Event driven and semantic based approach for data processing on IoT gateway devices

Internet of things (IoT) applications rely on networks composed of a set of heterogeneous sensors and smart devices, which have the capability to constantly monitor the surroundings and gather data. This heterogeneity is reflected in raw data collected by such type of systems. Additionally, these data are continuously streaming; thus leading to huge volumes of heterogeneous data, which are further transferred to centralized platforms for processing. Consequently, two main challenges have arisen. First, the heterogeneity aspect of IoT data makes high-level IoT applications' task of interpreting such data and detecting events in the real world more complex. Second, sending sensory data to a centralized platform leads to some issues, such as extensive consumption of IoT devices' limited resources, network traffic overloading, and latency, which might negatively impact the response time especially in systems that were designed to handle critical situations. In this paper, we propose a decentralized approach for IoT data processing, by delegating this task to distributed edge devices (Gateways) taking into consideration their limited resources and network bandwidth. To accomplish this, we proposed a two-layer data processing approach that employs a hyped model encompassed of complex event processing (CEP) and semantic web (SW) techniques. While the first is proposed for performing aggregation and classification tasks, we use the latter for performing semantic filtering and annotation tasks. We have evaluated the feasibility of our approach to process sensory data in the context of Air Quality Monitoring scenario using an experimentation involving established ontologies. Several benchmarks are considered such as overall runtime, data size, and response time.