Published papers

This paper presents an approach to supporting the development process for large-scale smart city applications that leverage edge computing resources. A smart city testbed called CitiFlow is developed, which uses Distributed Node-RED as the underlying middleware to facilitate the decomposition and communication among sub components of large scale smart city applications. For the evaluation, a lab-based setup and a real world deployment were executed and are presented.

A growing trend in smart cities is the use of machine learning techniques to gather city data, formulate learning tasks and models, and use these to develop solutions to city problems. However, although these processes are sufficient for theoretical experiments, they often fail when they meet the reality of city data and processes, which by their very nature are highly distributed, heterogeneous, and exhibit high degrees of spatial and temporal variance. In order to address those problems, we have designed and implemented an integrated development environment called CityFlow that supports developing machine learning applications. With CityFlow, we can develop, deploy, and maintain machine learning applications easily by using an intuitive data flow model. To verify our approach, we conducted two case studies: deploying a road damage detection application to help monitor transport infrastructure and an automatic labeling application in support of a participatory sensing application. These applications show both the generic applicability of our approach, and its ease of use; both critical if we wish to deploy sophisticated ML based applications to smart cities.

We present a new model for encouraging people to get involved with monitoring and taking part in the life of cities. Cities could be smarter if IoT and people could serve as engaged and pro-active data resources (i.e., crowd sensing). This study tackles two challenges: methods by which the privacy of people who act as sensors/actuators can be guaranteed and methods to create a unified programming model for crowd sensors alongside other IoT functions. To achieve these goals, we introduce a new concept called Lokemon (Location Monster). Each sensing space is characterized as a personified target. Lokemon asks users to imagine themselves to be monsters associated with target spots when achieving sensing tasks. Lokemon is also expressed as a PubSub node so that the data from Lokemon can be easily accessed in the same way as data from IoT is assessed. The article explains the concept of Lokemon and its programming model. We report our evaluation of the effectiveness of Lokemon in a campus experiment that was performed for four weeks.

Chemotherapy-induced peripheral neuropathy (CIPN) is a common side effect of anticancer medicines that causes muscular weakness and falls in cancer patients. Therefore, we compared the methods for identifying and estimating the factors that cause symptoms in CIPN based on factor analysis and four different machine learning algorithms. The results of factor analysis showed that the latent factors causing CIPN can be classified into three categories: "Morning body condition", "Pain in the legs"and "Dizziness". We chose models for machine learning based on the premise of multi-task learning and built numerous machine learning models to estimate various CIPN symptoms. According to the results of the comparison studies, TabNet has the best generalization performance.

Attempts to facilitate and streamline systems architecting have resulted in a great number of reusable principles, practices, mechanisms, frameworks, and tools. Such a practice is the use of architectural viewpoints and views. However, as systems change, these practices should also evolve. The increasing scale and complexity of systems resulting from an ever-growing pool of human needs and breakthroughs may lead, in some cases, to an increased gap between the abstraction activities attempting to capture the whole of a system, and the instantiation activities that produce concrete and detailed descriptions of a system’s architecture. To address this issue, this article introduces a new notion, that of architectural glimpse statements, fundamental questions acting as the building blocks for architectural views and products. This notion can help architects ask the right questions in the right manner to create fundamental statements, the elaboration on which can lead directly to concrete architectural products. Working on top of standardized and common approaches, the article introduces a language for the creation of architectural glimpse statements using the 5W1H maxim. Based on this language, a tool and guidelines are also provided to facilitate the usage of glimpses. Finally, the overall methodology is demonstrated in two case studies.