Landmark-based approaches for plan recognition tasks

• Ano de defesa: 2016
• Autor(a) principal: Pereira, Ramon Fraga
• Orientador(a): Meneguzzi, Felipe Rech
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Pontifícia Universidade Católica do Rio Grande do Sul
• Programa de Pós-Graduação: Programa de Pós-Graduação em Ciência da Computação
• Departamento: Faculdade de Informática
• País: Brasil
• Palavras-chave em Português: HEURÍSTICA (INFORMÁTICA); COGNIÇÃO; INFORMÁTICA
• Área do conhecimento CNPq: CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO
• Link de acesso: http://tede2.pucrs.br/tede2/handle/tede/6854

Recognition of goals and plans using incomplete evidence from action execution can be done efficiently by using automated planning techniques. In many applications it is important to recognize goals and plans not only accurately, but also quickly. In order to address this challenge, we develop recognition approaches based on planning techniques that rely on planning landmarks to filter candidate goals and plans from observations. In automated planning, landmarks are properties or actions that cannot be avoided to achieve a goal. We address the task of recognizing goals and plans without pre-defined static plan libraries, and instead we use a planning domain definition to represent the problem and the expected agent behavior. In this work, we show the applicability of planning techniques for recognition tasks in three settings: first, we use planning landmarks to develop a heuristic-based plan recognition approach; second, we refine an existing planningbased plan recognition approach; and finally, we use planning techniques to develop an approach for detecting plan abandonment. The plan abandonment detection approach we develop aims to analyze a sequence of observations and a monitored goal to determine if an observed agent is still pursuing, or has no intention to complete such monitored goal. These recognition approaches are evaluated in experiments over several planning domains. We show that our plan recognition approach yields not only accuracy comparable to other state-of-the-art techniques, but also substantially lower recognition time over such techniques. Furthermore, our plan abandonment detection approach yields high accuracy at low computational cost to detect which actions do not contribute for achieving a particular monitored goal.