Fog computing is an architectural style in which network components between devices and the cloud execute application-specific logic. We present the first review on fog computing within healthcare informatics, and explore, classify, and discuss different application use cases presented in the literature. For that, we categorize applications into use case classes and list an inventory of application-specific tasks that can be handled by fog computing. We discuss on which level of the network such fog computing tasks can be executed, and provide tradeoffs with respect to requirements relevant to healthcare. Our review indicates that: 1) there is a significant number of computing tasks in healthcare that require or can benefit from fog computing principles; 2) processing on higher network tiers is required due to constraints in wireless devices and the need to aggregate data; and 3) privacy concerns and dependability prevent computation tasks to be completely moved to the cloud. These findings substantiate the need for a coherent approach toward fog computing in healthcare, for which we present a list of recommended research and development actions.
For sustainable Internet of Things (IoT) systems, solar-power prediction is an essential element to optimize performance, allowing devices to schedule energy-intensive tasks in periods with excess energy. In regions with volatile weather, this requires taking the weather forecast into account. The problem is how to provide such solar energy predictions with high accuracy for large-scale IoT systems with various devices in an autonomous way, without manual adaptation effort. We present a detailed study on machine-learning approaches for the prediction of solar power intake for large scale IoT systems. We examine which machine learning models, feature sets and sampling rates gain the best results for a medium-term forecasting horizon. We also explore an operational setting in which devices are deployed without prior data and machine learning models are re-trained for each sensor continuously as a form of online learning. Our results show that prediction errors can be reduced by 20 % compared to the state of the art, despite strong weather volatility.
Research on autonomous management for large-scale deployments of constrained devices is still a maturing field in the Internet of Things (IoT). Although much research has been conducted on how to achieve autonomous management in specific cases, there is a need for literature investigating which mechanisms can achieve such behavior in a generalized way. In this review, we present a comprehensive and structured study of the mechanisms for autonomous device management of constrained IoT devices in the light of management tasks, operational environment, network topology, resource constraints, scalability and management categories. Data extracted from 32 relevant cases is first organized and analyzed according to a synthesized taxonomy of observed adaptation mechanisms, and then combined with state-of-the-art models of autonomous operations, identifying common patterns for autonomous management. Based on our findings we substantiate best practices for designing and implementing solutions around adaptation mechanisms. We then present a generalized model for autonomous device management that describes and explains the processes required for autonomous operation, unifying the insights from previous works as one cohesive archetype.
The future devices connected to the IoT must have the ability to solve their tasks intelligently, optimize their operations and adapt to changes autonomously. Machine learning, executed as part of device management, is the key to such intelligence. We discuss an evolution of device management systems, that also takes meta-knowledge into account, i.e., has the ability to manage the learning in the system. This is a step towards the vision of Cognitive IoT. As one important function for device management, we identified the selection among various prediction models. As an example, we use autonomous energy management of solar-powered sensor devices in an environment with volatile weather and seasonal changes. We investigate an algorithm that selects among a set of prediction models based on historic performance. Our results show that even though a classical machine learning model is able to learn quickly in some periods, it has trouble generalizing well over the data in other periods, compared to a simple physical model. We show that providing a system with the ability to select among a set of predictors can mitigate the bootstrapping problem for constrained devices and also help them stay in operation in periods when training data is missing.
Device management can enhance large-scale deployments of IoT nodes in non-stationary environments by supporting prediction and planning of their energy budget. This increases their ability for perpetual operation and is a step towards maintenance-free IoT. In this paper we consider how to accelerate the collection of relevant training data for nodes that are introduced into an existing deployment to increase the accuracy of their predictions. In particular, we investigate how nodes powered by solar energy can learn their energy intake faster and more accurately by using data from selected nodes that are working in similar conditions. We explore an architecture that utilizes different training data selection policies to manage the learning processes. For validation, we perform a case study to explore how nodes with correlated data can contribute to the learning process of other nodes. The obtained results indicate that this approach improves the accuracy of the predictions of a new node by 14%.