Augmenting connectivity and device management with certain data analytics features helps you to bring your IoT projects to the next level. In principle, there are two major integration paths: (1) Get the data from the devices (i.e. the edges), transfer it to a backend, and run analytics jobs on top. (2) Run simple analytics jobs e.g. on the edges and transfer only the results to (device managing systems in) the backend. In the past months, we have explored these options by integrating some of the Bosch IoT Analytics functionalities with the ProSyst stack. In a first step, we have read data (i.e. OSGi messages) coming from specific gateways and managed by ProSyst’s mPRM. We experienced that in this set-up OSGi messages are not out-of-the-box usable in analytics, hence, several transformation steps are necessary, e.g. if values coming as separate OSGi events, like motion with x, y, z, they need to be recombined as a vector including deduced values like root mean square. The data was subsequently used in anomaly detection jobs implemented with the respective Bosch IoT Analytics cloud service. This analytics cloud service helps you analyze a set of devices and identify the individual anomalous ones, i.e. those sending implausible data or just behaving strange. By deploying the thus trained anomaly detection algorithms on edge devices and thus reducing the noise in the backend, the two above mentioned paths are covered and the roundtrip is complete. In this talk, we describe our experiments and show some challenges we met as well as possible solutions.
- About Us