enabling the communication of spatial expressions. The Qualitative Ego-Sphere is a discretization of the space around an agent. Definitions have been established for the regions around the Qualitative Ego-Sphere so as to divide the sphere s equator into eight regions (or sectors), along three distances from the observer. Vertically the sphere is divided into four heights. With these definitions of space it was possible for a robotic agent to classify the location of objects in relation to another agent or itself. In the present dissertation four cases were investigated, representing potential situations of communication between agents in an environment. The situations involved the placement of various objects in an environment; the object s locations were initially obtained manually to create the Gold Standard in order to evaluate the system developed. Then the location of each object was achieved by using a vision system implemented with a version of the Qualitative Ego-Sphere. Finally, human trials were performed by asking the position of objects arranged on the environment. The values of the Gold Standard, the values obtained by the vision system and the values obtained by the responses of human were compared. It was thus possible to evaluate the extent of which the qualitative classification of a robotic agent (based on Qualitative Ego-Sphere) agrees with that of a human agent. The experiments demonstrated the existence of ambiguity in the human conceptualization of space making it necessary to adapt the Qualitative Ego-Sphere settings. It was also observed vagueness in the altitude and longitude Qualitative Ego-Sphere definitions, given the necessity to inform to the human some reference or criteria to classify an object in one of the distances, accordingly to the definitions of Qualitative Ego-Sphere. Regarding the sectors, we observed that the human answers are close to the definitions from Qualitative Ego-Sphere.