Towards an IoT Platform with Edge Intelligence Capabilities

A specific feature of the IoT systems consists in continuously generated data by sensors and smart devices, which makes necessary enabling real time pre-processing and filtering as close as possible to their location. To cope with this specificity, the intermediary architectural level of Fog computing has been considered for this class of systems. The paper presents a solution for implementing this architectural extension to a current, cloud-oriented pilot IoT platform. The theoretical background on which this solution is based includes as its main topics the Fog computing, the Edge analytics and the Publish/subscribe interaction model. Based on their analysis, the architecture extension requirements specific to each topic are detailed. The adopted approach combines features and functionalities of both content-based and topic-based publish/subscribe models, with the aim to promote Edge Analytics principles by moving computation closer to where data resides and providing required performance for data-in-motion analysis.