Jasenka Dizdarevic, Technical University of Braunschweig
Internet of Things (IoT) is becoming more relevant with the connection of millions of everyday objects to the network, generating a large amounts of data, building smart cities and facilitating the daily life. With the IoT expansion, it becomes obvious that the traditional cloud computing model will not be enough to handle the amount of data generated by the increasing number of devices. This leads us to fog computing, a model in which the processing of data and applications is concentrated on devices at the edge of the network, instead of completely in the cloud. In this way the data can be processed locally in an intelligent device, so called fog node, thus relieving the cloud of part of the work. The mF2C project is focused on managing the seamless interoperability of fog, cloud and IoT.
One of the many questions that emerges is if these smart devices can be deployed to identify maintenance issues, anomalies, or disruptions before they occur. We know that the IoT environments allow us to obtain information in real time from the sources, but we still have the problem of interpretation and understanding of that data. We cannot afford to pre-program rules to deal with the infinite combinations of input data and situations that appear in the real world. Instead of doing that, there is a need for devices to be able to self-program, that is, that can learn from their own experience. This is where machine learning becomes necessary, our devices must be able to make sense and extract the meaning hidden behind bytes that move through the network. Since IoT nodes do not have the computing and storage resources to perform analytics, and cloud servers, on the other hand, are too far away to process data and respond in time, Machine Learning tasks should be left to fog nodes.
This technology aims to give learning capacity to devices by programming highly intelligent algorithms. This means that the new devices will have comprehension, learning, prediction, adaptation and potential capabilities to operate autonomously, and will be able to learn and change their future behavior. Algorithms will go through data sets sent from IoT nodes and deduce information about the properties of that data and this extracted information will allow them to make predictions about other future data. For example, in the case of the smart grid, the grid relies on an accurate demand forecast to optimize the energy consumption. A powerful automatic learning algorithm allows you to track, predict and learn about the efficient use of energy from the existing information. This is possible because almost all non-random data contains patterns, and those patterns allow the machine to generalize. In order to generalize, a model is trained with which the important aspects of the data are determined. As devices become more sophisticated in identifying changes, disruptions and potential problems with the equipment, they can be programmed to solve the problem in real time rather than sending alerts to the cloud.
The combination of computational power, advanced algorithms and data sets that arrive massively to feed various Machine Learning algorithms is leading to birth of a new technological era. The moment we live in today demands the convergence of the cloud computing, fog computing and IoT, as well as the exploration of the new emerging technological solutions (such as Machine Learning). All this is tackled by the mF2C project with the aim to create an interoperable fog-to-cloud framework.
Machine Learning in fog-to-cloud environment