Antonio Campos

Environmental user-preference learning for smart homes: An autonomous approach

The ubiquitous computing concept was introduced by Weiser in 1991. Despite the many advances since then, the use of ubiquitous spaces is not yet widespread in home environments, due to the difficulty in using context information and artificial intelligence algorithms. This work combines context-management architecture and machine-learning techniques to infer environmental user preferences automatically. Context information is represented with ontologies and the system is organized through a service-oriented architecture. The system learns all the parameters that the user configures in the home. These parameters are taken as the users preferences and will be learned automatically by means of an algorithm based on the KNN machine learning algorithm. These preferences will enable the development of context-aware applications for the home environment that will proactively adjust, depending on the learned values and the context. These kinds of applications will further the concept of ambient-assisted living and, hence, aid in improving quality of life in the home.

Multilevel and hybrid architecture for device abstraction and context information management in smart home environments

Hardware device management, and context information acquisition and abstraction are key factors to develop the ambient intelligent paradigm in smart homes. This work presents an architecture that addresses these two problems and provides a usable framework to develop applications easily. In contrast to other proposals, this work addresses performance issues specifically. Results show that the execution performance of the developed prototype is suitable for deployment in a real environment. In addition, the modular design of the system allows the user to develop applications using different techniques and different levels of abstraction.

Offloading Approach for Flexible Provisioning and Execution of Remote Data-gathering Applications

Nowadays, applications must be provided for mobile devices with very different resources. In order to provide the user with a satisfying experience, some of the computation can be offloaded to a remote server. This paper presents an approach for this offloading process for remote data-gathering applications. The set of requirements to be fulfilled by the approach is identified and presented. This approach takes into account resources such as CPU, memory, storage and connection bandwidth, both at provisioning and execution time, as well as allowing different granularities for the tasks to be offloaded. In addition, it takes into account other tasks being executed in the mobile device and in the server. The approach is based on client-server architecture. A first-approach solution is presented, describing the decisions to be made both at application-provisioning time and during runtime. A prototype of a.NET application for sputum analysis has been implemented in order to show the applicability of the proposed approach. Execution time, memory usage, storage and transmission time have been tested with a variety of devices. The results show how this approach can be applied in practice, even with the limitations that operating systems and real devices impose.

From UAProf towards a Universal Device Description Repository

Techniques to create software and content that adapt to different apparatus require gathering information about device features. Traditionally, Device Description Repositories (DDRs) have provided limited descriptions, in terms of description granularity and of the amount of devices included. A Universal DDR (UDDR) would allow any software developer or content creator to have complete, up-to-date and trustworthy device descriptions for any application domain. Collaboration of all stakeholders in the adaptation business would be necessary to populate the UDDR, but without compromising the quality of the information. Device manufacturers usually publish first-hand device descriptions using UAProf. Unfortunately, UAProf documents are known to contain mistakes or inaccurate/incomplete information. This work suggests a multi-step process to manipulate UAProfs in order to correct their most common mistakes, to extend their expressiveness and to allow amendments from different contributors. More specifically, amendments are annotated with provenance information, enabling device description consumers to decide whether to trust them.

A real-time thermal model for the intelligent control of cooling systems for hot strip mill work rolls

This article describes a mathematical model which calculates, in real time, the temperature and the thermal expansion of the work rolls of hot strip mills. These results are used to feed a real-time expert system, which controls the cooling system of the work rolls. These two magnitudes can not be measured online which, over the last few decades, has motivated several authors to develop different models to obtain the thermal evolution of the work rolls of hot strip mills. However, while their motivation was to achieve the greatest possible accuracy working off-line, the motivation of this work is very different: to achieve a mathematical model that, while maintaining acceptable accuracy, is capable of providing results in real-time and with the minimum consumption of computational resources.