Starting from the broader definition that the Internet of things is not a technology, but a complex ecosystem with a specific implication in the industry and at the same time a user-encapsulated interaction system, we can further describe its conceptual map as an interaction of people and objects as well as the individual interaction of objects within the IoT ecosystem. The existence of an ecosystem based on IoT technology is the link between these interactional elements that have the basis for interoperability – which uses data as a basic trigger and bearer interaction.
Considering the interaction process, it is necessary to start from the source of the data i.e. the place of its creation. This is happening at the object no matter whether the data producers are things – machines or people. In this context, we have the source as Physical Data and Machine data. Both sources, regardless of the different nature of existence, produce data at the level of the sensors and exchange them through a networked environment. Observing the form of these data, they can be static (produced data at the rest phase) and active i.e. generated in the dynamic process of changing the state of the objects or the movement of the living beings (data in motion).
At this level are the basic interactions of IoT elements. Although these data are considered to be very simple, they may in fact be very complex because they are built of various data formats, images and videos and very often as non-structured data. In this, raw data transformation is used to transform non-structured forms into a classified structured data. Transformed data, with the help of local processing units (more recently at Edge level), are processed close to the source itself. Thus, the data is immediately transmitted to the data element analysis process because it shortens the process of interaction from the original data and its interpretation by the end user. This, not so new concept in the interaction process, shortens the data path and avoids central storage of large amounts of data. After processing in Edge, the analyzed and processed data is sent to storage and prepared to release the report to the system user.
The process of interaction within the IoT system do not end here, but the process of interaction continues on the set of data available within the system. Namely, classified structured data is moving to the cloud. The first in a series of interaction processes is the Data accumulation. This is the first level where IoT data is actually stored. Data interaction at this level takes over the single cloud API. This is necessary because data production is a continuous cycle and it is necessary to have control and insight both in the newly created data and in the data archive. Data accumulation is in this process of interaction a method of storage that enables unlimited data quantities to be stored durably and economically.
Next to this, there is the process of data abstraction and data aggregation. These two processes interact together within the IoT system, although they sometimes need to be considered separately. Specifically, the data aggregation is introduced over IoT data elements as an abstraction that transforms the relationship between objects in a given level of data or data collection. Aggregated objects i.e. data collections are in fact real-world models of objects represented in digital form. This process is in the process of interaction within the IoT ecosystem very important. In most cases it represents the data that are processed by moving objects. Such objects are difficult to visualize because there are numerous cuts and overlaps between the trajectories when the interaction occurs and make the display very bursting and unreadable. For this reason, together with the data aggregation process, data abstractions are performed by mapping the control parameters, thus generating methods of generalization and simplification of the visual representation, which also enables the optimization of the monitoring and prediction of processes within the IoT ecosystem. This process is especially important if the data is further put into interaction on a semantics basis, i.e. actions intelligence process. This in fact presents behavior’s strategic approach to solving complex problems and relates to the interaction between problem and task solver. If all the data has been correctly prepared, this process can easily be done by converting information into intelligent data based on linked data and shared ontologies. This is a prerequisite for corporate intelligence solutions where IoT finds its place, especially in the management and analysis of industrial (IIoT) processes.
Interaction within the IoT ecosystem is further reflected in the fact that all new data is processed. Whether the data is processed by grouping new data and so the whole group is processed at a specific time (batch processing), or each data is processed immediately upon arrival in the system without waiting for the previous process to be executed (stream processing). The existence of differences in the way of functioning of these two processes has implications on the applicative part of the analysis system. It is necessary to take account of this process and the way of data interaction when determining the architecture of the system and the design of the resources. For example, systems that are designed with batch processing by definition work with delay and the system in the final instance cannon rely on consistent responses which is a very important element for the design of the IoT system. According to this choice, the analytics that is later performed with the help of the Cloud API or other user-defined applications may depend on the accuracy, the reliability of decision-making and the timeliness of reporting on the IoT ecosystem process.
Topics about IoT analytics, data visualization as well as other interactions – predictive detection processes, anomaly detection that take place in the IoT ecosystem will be described in the following separate topics.