Toward context awareness in the cooperation of underwater robots
- Autores: Xin Li
- Directores de la Tesis: José Fernán Martínez Ortega (dir. tes.)
- Lectura: En la Universidad Politécnica de Madrid ( España ) en 2017
- Idioma: español
- Tribunal Calificador de la Tesis: Tomás Robles Valladares (presid.), Lourdes López Santidrián (secret.), Miguel Angel López Carmona (voc.), Carlos García Rubio (voc.), Mikael Ekström (voc.)
- Programa de doctorado: Programa de Doctorado en Ingeniería de Sistemas y Servicios para la Sociedad de la Información por la Universidad Politécnica de Madrid
- Materias:
- Enlaces
- Tesis en acceso abierto en: Archivo Digital UPM
- Resumen
Nowadays, underwater robots are becoming more and more popular to carry out maritime operations, such as oil spill detection and berm construction. Knowing and understanding the physical environment where specific operations are to take place is key for accomplishing effectively and efficiently such operations and associated activities. On the other hand, when the considered activities become more complex demanding a dynamic setup, the knowledge of environment can allow enhancing performance and make adaptations. Nonetheless, it is very difficult for robots or operators to know the underwater environment, which is highly dynamic and uncertain, a priori, or even in a rough way. Facilitating context awareness, which is a capability for entities to be context-aware, in the cooperation of underwater robots still remains a challenge.
The aim of this thesis is to present a comprehensive study on delivering context awareness in the cooperation of underwater robots and overcome the most challenging problems of this topic. To achieve this objective, this thesis is carried out centering on three main research areas.
The first area aims to provide a general solution for delivering context awareness in underwater robots. An architectural proposal of a context-aware framework for underwater vehicles is presented in this thesis. The proposed context-aware framework provides a complete and well-defined context management, including context acquisition, context modeling, context reasoning, context distribution, and context dissemination. It can be an enabler of context awareness to be integrated into existing underwater robotic middleware architectures. Services provided by this framework can be exploited in different ways, such as being used by robots to understand the surrounding and for operators to conceive mission plans.
The second area concentrates on properly modeling context information that is necessarily exchanged between robots. In this area, three main contributions are made. Firstly, a fuzzy ontology development methodology (FODM) is proposed for guiding the building of fuzzy ontologies from scratch. Secondly, an ontology proposal, named the SWARMs ontology, is presented and implemented. The SWARMs ontology consists of a core ontology and four domain-specific ontologies, including mission & planning, robotic vehicle, environment recognition & sensing, and communication & networking. It is also able to be extended with fuzzy and probabilistic annotations to represent context uncertainty. Especially, the Probabilistic Web Ontology Language (PR-OWL) ontology is adopted to express context uncertainty based on the Multi-Entity Bayesian Network (MEBN) theory. In this way, the SWARMs ontology can not only present a comprehensive and principled representation of context and its associated uncertainty but also provide support for uncertainty reasoning. Finally, a proposal of applying the Stochastic Reduced Order Model (SROM) algorithm to quantify uncertainties propagated in mathematic relationships in the SWARMs ontology is presented. This proposal can guarantee a considerable degree of accuracy in approximating the statistics of uncertain ontological elements but with much fewer calculations. It is worth noting that this proposal is general enough to be applied to quantify uncertainties in any mathematics-embedded ontologies.
The last area under investigation focuses on how to effectively reason about context information and its uncertainty in the underwater robot field. None of the existing context reasoning methods can individually meet the reasoning requirements in the underwater robot field. Therefore, a hybrid context reasoning mechanism is proposed in this thesis. The proposal is to loosely couple three different context reasoning methods, namely, the ontological, rule-based, and MEBN reasoning techniques. With the combination of the different reasoning methods, it is flexible to mitigate each reasoning’s weaknesses by using others’ strengths. A set of Java Application Programming Interfaces (APIs) is implemented to realize the hybrid context reasoning proposal. The implementation provides simple interfaces to use the existing context reasoners, including the Jena OWL reasoner, Pellet reasoner, and the UnBBayes-MEBN reasoner, to provide reasoning capabilities. In addition, the implementation of this proposal is validated in terms of usefulness. A preliminary performance analysis on the hybrid context reasoning mechanism is also provided and it shows that the hybrid context reasoner can provide reasoning capabilities within an acceptable time span.
