Towards a Hybrid Approach to Context Reasoning for Underwater Robots

Abstract

Abstract:Ontologies have been widely used to facilitate semantic interoperability and serve as a common information model in many applications or domains. The Smart and Networking Underwater Robots in Cooperation Meshes (SWARMs) project, aiming to facilitate coordination and cooperation between heterogeneous underwater vehicles, also adopts ontologies to formalize information that is necessarily exchanged between vehicles. However, how to derive more useful contexts based on ontologies still remains a challenge. In particular, the extreme nature of the underwater environment introduces uncertainties in context data, thus imposing more difficulties in context reasoning. None of the existing context reasoning methods could individually deal with all intricacies in the underwater robot field. To this end, this paper presents the first proposal applying a hybrid context reasoning mechanism that includes ontological, rule-based, and Multi-Entity Bayesian Network (MEBN) reasoning methods to reason about contexts and their uncertainties in the underwater robot field. The theoretical foundation of applying this reasoning mechanism in underwater robots is given by a case study on the oil spill monitoring. The simulated reasoning results are useful for further decision-making by operators or robots and they show that the consolidation of different reasoning methods is a promising approach for context reasoning in underwater robots.

Construção do Estado da Arte em Estratégias Híbridas para Raciocínio de Contexto: uma Abordagem explorando Revisão Sistemática da Literatura

“…No artigo [Li et al 2017]é proposta uma estratégia híbrida para o raciocínio de contexto para um framework de ciência de contexto. A estratégia propostaé constituída de três técnicas de raciocínio: (i) ontológico, sendo utilizados axiomas internosà ontologia com o intuito de inferir conhecimento relacionado as classes presentes na mesma; (ii) baseado em regras, que consiste de regras que processam os dados contextuais presentes na ontologia, sendo utilizada a linguagem SWRL (Semantic Web Rule Language) para a concepção das regras; (iii) Multi-Entity Bayesian Network, o qual consiste de um sistema lógico que integra a lógica da primeira ordem com a teoria da probabilidade bayesiana.…”

“…Os autores destacam que, com a incorporação das três técnicas,é proporcionado um melhor desempenho para o raciocínio de contexto, aproveitando-se dos pontos positivos de cada técnica, com o intuito de compensar as fraquezas das mesmas. Os autores destacam que a estratégia híbrida propostaé eficaz para fornecer diferentes capacidades de raciocínio e satisfazer as diferentes necessidades das aplicações [Li et al 2017].…”

“…Dentre os sete trabalhos selecionados, três deles [Lopes et al 2012, Strobbe et al 2012, Yuan and Herbert 2014 se caracterizam por poderem ser aplicados em diferentes domínios. Enquanto os outros trabalhos [Smaaberg et al 2014, Baz et al 2016, Razzaq et al 2017, Li et al 2017] são aplicados respectivamente em: sistemas de recomendação; sistemas de classificação; paradigma de saúde e bem estar; robôs subaquáticos. RQ3: Que metodologiaé empregada na avaliação?…”

MEBN-RM: A Mapping between Multi-Entity Bayesian Network and Relational Model

“…MEBN is sufficiently expressive for general-purpose knowledge representation and reasoning in an uncertain and 2 complex world. Because MEBN is flexible enough to represent a variety of complex and uncertain situations, it has been applied to systems for Predictive Situation Awareness (PSAW), the problem of understanding and predicting aspects of a temporally evolving situation [Laskey et al, 2000] [Wright et al, 2002] [Suzic, 2005] [Costa et al, 2009] [Carvalho et al, 2010] [Park et al, 2014] [Golestan, 2016] [Li et al, 2017] [Park et al, 2017]. In a recent review of knowledge representation formalisms, Golestan et al [2016] recommended MEBN as having the most comprehensive coverage of features needed to represent complex problems among several artificial intelligence (AI) models including statistical relational models (e.g., Probabilistic Relational Models [Getoor & Taskar, 2007] and Markov Logic Networks [Richardson & Domingos, 2006]).…”

Proposal of an Automated Mission Manager for Cooperative Autonomous Underwater Vehicles

“…It is also in charge of storing the information obtained from different domains, including mission and planning, and environment recognition and sensing among others. Besides the information model, the ontology is also used to store the last updated values, as they are also used automatically to infer events by means of the use of semantic reasoners [43][44][45].…”