Integrating automated planning with a multi-agent system development framework

• Ano de defesa: 2014
• Autor(a) principal: Cardoso, Rafael Cauê
• Orientador(a): Bordini, Rafael Heitor
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Pontifícia Universidade Católica do Rio Grande do Sul; Porto Alegre
• Palavras-chave em Português: INFORMÁTICA; SISTEMAS MULTIAGENTES; AGENTE INTELIGENTE (SOFTWARE)
• Link de acesso: http://hdl.handle.net/10923/7044

Automated planning is an important capability to have in intelligent agents. Extensive research has been done in single-agent planning, but so far planning has not been fully explored in multiagent systems because of the computational costs of multi-agent planning algorithms. With the increasing availability of distributed systems and more recently multi-core processors, several new multi-agent planning algorithms have been developed, such as the MAP-POP algorithm, which we integrate into the JaCaMo multi-agent system framework. Our work provides offline multi-agent planning capabilities as part of a multi-agent system development framework. This framework supports complex multi-agent problems based on agent-oriented programming. In summary, the main contribution of this work is to provide the developers with an initial multiagent system implementation for a target scenario, based on the solutions found by the MAPPOP multi-agent planner, and on which the developer can work further towards a fully-fledged multi-agent system.